Self-Calibrated Ion-Selective Electrodes

Abstract

Ion-selective electrode (ISE) potentiometry is reliable only if on-site calibration using a standard solution is performed before ion measurements. The complex device and operation required for calibration hinder the implementation of ISEs in decentralized sensing. Reported herein is a new type of ISE that is calibrated by a built-in component of the sensor without requiring any fluid handling processes. The indicator and reference electrodes are connected by a thin ionic conductor such as an aqueous phase containing the measuring ions in a capillary tube. This connection establishes a baseline electromotive force (EMF) that incorporates phase boundary potentials across multiple interfaces of the electrochemical cell and serves as a one-point calibration. Unlike conventional ISEs that rely on one EMF reading for each measurement, the proposed sensor utilizes a sample-induced EMF change relative to the baseline for each ion measurement. The variability in relative EMF is found to be <2.0 mV for multiple full potentiometric sensors consisting of plasticizer-based K+ ISEs and hydrogel-based Ag/AgCl reference electrodes. This value is significantly smaller than the variability of absolute EMF readouts in similar sensors without the self-calibration design. Moreover, when the ion-conducting calibration bridge has a low concentration of primary ions, low ion mobility, and/or a small contact area with the indicator and reference phases, it does not compromise the Nernstian response slope toward the analyte ions in the sample and therefore does not need to be removed for sample testing. The accuracy of the single-use self-calibrated K+ sensor in testing undiluted human blood samples is validated using a commercial blood gas analyzer as the reference method. Although this study focuses on disposable sensors consisting of tubes, the fluidics-free self-calibration strategy may be adapted to other sensor configurations such as planar sensors.

Full text

Self-Calibrated Ion-Selective Electrodes
Nafesa Adil, Renjie Wang, and Xuewei Wang*
Cite This: Anal. Chem. 2023, 95, 11149−11156
Read Online
ACCESS Metrics & More Article Recommendations *
sı Supporting Information
ABSTRACT: Ion-selective electrode (ISE) potentiometry is
reliable only if on-site calibration using a standard solution is
performed before ion measurements. The complex device and
operation required for calibration hinder the implementation of
ISEs in decentralized sensing. Reported herein is a new type of ISE
that is calibrated by a built-in component of the sensor without
requiring any fluid handling processes. The indicator and reference
electrodes are connected by a thin ionic conductor such as an
aqueous phase containing the measuring ions in a capillary tube.
This connection establishes a baseline electromotive force (EMF)
that incorporates phase boundary potentials across multiple
interfaces of the electrochemical cell and serves as a one-point calibration. Unlike conventional ISEs that rely on one EMF
reading for each measurement, the proposed sensor utilizes a sample-induced EMF change relative to the baseline for each ion
measurement. The variability in relative EMF is found to be <2.0 mV for multiple full potentiometric sensors consisting of
plasticizer-based K+ISEs and hydrogel-based Ag/AgCl reference electrodes. This value is significantly smaller than the variability of
absolute EMF readouts in similar sensors without the self-calibration design. Moreover, when the ion-conducting calibration bridge
has a low concentration of primary ions, low ion mobility, and/or a small contact area with the indicator and reference phases, it
does not compromise the Nernstian response slope toward the analyte ions in the sample and therefore does not need to be
removed for sample testing. The accuracy of the single-use self-calibrated K+sensor in testing undiluted human blood samples is
validated using a commercial blood gas analyzer as the reference method. Although this study focuses on disposable sensors
consisting of tubes, the fluidics-free self-calibration strategy may be adapted to other sensor configurations such as planar sensors.
■INTRODUCTION
Ionophore-based ion-selective electrodes (ISEs) oer excep-
tional selectivity, high logarithmic linearity, rapid response
time, and excellent reversibility, making them ideal for
measuring electrolyte activities in biofluids such as blood,
plasma, and urine. They have achieved tremendous commer-
cial success in fully automated blood gas/electrolyte analyzers
and clinical chemistry analyzers used by healthcare profes-
sionals. In recent years, there has been a growing interest in
aordable and accessible electrolyte measurements in decen-
tralized locations. For instance, the development of home-use
electrolyte sensors capable of frequent testing of K+, Ca2+, and
Na+in capillary blood holds the potential to revolutionize the
self-management of chronic renal, heart, parathyroid, and
hypothalamus disorders.
1−4
Epidermal, transdermal, and
implanted electrolyte sensors for sweat and interstitial fluids
may permit real-time monitoring of electrolyte homeostasis to
guide timely interventions.
5−10
Since the composition and
fabrication of ISEs have been well established for decades, it
might seem that ISEs can be rapidly integrated into wearable,
implantable, and at-home sensing devices once appropriate
engineering is applied. However, despite significant engineer-
ing advancements,
5−10
the use of ISEs as accurate ion monitors
in decentralized settings has not been successful to the best of
our knowledge.
One major hurdle in utilizing ISEs for decentralized ion
monitoring is the need for on-site calibration. In commercial
handheld, benchtop, and laboratory electrolyte analyzers, each
ISE is calibrated using standard solutions with known analyte
activities before and/or between measurements. This calibra-
tion process involves complex instruments comprising standard
solutions, fluidic channels, and fluid control modules such as
pumps, actuators, and valves. Batch calibration and factory
calibration are possible approaches to eliminating the
mandatory individual calibration at the point of use. The
prerequisite for batch calibration is that all ISEs fabricated in
the same large batch have the nearly same calibration curves
(the standard potential, E°, and the response slope).
Electrolyte measurements for medical diagnostics often require
an electromotive force (EMF) error of no more than 1−2 mV.
