Determination of the Selectivity of Printed Wearable Sweat Sensors

Alicia Zörner

, Susanne Oertel

, Björn Schmitz

, Nadine

Lang

Michael P. M. Jank

and Lothar Frey

Fraunhofer Institute for Integrated Systems and Device Technology IISB, Schottkystr. 10, 91058 Erlangen, Germany

Fraunhofer Institute for Integrated Circuits IIS, Am Wolfsmantel 33, 91058 Erlangen, Germany

Keywords: Screen-printed, Biosensor, Sweat, Ammonium, Exercise, Ion-Selective Electrode, Wearable Device,

Bluetooth, Selectivity, Wearable Healthcare.

Abstract: The characterization and system integration of a fully screen-printed electrolyte biosensor is described. The

purpose of this sensor is to determine the state of fitness during sports activity by measuring the ammonium

concentration in sweat. Focusing on the selectivity of the ammonium sensor against interfering sodium and

potassium ions, the separate solution method (SSM) and the fixed interference method (FIM) are compared

on the basis of a single sensing device. The latter is mainly supported by the excellent stability of the

sensors. For both interfering ions, the FIM analysis shows a sufficient margin for the operation of the sensor

in the desired application in wearable health and fitness monitoring from sweat. The selectivity coefficients

are better than 0.01 for sodium and still better than 0.1 for potassium. SSM delivers higher selectivity in

both cases, although the discrepancies in selectivity point towards further optimization potential in the

sensor architecture or materials combination.

1 INTRODUCTION

In recent years, many efforts were carried out for the

development of printed sensors that analyze ions in

bodily fluids. In latest publications from Dam et al.,

(2015) and Matzeu et al., (2016), ion-selective

electrodes (ISEs) are used for detection of chloride

and sodium concentrations in sweat. In our previous

work (Oertel et al., 2016) we already presented a

printed sensor for analysis of ammonium in sweat.

The content of ammonium in sweat is correlated

to physical overstrain as published by Ament et al.,

(1997), Alvear-Ordenes et al., (2005) and Meyer et

al., (2007). Therefore, using a sensor for the direct

monitoring of ammonium in sweat is very attractive

for fitness as well as for healthcare applications. A

combination with other body parameters, e.g.

cardiovascular signals, improves the assessment of

the body status in sports performance diagnostics as

well as the health status for diagnostics and therapy

of diseases.

The advantage of an ion-selective electrode and

its main feature is the selectivity for a primary ion

with respect to ions interfering from the background.

For all ion-selective membranes, interfering ions that

have similar size of the ionic radius and the same

ionic charge could create a distracting potential at

the working electrode. Therefore, it is important to

determine the selectivity coefficient of the ISE.

For an ammonium-selective electrode using

nonactin as ionophore, especially sodium and

potassium can act as interfering ions. Both ions have

the same charge as ammonium and nearly the same

ionic radius. Therefore, they can affect the proper

evaluation of the ammonium level. This is even

more critical as both ions are main components of

sweat, which are active in ranges between 25 mmol/l

and 60 mmol/l for sodium and from 5 mmol/l to 18

mmol/l for potassium (Meyer et al., 2007).

For the application-oriented evaluation of the

selectivity coefficient, it is very important to

consider the tolerated values of the specific area of

interest. Furthermore, for comparison against the

results of other groups it is necessary to take into

account the determination method of the selectivity

coefficient (Bakker et al., 2000 and Egorov et al.,

2014).

Until now, other works presenting printed

ammonium-selective electrodes use only a single

method for the determination of the selectivity

ZÃ˝urner A., Oertel S., Schmitz B., Lang N., Jank M. and Frey L.

Determination of the Selectivity of Printed Wearable Sweat Sensors.

DOI: 10.5220/0006296400810087

In Proceedings of the 10th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2017), pages 81-87

ISBN: 978-989-758-216-5

coefficient. For example, Guinovart et al., (2013)

utilize the separate solution method (SSM), whereas

Koncki et al., (1999) apply the fixed interference

method (FIM) for the characterization of screen-

printed ammonium-selective electrodes. The SSM is

simple, but often is not representative for a real, i.e.

containing different ions, solution. In contrast, the

FIM agrees more with the experimental conditions

(Spichiger-Keller, 1998), but may offer a limited

range of evaluation.