Received: May 16, 2023
Accepted: July 3, 2023
Published: July 13, 2023
Articlepubs.acs.org/ac
© 2023 American Chemical Society 11149
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
Downloaded via VIRGINIA COMMONWEALTH UNIV on August 13, 2023 at 16:55:50 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

However, such a small variation is extremely challenging to
obtain in an EMF range spanning hundreds of mV. The EMF
measurement is susceptible to subtle variations at any
interfaces of the electrochemical cell involving the metal or
carbon electrode, the ion-to-electron transducer, the sensing
membrane, and the reference electrode components.
11,12
The
prerequisite for individual factory calibration is that the
electrodes do not undergo any changes during storage. Data
on the storage stability of ISEs are very limited, but aging of
any components, interpenetration between adjacent phases,
and exposure to ambient gases, moisture, and light can lead to
unpredictable EMF drifts.
13,14
The groups of Bakker, Bobacka,
and Yoshida explored coulometry and constant potential
coulometry as alternative readout modes of ion-selective
membranes with motivations including eliminating the
calibration process.
15−19
However, the authors noted limi-
tations such as fluid handling requirements and systematic
errors, and the coulometric response based on exhaustive ion
transfer may be hematocrit-dependent in whole blood tests.
In recent years, a wide variety of materials and chemicals
such as redox polymers/molecules, capacitive carbon materials,
nanostructured noble metals, metal−organic frameworks,
intercalation compounds, and their combinations have been
studied as the ion-to-electron transduction layer of solid-
contact ISEs with the aim of improving the electrode-to-
electrode consistency and other analytical performance
characteristics.
11,12
The deposited layer or suspension of
redox-active solid contact materials such as conducting
polymers may be further pre-polarized to unify standard
potentials of multiple electrodes.
20−25
Although a few mV or
less of standard deviations (SD) in E°have been observed
from the same batch of electrodes and occasionally from
dierent batches, the reported variation is only for indicator
electrodes as a shared reference electrode has been used in
most studies. Since the EMF variation of reference electrodes
is also a couple of mV or more,
11
full two-electrode
potentiometric sensors are unlikely to have medically accept-
able precision. Furthermore, it is important to consider that
many small E°SDs have been obtained after soaking the
electrodes in a standard solution or subjecting the electrodes to
electrochemical polarization for specific durations, typically
ranging from 1 to 48 h.
26
This lengthy and cumbersome
conditioning process required right before conducting
measurements is as impractical as the calibration process itself.
The Buhmman group designed paper-based full potentiometric
sensors that show an E°deviation of 3.2 mV in serum without
conditioning, but “inner filling solutions” need to be added to
complete the electrochemical cell at the point of use.
27
Recently, the pre-hydration of ion-selective membranes during
the electrode preparation step has been investigated as a
potential solution to mitigate the need for user-based electrode
conditioning.
25
ISEs conditioned in a KCl solution and stored
in a sealed package exhibit stable and reproducible response
without prolonged on-site conditioning, but sensors were
stored for only “at least 24 h”, and bulky electrodes were used.
Alternative methodologies aimed at reducing the conditioning
times of ISEs focus on special analyte ions or rely on complex
sensor arrangements that are not suitable for non-laboratory
applications.
28,29
Therefore, it remains challenging to prepare
simple, compact, and reliable ISE sensors that are free of
calibration and conditioning at the point of use.
Herein, we report a completely new design of potentio-
metric sensors to enhance the reproducibility of EMF
responses of ISEs. An ionic conductor such as an electrolyte
solution is used to connect the sensing membrane and
reference phase, which establishes an EMF baseline that
corrects for sensor-to-sensor variations. When a sample is
added, it induces a change in EMF, which is measured as the
potentiometric response. This approach diers significantly
from traditional potentiometry which relies on absolute EMF
readings without a concept of baseline. Notably, there is a
report on “self-calibrated” ISE,
30
but the principle of “self-
calibration” is completely dierent, and the sensor perform-
ance does not meet the requirements for decentralized
electrolyte monitoring (12 h conditioning; large EMF errors).
■EXPERIMENTAL SECTION
Reagents and Materials. Potassium ionophore I
(valinomycin), potassium tetrakis[3,5-bis-(trifluoromethyl)-
phenyl]borate (KTFPB), 2-nitrophenyl octyl ether (NPOE),
all of Selectophore grade, along with inorganic salts including
potassium chloride (KCl), sodium chloride (NaCl), lithium
chloride (LiCl), calcium chloride (CaCl2), magnesium chloride
(MgCl2), lithium acetate (LiOAc), and sodium phosphate
monobasic were purchased from MilliporeSigma. Low-
molecular-weight polyethylene glycol diacrylate (PEGDA, n
= approx. 9), amorphous fumed silica particles (surface
treatment with dimethyldichlorosilane, ∼325 mesh powder),
sodium poly(4-styrenesulfonate) (NaPSS), 3,4-ethylenediox-
ythiophene (EDOT), lithium phenyl-2,4,6-trimethylbenzoyl-
phosphinate, Ag wires (99.9%, 0.5 mm diameter), Au wires
(99.9%, 0.5 mm diameter), Pt wires (99.9%, 0.3 mm
diameter), and polymicro flexible fused silica capillary tubing
(1068150023; ID: 100 μm, OD: 360 μm) were purchased
from Fisher Scientific. HelixMark standard silicone tubing (60-
011-07; ID: 1.58 mm, OD: 2.41 mm) was obtained from
Freudenberg Medical. PTFE #30 AWG thin wall tubing
(06417-11; ID: 0.30 mm, OD: 0.76 mm) was purchased from
Cole-Parmer.
Heparinized blood specimens were obtained from the
Virginia Commonwealth University Medical Center. They
are leftover blood from the Blood Gas Lab and deidentified
before being collected by us. Heparinized human plasma was
purchased from Innovative Research.
Preparation of K+ISEs. The indicator electrode comprises
a PEDOT-coated Au wire inserted into a segment of plasticizer
in a 1 cm long silicone tube. The cleaned Au wire is coated
with PEDOT via galvanostatic polymerization in a three-
electrode electrochemical cell consisting of a Ag/AgCl
reference electrode and a Pt wire as the counter electrode.
31
The electrochemical cell contains a deaerated solution of 0.015
M EDOT and 0.1 M NaPSS. The electrosynthesis of PEDOT
film on the ∼1.5 cm end of a Au wire is performed at ∼0.2
mA/cm2current density for 1428 s. The PEDOT-coated Au
wire is air dried and stored in the dark overnight before use.
The K+-selective oil is NPOE containing 1.0 wt % potassium
ionophore I and 0.4 wt % KTFPB. The mixture is sonicated for
30 min to ensure complete dissolution of the sensing
chemicals. Hydrophobic fumed silica nanoparticles are further
mixed into the plasticizer solution at a ratio of 6.5% w/v to
enhance its viscosity and mechanical strength.
Preparation of Reference Electrodes. The reference
electrode comprises a Ag/AgCl wire coupled with a segment of
PEGDA hydrogel containing inorganic salts. The Ag/AgCl
wire is prepared via anodic electrodeposition of AgCl on a
cleaned Ag wire in a 0.1 M NaCl solution. A 1 cm long silicone
Analytical Chemistry pubs.acs.org/ac Article
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
11150

tube is filled with the low-molecular-weight PEGDA mixed
with 49.9 wt % aqueous solution and 1.0 wt % photoinitiator,
lithium phenyl-2,4,6-trimethylbenzoylphosphinate. The aque-
ous solution contains 0.5 M LiOAc, 0.1 M LiCl, and 1 mM
KCl unless otherwise specified. After the Ag/AgCl wire is
inserted into the liquid mixture in the silicone tube, the tube is
exposed to a planar UV light (Everbeam 365 nm 100 W UV
LED Black Light) for 30 s to obtain a photocured PEGDA-
based reference electrode.
Preparation of Self-Calibrated Sensors. When the
calibration phase is an aqueous solution, the solution is
injected into a fused silica capillary tube or a PTFE tube. Two
ends of the calibration tube are inserted into the plasticizer
phase and the uncured PEGDA solution, respectively, with a
depth of 2−3 mm on each side before photo-cross-linking is
initiated for the reference electrode. When the calibration
phase is an aqueous solution with PEGDA, it is first
photocured in a calibration tube before the tube is inserted
into the plasticizer phase and the reference solution. Then, the
reference solution is photocured to obtain a full self-calibrated
sensor. The length of the calibration tube is ∼1 cm for all self-
calibrated sensors. All sensors are tested without any
conditioning.
EMF Measurements. All EMF measurements are carried
out at room temperature using an EMF-16 Precision
Electrochemistry EMF Interface (Lawson Labs Inc.). For
ion-selective sensors without the calibration bridge, the EMF is
measured after a drop of aqueous sample is added into the gap
between two silicone tubes. For ion-selective sensors with the
calibration bridge, the EMF is measured before and after the
addition of a drop of aqueous sample. A piece of Parafilm is
always used underneath the sensor to prevent spreading of the
aqueous solution. A Metrohm double-junction Ag/AgCl
reference electrode is used as the reference electrode to test
the performance of the PEGDA-based reference electrodes.
The commercial and home-made reference electrodes are
manually dipped into 2 mL of dierent solutions to test the
EMF response of the PEGDA reference electrode to salts.
Liquid junction potential is calculated according to the
stationary Nernst−Planck equation
32
using LJPcalc software
(https://swharden.com/LJPcalc). Activity coecients are
calculated using the Debye−Huckel equation. For storage
stability tests, self-calibrated sensors with the optimal reference
and calibration composition are stored in a sealed two-layer
plastic container with water in the bottom layer and sensors on
the top layer. Response of fresh sensors and sensors after 3 and
6 weeks of storage is tested toward 10 mM KCl.
■RESULTS AND DISCUSSION
Calibration Capability of the Ionic Conductor
Integrated between the Indicator and Reference
Electrodes. PEDOT-based solid contact ISEs are used as an
example to demonstrate the concept of self-calibrated
potentiometric sensors because conducting polymer is one of
the most commonly used solid contact materials. Potassium
ionophore I and KTFPB are dissolved in NPOE via sonication
without using any other solvents like tetrahydrofuran. The
NPOE solution is further mixed with 6.5% w/v hydrophobic
fumed silica particles as a thickening agent. The resulting
NPOE−silica particle mixture is still easy to be transferred into
the silicone tube via pipetting to create an ISE along with a
PEDOT-coated Au wire, but the enhanced viscosity renders
this “oil” phase mechanically more stable. The reference
electrode is a Ag/AgCl wire inserted in a photo-cross-linked
PEGDA hydrogel containing chloride ions. There is no
microporous frit between the reference electrode and the
sample, but the high rigidity of the solidified PEGDA hydrogel
minimizes the mixing of the reference electrolyte and the
sample (see below for optimization of the reference electrode).
Figure 1A shows the setup of a tube-type K+sensor. Figure 1B
shows the EMF reading of n= 5 sensors toward 1 mM KCl as
the sample. Not surprisingly, the EMF reading of the full
potentiometric sensor varies. The SD is 40.5 mV, which is
consistent with the sensor-to-sensor variability of similar full
sensors in a previous report.
9
The inconsistency is because the
constituent conductors and interfaces such as the solid contact
layer and the AgCl layer vary among multiple sensors even
though they are identically fabricated.
11
Figure 2A,B shows a prototype of the self-calibrated sensor
using the tube-type design. The ion-selective oil and reference
hydro gel are connected by a fused silica capillary tube with an
ID of 0.10 mm and an OD of 0.36 mm. The tube has been
filled with a solution containing the analyte ions (0.1 M KCl in
this example). With this ionically conducting bridge, an EMF
baseline can be first obtained in the absence of a sample, which
is a unique feature of the proposed self-calibrated sensor as
opposed to all previous potentiometric sensors. Then, a drop
of aqueous sample is added to the space between two large
silicone tubes to cover the exposed surface of the plasticizer
phase in the indicator electrode and the hydrogel phase in the
reference electrode. The EMF change upon sample addition is
recorded and used for the quantification of K+in the sample.