In this work, we directly compare the determina-

tion of the selectivity coefficients by both respective

methods and discuss the differences of the

evaluation.

Measurements employing artificial sweat are

conducted to evaluate the sensor in an environment

close to the projected application. The data

collection is based on a fully integrated electronic

system transferring the derived data wirelessly to a

mobile device.

In the experimental part of this paper, the

fabrication of the sensor as well as the

characterization methods used for determination of

the selectivity of the sensor is described. Besides,

the data acquisition and transfer is shortly presented.

In the second part of the paper, the results of the

selectivity coefficient for one sensor using both

methods are shown for each interfering ion, sodium

and potassium. Furthermore, first results measuring

artificial sweat are presented. In the followed part

the difference of the resulting selectivity coefficient

are discussed.

2 EXPERIMENTS

The fabrication of the sensor using screen printing

and two different methods for characterization of the

sensor selectivity are described. A system for data

acquisition and transfer were also developed.

2.1 Sensor Fabrication by Screen

Printing

For screen-printing of the electrodes, the same

process as described previously (Oertel et al., 2016)

were used. However, the used screen-printed pastes

were changed for optimization of the printed design.

Silver-based (Loctite ECI 1006 E&C) and carbon

filled (Electrodag PF-407C) pastes for screen

printing were now acquired from Henkel AG & Co.

KGaA (Düsseldorf, Germany). The previously used

silver-silver chloride paste with a ratio of 65:35

(125-21) for the reference electrode was acquired

from Creative Materials (Ayer, MA, USA). For the

insulating layer, a Barium titanate (PE-BT 101)

paste from Conductive Compounds was used.

The pastes were printed on flexible polyethylene

naphthalate (PEN, QA65HA, 125µm, Teijin DuPont

Films Japan Limited) and polyimide (PI, Kapton

HN, 125µm, Müller GmbH) foils.

A stabilizing layer on top of the reference

electrode (Guinovart et al., 2013) was prepared from

a mixture of sodium chloride (NaCl, ≥ 99.5%,

BioXtra), methanol (99.8%, anhydrous), and

polyvinyl butyral (PVB, Butvar® B-98, all

purchased from Sigma Aldrich).

Different mixtures of the ion-selective membrane

were tested and the best result is presented. Here, the

Ammonium ionophore (nonactin in Cocktail B,

Fluka) was obtained from Sigma Aldrich for

fabrication of the ion-selective electrode. Cocktail B

consisted of the ionophore Nonactin (3.5 wt%),

polyvinyl chloride (0.9 wt% PVC) as polymer

matrix, tetrahydrofuran (62.3 wt% THF) as solvent,

dibutyl sebacate (32.95 wt% DBS) as plasticizer,

and potassium tetrakis(4-chlorophenyl)borate

(0.35 wt% KTClPB) as anionic sites.

Analytical grade salts of ammonium chloride,

potassium chloride, and sodium chloride were

purchased from Bernd Kraft for standard calibration

solutions.

The artificial sweat (1020-1-D and 1006-1-B)

was purchased from synthetic urine. The composite-

on of this artificial sweat corresponds to DIN ISO

9022-12 and DIN ISO 105-E04 norm and contains

typical pH value of sweat and typical concentrations

of sweat ingredients like urea, lactic acid, sodium

and ammonium chlorides, acetic acid etc., in case of

1006-1-B, it is without ammonium chloride. We

used this artificial sweat by addition a defined

ammonium ion concentration of 0.1 mol/l.

For the fabrication of the sensor, an automated

screen- and pattern printer from Ekra (series X1)

was used. The silver-silver chloride paste was

printed using a polyester screen (110 µm mesh - 34

µm wire thickness x 22.5° cover angle, 10 µm -15

µm emulsion over mesh, EOM). The other pastes,

silver, carbon as well as the insulator were printed

using a stainless steel screen (VA 270-0.036x22.5°,

5-10µm EOM).

The printed layers on foils were annealed at

150°C for different times (5 minutes up to

15 minutes) on a hot plate (PZ 28-2 EZ, Harry

Gestigkeit). For potentiometric measurements, a

2636B Sourcemeter from Keithley instruments

(Cleveland, OH, USA) was used.