Figure 2C shows the baseline EMF and the EMF change after
the addition of 12 μL of 10−3M KCl. The baseline is further
normalized to zero, as shown in Figure 2D, to aid in
visualization of the relative EMF. Although the original EMF
baseline varies by tens of mV for n= 5 sensors, the SD of the
EMF change relative to the baseline (ΔEMF) is only 1.7 mV.
All EMF variabilities from the electron conductors, solid
contact layer, AgCl layer, and their interfaces are included in
this baseline and compensated when ΔEMF is used for the
response. The inconsistency in ΔEMF will only result from the
variabilities in the interfaces between the two aqueous
solutions (calibration solution and sample) and the two
liquid-based electrode phases (plasticizer and hydrogel). This
inconsistency turns out to be much smaller probably because
electric potential dierences across interfaces between ionic
conductors are well defined by the phase boundary potential
model and the liquid junction potential theory.
Figure 1. (A) Photograph of a tube-type potentiometric sensor
consisting of a Ag/AgCl reference electrode and an ionophore-based
solid-contact ISE. Two tubes are silicone tubes to accommodate the
hydrogel and the plasticizer phase for the reference and indicator
electrodes, respectively. The aqueous sample at a volume of 12 μL is
added in between two silicone tubes. (B) EMF readings of n= 5
sensors to 10−3M KCl.
Analytical Chemistry pubs.acs.org/ac Article
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
11151

Figure 3 shows the stepwise EMF response of the non-
calibrated and self-calibrated sensors when the KCl concen-
tration in the sample is increased from 10−4to 10−1M. The
response slope of the self-calibrated sensors is 57.4 mV/
decade, which is nearly identical to the slope of 57.8 mV/
decade for sensors without the calibration tube, suggesting that
the presence of an appropriate calibration bridge does not
compromise the EMF response of the sample while serving for
the calibration purpose. The EMF baseline established by the
calibration bridge is similar to the one-point calibration
performed in the cartridge of the Abbot i-STAT blood gas/
electrolyte analyzer. The dierence is that the calibration phase
in the present design does not need to be delivered or removed
since it is permanently integrated in the sensor.
Optimization of the Calibration Bridge. In our sensor
arrangement, one side of the ion-selective oil and the reference
hydrogel are simultaneously exposed to two aqueous phases
after the sample addition. To the best of our knowledge,
similar designs have not been previously reported. An ISE or
perhaps any indicator electrode of an electrochemical sensor is
exposed directly to one external solution (usually a calibration
solution or a sample) at one time. One immediate question
about the new design is that how the presence of a calibration
phase aects the potentiometric response of the sample. Figure
4shows EMF baselines created by dierent calibration phases
and EMF changes after the addition of samples containing
10−3, 10−2, and 10−1M KCl. When the calibration tube has an
ID of 0.30 mm and an OD of 0.76 mm and the calibration
solution in this tube is 0.1 M KCl (Figure 4A), there are
obvious upward EMF drifts after the addition of a KCl solution
that is less concentrated than the calibration solution. In
designing low-detection-limit ISEs based on classical planar
polymeric membranes, the groups of Erno Pretsch and Eric
Bakker studied the transmembrane ion flux from the inner
filling solution with a higher primary ion concentration to the
sample without primary ions.
33−35
Due to this chemical
gradient-induced ion flux, the sample solution has an elevated
primary ion concentration in the surface layer adjacent to the
ion-selective membrane than the bulk. Since the ion-selective
membrane senses the activity of primary ions in the interface
zone of the aqueous sample, the EMF response appears greater
than estimated based on the ion activity of the bulk sample
solution. In our self-calibrated sensor configuration, the ion-
selective oil is also exposed to two solutions with possibly
dierent ion concentrations although both solutions are at the
same side. Therefore, there should be a similar flux of primary
ions down the concentration gradient. In other words, the
concentrated calibration solution contaminates the sample in
our configuration similarly as the concentrated inner filling
solution contaminates the sample in a classical polymeric
membrane ISE. As EMF is proportional to the logarithm of the
primary ion activity according to the Nernstian equation, the
EMF increase caused by the contamination of a low-
concentration solution is greater than the EMF decrease
caused by the depletion of the concentrated solution. As a
result, upward EMF drifts are observed in Figure 4A upon
addition of a sample with a K+concentration low than that in
the calibration tube.
Three strategies are examined to suppress the observed
sample contamination and the resulting EMF drift. First, as
shown in Figure 4B, 10−3M instead of 10−1M KCl as the
calibration solution does not contaminate samples and
therefore cannot induce significant EMF drifts. A slight upward
drift for the 10−1M KCl sample is likely caused by K+
contamination in the opposite direction (from the sample to
the calibration solution). As the electrolyte concentration in
real samples especially body fluids usually have a narrow and
known range, using a calibration solution containing a
comparable concentration of the measuring ion should easily
minimize contamination and EMF drifts. Second, when the
contact area between the ion-selective oil and the calibration
Figure 2. (A) Photographs of a self-calibrated potentiometric sensor
before and after the sample addition. (B) Surface of the ion-selective
oil that will be in contact with the sample. The fused capillary tube
with the calibration solution is inserted into the oil. (C) Original EMF
traces of n= 5 self-calibrated sensors before and after adding 12 μL of
10−3M KCl as the sample. (D) EMF traces with all baseline values
normalized to zero. Black arrows in C and D indicate the addition of
the sample.
Figure 3. EMF traces and corresponding calibration curves of the
potentiometric sensors without (A,B) and with (C,D) calibration
bridge. The concentration denotes that of KCl in the sample. Each
sample is tested for 30 s and carefully removed before a new sample
with a higher KCl concentration is added. N= 3 sensors are tested for
each type.
Analytical Chemistry pubs.acs.org/ac Article
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
11152

solution is much smaller than that between the oil and the
sample, ion contamination caused by the calibration solution
becomes less obvious. As can be seen from Figure 4C, EMF
drifts are unnoticeable when the calibration tube ID is
decreased to 0.1 mm although the calibration solution still
has a high concentration of KCl. The oil-calibration solution
contact area is only 0.4% of the oil-sample contact area (0.008
vs 1.889 mm2), and the volume of the calibration solution is
150 times lower than the sample volume (0.08 vs. 12 μL). In
contrast, the oil-calibration solution contact area is 0.071 mm2
(4.6% of the oil-sample contact area), and the calibration
solution volume is 0.76 μL (vs 12 μL sample) when the ID of
the calibration tube is 0.3 mm (Figure 4A). A capillary tube
with a tiny amount of salt solution is still able to complete the
electrochemical cell to create a baseline, but it does not impair
the sample response. Third, the EMF drift is reduced when the
calibration phase is solidified by photo-cross-linked PEGDA
(Figure 4D vs 4A while using the same concentration of KCl in
the 0.3 mm ID tube). Chemical diusion in the PEGDA
hydrogel is much lower than that in an aqueous solution,
36
which suppresses the ion flux from the calibration phase to the
sample. If the KCl concentration in PEGDA is reduced to 10−3
M, the EMF drift is eliminated (Figure 4E). Although both
liquid (Figure 4B,C) and hydrogel (Figure 4E) calibration
phases enable drift-free EMF responses in these tube-type
sensors, solidified hydrogel is more practical for other sensor
configurations such as mass-producible planar sensors for long-
term goals. Therefore, we will focus more on the photocured
PEGDA calibration phase.
Classical response theories of ISEs such as the phase
boundary potential model and the Nernst−Planck and Poisson
equation were developed for well-defined, flat membranes with
only one-dimensional ion gradients and electric potential
profiles perpendicular to the membrane surface.
37,38
In the self-
calibrated sensors, the presence of a common calibration phase
with a fixed ion concentration may compromise the response
to the analyte ions in the sample. Interestingly, the Nernstian
response slope is maintained when the calibration phase has a
low concentration of primary ions (Figure 4B,E) or a small
contact area with the ion-selective oil (Figure 4C). Before
application of the sample, the plasticizer-based sensing oil and
the hydrogel-based reference phase are connected by an
aqueous phase containing primary ions, which is analogous to a
regular one-point calibration process of an ISE. After
application of a sample that has a dierent composition from
the calibration phase, the phase boundary potential across the
sample-plasticizer interface and the liquid junction potential
across the sample-hydrogel interface will dier from those for
the calibration phase. As a result, there will be a current flowing
in the closed loop of plasticizer |calibration phase |reference
hydrogel |sample |plasticizer. Since the resistance of the
calibration phase is much higher than that of the sample when
Figure 4. EMF traces of self-calibrated K+-selective sensors before and after the addition of 12 μL of 10−3, 10−2, or 10−1M KCl as the sample.
Arrows indicate the sample addition. The EMF change relative to the baseline (ΔEMF after 400 s) for each KCl sample is labeled as average ±SD
for n= 3 sensors. The EMF dierence between 10−3and 10−2M as well as 10−2and 10−1M KCl is labeled on the right to indicate the response
slope.
Analytical Chemistry pubs.acs.org/ac Article
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
11153

the calibration phase is in a narrow-bore capillary and/or has a
low ionic strength, the Ohmic drop is large in the calibration
bridge but insignificant in the sample. Therefore, the EMF of
the electrochemical cell can be calculated according to the
plasticizer-sample-reference hydrogel pathway. This is like a
regular ISE measurement, for which the EMF response is
governed by the primary ion activity in the sample. The same
cell EMF should be obtained from the plasticizer-calibration
phase-reference hydrogel pathway because it includes the
significant Ohmic drop in the calibration phase. In summary,
an appropriately formulated and sized calibration phase causes
neither EMF drifts nor response slope reductions while
providing a baseline for one-point calibration.
Optimization of the Reference Electrode. When the
calibration phase and the reference phase have dierent
compositions, chemical diusion between two aqueous phases
may happen during storage and impair the reliability of self-
calibration. Therefore, it is best to employ the same
composition for the calibration and reference phases. They
are based on a PEGDA hydrogel instead of an aqueous
solution because the hydrogel is preferred for the calibration
phase and necessary to create a reference electrode that does
not mix with the sample. The hydrogel should contain the
analyte ion, K+, for the calibration purpose, but the K+
concentration should remain low to prevent contamination
to the sample. Also, a high concentration of Cl−is needed in
the hydrogel to maintain a stable potential of the Ag/AgCl
reference electrode. We systematically tested the performance
of reference electrodes with dierent PEGDA-based hydrogel
formulations to find one that meets the aforementioned
requirements and exhibits most stable EMF as the sample
composition changes. Figure 5 shows the EMF response of
dierent reference electrodes to NaCl and KCl against a
commercial double junction Ag/AgCl reference electrode. A
high concentration of KCl or NaCl in the hydrogel makes the
electrode more responsive to KCl or NaCl, respectively. The
use of 0.1 M LiCl as the chloride salt reduces the EMF
response toward both KCl and NaCl. Since the calibration
phase needs a low concentration of K+, a combination of 0.1 M
LiCl and 1 mM KCl is reasonable. The addition of LiOAc, a
commonly used equitransferent salt in reference electrodes,
further reduces the EMF response especially toward NaCl.
Therefore, the optimal formulation of our PEGDA-based
reference phase has 1 mM KCl, 0.1 M LiCl, and 0.5 mM
LiOAc. This Ag/AgCl reference electrode has an EMF
response of −2.7 mV when NaCl is increased from 10−2to
10−1M. Since the commercial double-junction reference
electrode using 1 M LiOAc as the bridge electrolyte has a
calculated liquid junction potential of −2.9 and 0.3 mV for
10−2and 10−1M NaCl, respectively,
32
the observed EMF
change (Figure 5B, magenta line) is largely due to the response
of the commercial reference electrode instead of the PEGDA-
based reference electrode. Accordingly, this PEGDA reference
electrode is unlikely to be sensitive to Na+fluctuations in most
biological samples. Similarly, the junction potential for 10−3,
10−2, and 10−1M KCl is −4.6, −3.2, and −1.4 mV,
respectively, which accounts for a large portion of the observed
KCl response in Figure 5A (magenta line) and thereby
suggests low sensitivity of the PEGDA-based reference
electrode to KCl fluctuations. Moreover, the EMF of the
PEGDA-based reference electrode remains constant when it is
transferred from simple NaCl solutions in DI water to plasma
and blood (Figure 6), indicating that there are no matrix
eects from these biological samples.
Traditional Ag/AgCl reference electrodes including those
using hydrogels are separated from the sample via a pinhole, a
capillary, a microporous junction, or a polymer layer to
minimize mixing of the reference electrolyte and the
sample.
39−41
In our design, the hydrogel in the silicone tube
is in direct contact with the sample, but the mixing is
suppressed by the cross-linked PEGDA network. Unlike
traditional hydrogels such as agar, the hydrogel formed from
photo-cross-linking of low-molecular-weight PEGDA is rigid
with a mesh size is only about 1−2 nm,
36
which is even smaller
than the pore size of most junctions used in conventional
reference electrodes.
42
While the aqueous solution trapped in
the PEGDA network does not quickly mix with the aqueous
sample, ions can diuse across the interface, generating liquid
junction potentials. The ion mobility in PEGDA diers from
that in simple aqueous solutions because a portion of water
molecules are bound to polymers in hydrogels.
43
The ethylene
glycol structure with electronegative O is also expected to bind
to cations to reduce their mobility. As a result, it is hard to
calculate the liquid junction potential of each hydrogel
formulation to explain the EMF response observed in Figure
5. Since the focus of this work is the novel self-calibration
concept based on the ion-conducting bridge, the detailed
working mechanism of the PEGDA-based reference electrode
will be explored in future work.
Figure 5. EMF response of dierent tube-type reference electrodes to
KCl (A) and NaCl (B) against a Metrohm double-junction Ag/AgCl
reference electrode. The examined reference electrodes comprise Ag/
AgCl wires coupled with PEGDA hydrogels containing dierent salts.
Figure 6. EMF response of the optimal PEGDA-based Ag/AgCl
reference electrode in dierent aqueous solutions and body fluids
against a commercial double junction Ag/AgCl reference electrode.
Analytical Chemistry pubs.acs.org/ac Article
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
11154

Reproducibility, Blood Tests, and Storage Stability.
The electrodes have been assumed to be single-use in
optimizing the calibration phase and the reference electrode.
This is because home-use sensors for body fluids such as blood
are supposed to be disposable, and at-home electrolyte
measurements using capillary blood are an urgent unmet
need for the management of chronic diseases. Since blood K+
concentrations fall between 2 and 8 mM with a normal range
of 3.5−5.5 mM, a calibration curve of the optimal self-
calibrated K+-selective sensor over 1−10 mM K+is constructed
in Figure 7A. All standard solutions have an ionic strength of
approximately 0.16 M and a pH of 7.4 to mimic the blood
composition. Notably, we did not attempt to calculate the
theorical ΔEMF since the external calibration method is more
convenient and reliable. The reproducibility of the relative
EMF response of multiple sensors is further confirmed for 1
and 10 mM KCl (n= 10 sensors for each concentration). As
can be seen from Figure 7B, the SD for 1 and 10 mM KCl is
only 1.4 and 1.5 mV, respectively. According to the Nernstian
equation, a 1.5 mV SD in EMF corresponds to a concentration
SD of 0.3 mM for 5.0 mM K+, which is below the acceptable
K+measurement error of 0.5 mM according to the U.S. Code
of Federal Regulations.
44
Since TFPB in the plasticizer is in its
K+salt form prior to the sensor assembly, no K+diusion from
the calibration phase to the plasticizer is expected. Indeed,
when self-calibrated sensors are used 2 h after the assembly,
the EMF response to 1 and 10 mM KCl is −27.8 and +27.1
mV, respectively, matching that of fresh sensors. The self-
calibrated sensor is reversible (Figure S1), but its stability
during multiple and continuous use needs to be demonstrated
with modified sensor designs when wearable, implantable, and
in-field monitoring applications are pursued in future work.
Using the calibration curve in Figure 7A, the self-calibrated
K+sensors are evaluated in seven human blood specimens
using an Abbott i-STAT blood gas analyzer as the reference
technology. As shown in Table 1, the percent error ranges from
2.5 to 10.8%, indicating high accuracy of our sensors even
though they are manually assembled. We noticed that the K+
concentration determined by the self-calibrated sensor is
always higher than that from the i-STAT analyzer. This
positive deviation might be because of some systematic errors
that are not identified such as errors of pipettes and balance
used to prepare standard solutions. Furthermore, we
conducted a preliminary study on the storage stability of the
self-calibrated sensors. Since hydrogel in the reference and
calibration phases may evaporate, sensors are stored in
humidified air created in a sealed container. The response of
10 mM KCl on fresh sensors, sensors after 3 week storage, and
sensors after 6 week storage is 27.1 ±1.2, 28.0 ±0.6, and 28.1
±1.1 mV, respectively. Each sensor is used for a single
measurement, and n= 3 sensors are tested at each time point.
This preliminary result is promising, and the storage stability
may be further enhanced by utilizing professional packaging
with more accurate humidity control.
■CONCLUSIONS
A potentiometric sensor possesses an EMF baseline when the
indicator and reference electrodes are linked through an ionic
conductor. This integrated ionic conductor enables one-point
calibration without compromising the response of the sample.
Unlike traditional on-site calibration processes, this new
calibration strategy does not require any fluid handling
modules to deliver and clean calibration solutions. Design of
planar ion sensors and multiplexed ion sensors with this self-
calibration capability is underway in our laboratory. Other
conductors such as conjugated polymers, carbon materials, and
metals may also connect the indicator and reference electrodes
to allow for the establishment of a baseline EMF. The storage
and fabrication of these materials could be easier than aqueous
solutions and hydrogels although their interfacial potentials
may be harder to control. More electrochemical readout
modes beyond open-circuit potential are also being explored
on the ion-selective sensors with an integrated bridge in our
laboratory.
■ASSOCIATED CONTENT
*
sı Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.analchem.3c02135.
Selectivity coecients of the self-calibration potassium
ISE and reversibility of the self-calibration potassium ISE
(PDF)
Figure 7. (A) Calibration curve of the optimal self-calibrated K+
sensor using identical PEGDA-based calibration and reference phases.
Each sensor is used one time, and the EMF change relative to the
baseline after 1 min is calculated as the ΔEMF. All standard KCl
solutions are prepared in 150 mM NaCl, 1 mM CaCl2, 1 mM MgCl2,
and 2 mM phosphate buer at pH 7.4. N= 3 sensors are tested for
each concentration. (B) Reproducibility of the self-calibrated sensor
in testing 10−3and 10−2M KCl, respectively. The baseline EMF is
normalized to zero for all sensors, and the relative EMF is shown as
average ±SD for n= 10 sensors.
Table 1. K+Concentration of Seven Human Blood Samples
Determined by the Self-calibrated Sensors and the Abbott i-
STAT Blood Gas Analyzer Equipped with CHEM8+
Cartridges
blood 1 2 3 4 5 6 7
this work (mM) 3.6 4.1 4.6 3.7 4.1 5.2 3.7
i-STAT (mM) 3.4 4.0 4.2 3.5 3.7 5.0 3.5
% error 5.9 2.5 9.5 5.7 10.8 4.0 5.7
Analytical Chemistry pubs.acs.org/ac Article
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
11155

■AUTHOR INFORMATION
Corresponding Author
Xuewei Wang −Department of Chemistry, Virginia
Commonwealth University, Richmond, Virginia 23284,
United States; orcid.org/0000-0001-9604-3045;
Email: wangx11@vcu.edu
Authors
Nafesa Adil −Department of Chemistry, Virginia
Commonwealth University, Richmond, Virginia 23284,
United States
Renjie Wang −Department of Chemistry, Virginia
Commonwealth University, Richmond, Virginia 23284,
United States; Present Address: Department of Chemistry
and Biochemistry, Florida Atlantic University, 777 Glades
Road, Boca Raton, FL 33431, United States
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.analchem.3c02135
Author Contributions
The manuscript was written through contributions of all
authors. All authors have given approval to the final version of
the manuscript.
Notes
The authors declare no competing financial interest.
■ACKNOWLEDGMENTS
This work is financially supported by the Virginia Common-
wealth University (Startup Grant for X.W.) and Virginia
Innovation Partnership Corporation (VIPC; CCF23-0065-
HE). We thank reviewer 1 for suggesting the theory involving
dierent Ohmic drops in the sample and calibration phase.
■REFERENCES
(1) Brandi, M. L.; Bilezikian, J. P.; Shoback, D.; Bouillon, R.; Clarke,
B. L.; Thakker, R. V.; Khan, A. A.; Potts, J. T., Jr. J. Clin. Endocrinol.
Metab. 2016,101, 2273−2283.
(2) Chang, A. R.; Sang, Y.; Leddy, J.; Yahya, T.; Kirchner, H. L.;
Inker, L. A.; Matsushita, K.; Ballew, S. H.; Coresh, J.; Grams, M. E.
Hypertension 2016,67, 1181−1188.
(3) Nilsson, E.; Gasparini, A.; A
rnlöv, J.; Xu, H.; Henriksson, K. M.;
Coresh, J.; Grams, M. E.; Carrero, J. J. Int. J. Cardiol. 2017,245, 277−
284.
(4) Pabich, S.; Flynn, M.; Pelley, E. J. Endocr. Soc. 2019,3, 882−886.
(5) Gao, F.; Liu, C.; Zhang, L.; Liu, T.; Wang, Z.; Song, Z.; Cai, H.;
Fang, Z.; Chen, J.; Wang, J.; Han, M.; Wang, J.; Lin, K.; Wang, R.; Li,
M.; Mei, Q.; Ma, X.; Liang, S.; Gou, G.; Xue, N. Microsyst. Nanoeng.
2023,9, 1.
(6) Heikenfeld, J.; Jajack, A.; Feldman, B.; Granger, S. W.; Gaitonde,
S.; Begtrup, G.; Katchman, B. A. Nat. Biotechnol. 2019,37, 407−419.
(7) Zhao, J.; Guo, H.; Li, J.; Bandodkar, A. J.; Rogers, J. A. Trends
Chem. 2019,1, 559−571.
(8) Huang, X.; Zheng, S.; Liang, B.; He, M.; Wu, F.; Yang, J.; Chen,
H.; Xie, X. Microsyst. Nanoeng. 2023,9, 25.
(9) Gao, W.; Emaminejad, S.; Nyein, H.; Challa, S.; Chen, K.; Peck,
A.; Fahad, H. M.; Ota, H.; Shiraki, H.; Kiriya, D.; Lien, D.; Brooks, G.
A.; Davis, R. W.; Javey, A. Nature 2016,529, 509−514.
(10) Parrilla, M.; Cuartero, M.; Crespo, G. A. TrAC, Trends Anal.
Chem. 2019,110, 303−320.
(11) Rousseau, C. R.; Buhlmann, P. TrAC, Trends Anal. Chem. 2021,
140, 116277.
(12) Cheong, Y. H.; Ge, L.; Lisak, G. Anal. Chim. Acta 2021,1162,
338304.
(13) Hu, J.; Stein, A.; Buhlmann, P. TrAC, Trends Anal. Chem. 2016,
76, 102−114.
(14) Michalska, A.; Wojciechowski, M.; Bulska, E.; Maksymiuk, K.
Talanta 2010,82, 151−157.
(15) Grygolowicz-Pawlak, E.; Bakker, E. Electrochem. Commun.
2010,12, 1195−1198.
(16) Shvarev, A.; Neel, B.; Bakker, E. Anal. Chem. 2012,84, 8038−
8044.
(17) Vanamo, U.; Hupa, E.; Yrjänä, V.; Bobacka, J. Anal. Chem.
2016,88, 4369−4374.
(18) Han, T.; Mattinen, U.; Bobacka, J. ACS Sens. 2019,4, 900−906.
(19) Tatsumi, S.; Omatsu, T.; Maeda, K.; Mousavi, M. P. S.;
Whitesides, G. M.; Yoshida, Y. Electrochim. Acta 2022,408, 139946.
(20) He, N.; Papp, S.; Lindfors, T.; Höfler, L.; Latonen, R. M.;
Gyurcsányi, R. E. Anal. Chem. 2017,89, 2598−2605.
(21) Vanamo, U.; Bobacka, J. Electrochim. Acta 2014,122, 316−321.
(22) Vanamo, U.; Bobacka, J. Anal. Chem. 2014,86, 10540−10545.
(23) Papp, S.; Kozma, J.; Lindfors, T.; Gyurcsányi, R. E.
Electroanalysis 2020,32, 867−873.
(24) Bahro, C.; Goswami, S.; Gernhart, S.; Koley, D. Anal. Chem.
2022,94, 8302−8308.
(25) Kozma, J.; Papp, S.; Gyurcsányi, R. E. Anal. Chem. 2022,94,
8249−8257.
(26) Conditioning times of dierent electrodes in Table 1 of ref 12:
ref 91 (48 h), 142 (overnight), 104 (overnight), 145 (no), 117 (1 h),
64 (1 h), 113 (not specified), 62 (24 h), 124 (24 h), 126 (1 day), 112
(16 h), 92 (>1 h), 83 (>12 h), 87 (not specified), 131 (overnight),
109 (1 h), 84 (not specified), 63 (2 days), 66 (>6 h).
(27) Hu, J.; Stein, A.; Buhlmann, P. Angew. Chem., Int. Ed. 2016,55,
7544.
(28) Damala, P.; Zdrachek, E.; Forrest, T.; Bakker, E. Anal. Chem.
2022,94, 11549−11556.
(29) Rich, M.; Mendecki, L.; Mensah, S. T.; Blanco-Martinez, E.;
Armas, S.; Calvo-Marzal, P.; Radu, A.; Chumbimuni-Torres, K. Y.
Anal. Chem. 2016,88, 8404−8408.
(30) Wang, F.; Liu, Y.; Zhang, M.; Zhang, F.; He, P. Anal. Chem.
2021,93, 8318−8325.
(31) Guzinski, M.; Jarvis, J. M.; Perez, F.; Pendley, B. D.; Lindner,
E.; De Marco, R.; Crespo, G. A.; Acres, R. G.; Walker, R.; Bishop, J.
Anal. Chem. 2017,89, 3508−3516.
(32) Marino, M.; Misuri, L.; Brogioli, D. A New Open-Source
Software for the Calculation of the Liquid Junction Potential Between
Two Solutions According to the Stationary Nernst-Planck Equation.
2014, arXiv preprint 1403.3640.
(33) Sokalski, T.; Ceresa, A.; Zwickl, T.; Pretsch, E. J. Am. Chem.
Soc. 1997,119, 11347−11348.
(34) Mathison, S.; Bakker, E. Anal. Chem. 1998,70, 303−309.
(35) Szigeti, Z.; Vigassy, T.; Bakker, E.; Pretsch, E. Electroanalysis
2006,18, 1254−1265.
(36) Cruise, G. M.; Scharp, D. S.; Hubbell, J. A. Biomater 1998,19,
1287−1294.
(37) Bobacka, J.; Ivaska, A.; Lewenstam, A. Chem. Rev. 2008,108,
329−351.
(38) Bakker, E.; Buhlmann, P.; Pretsch, E. Talanta 2004,63, 3−20.
(39) Ciobanu, M.; Wilburn, J. P.; Buss, N. I.; Ditavong, P.; Lowy, D.
A. Electroanalysis 2002,14, 989−997.
(40) Ha, J.; Martin, S. M.; Jeon, Y.; Yoon, I. J.; Brown, R. B.; Nam,
H.; Cha, G. S. Anal. Chim. Acta 2005,549, 59−66.
(41) Liao, W. Y.; Chou, T. C. Anal. Chem. 2006,78, 4219−4223.
(42) Anderson, E. L.; Troudt, B. K.; Buhlmann, P. Anal. Sci. 2020,
36, 187−191.
(43) Gun’ko, V. M.; Savina, I. N.; Mikhalovsky, S. V. Gels 2017,3,
37.
(44) Laboratory Requirements, Code of Federal Regulations,
Section 493.931, Title 42, 2003.
Analytical Chemistry pubs.acs.org/ac Article
https://doi.org/10.1021/acs.analchem.3c02135
Anal. Chem. 2023, 95, 11149−11156
11156