BIODEVICES 2017 - 10th International Conference on Biomedical Electronics and Devices

The printed sensors were connected to a

commercial flexible flat connector (FFC) with a

pitch of 2.54 mm with seven contacts. The sensor

layout was designed based on the typical pitches of

commercial FFCs. The FFC was placed on a printed

circuit board (PCB). The analog circuit consists of

an instrumentation amplifier (G=1) and low pass

filter (G=1). For data acquisition and transfer, a

system on a chip (SOC) including an 8051 based

low power microcontroller and RF front end is

utilized. The MCU is featured by 12 bit successive

approximation ADC. Data transfer is performed via

Bluetooth Smart (BLE). To allow wired data

transfer, an additional connector is included in the

circuitry.

The data was transferred via Bluetooth Low

Energy to an Android (Google, Mountain View,

USA) device.

The sensor design was described in previous

work by Oertel et al., (2016) and consists of a

bielectrode system that combines a silver working

electrode with a silver-silver chloride reference

electrode. A schematic and a photograph of the

sensor can be seen in Figure 1.

Figure 1: (a) Schematic and (b) photograph of screen-

printed ammonium selective potentiometric sensor with

additionally drop-casted reference and ion-selective

membranes.

The sensor was printed on flexible PEN and PI

foils. Moreover, the working electrode was coated

with an additional carbon layer between the ion-

selective membrane and the silver electrode for

achieving chemical inertness. In a final printing step,

an insulating layer with contact openings for

electrical contacts and the active area for later

deposition of the ion-selective membrane was done.

After screen-printing, the electrodes were

modified (see Figure 1a). The reference electrode

was covered with a mixture of PVB, methanol, and

NaCl as it was published by Guinovart et al., (2013).

The mixture was drop-casted and dried for about 24

hours at room temperature. Furthermore, the ion-

selective membrane was produced by drop-casting a

cocktail of ammonium ionophore onto the active

area of the working electrode. It was again dried

overnight at room temperature.

2.2 Sensor Characterization

The selectivity coefficient is derived from the

response of the sensor to different ionic analytes and

their respective activities. It can be calculated from

the Nikolsky-Eisenman equation (Umezawa et al.,

2000 and Spichiger-Keller, 1998), an extension of

the Nernstian equation (Bakker et al, 2000) that

considers the presence of an interfering ion:

log



















.











log



log



(1)

with K

as the selectivity coefficient, E

the

potential of the interfering ion j, E

the potential of

the primary ion i, z

the valency of the primary ion i,

the valency of the interfering ion j, R the universal

gas constant, T the temperature, F the Faraday

constant, a

the activity of the primary ion i and a

the activity of the interfering ion j.

Using the separate solution method (SSM), the

calibration curves of separate standard solutions of

both, the primary ion and the respective interfering

ion, are prepared with varying ion activities.

(Umezawa et al., 2000). In our case the

concentration range of the solutions was between

-5

mol/l (pC 5) and 10

-1

mol/l (pC 1) for each of

the ions, ammonium, sodium, and potassium. The

range refers to the typical ammonium level in sweat

before and during physical overstrain (Czarnowski

et al., 1992, Guinovart et al., 2013).

The electromotive force (EMF) of the primary

ion and the interfering ions were recorded

successively starting from low concentrations. A

current-less measurement that does not distort the

output voltage is conducted applying the high-

impedance voltage measurement mode of the 2636B

source meter delivering 10

ohms of internal

resistance. Each data point was collected after 7 min

of settling time. After measurement, the sensor was

rinsed with deionized water and submerged into the

subsequent standard solution.

For determination of the selectivity coefficient

via SSM, the resulting EMFs of primary ions and

interfering ions are evaluated at a certain

concentration level. At equal a

and a

, equation 1

can be reduced to









































.

(2)

Determination of the Selectivity of Printed Wearable Sweat Sensors

For the fixed interference method (FIM) the

electromotive force for solutions of constant ion

activity of the interfering ion and varying activity of

the primary ion is measured. For the determination

of the selectivity, the EMF is plotted versus the

logarithm of the ion activity of the primary ion

(Umezawa et al., 2000). If the effect of the

interfering ion on the sensor is dominant, typically at

low concentrations of the primary ion, the latter is

completely shielded, leading to a constant EMF.

However, when the concentration of the primary ion

is increased above a critical level, the EMF should

follow the concentration of the primary ion only,

leading to a linear increase of the EMF. Fitting the

constant branch depicting the EMF of the interfering

ion and the range, where the primary ion dominates,

delivers two lines that intersect at a characteristic

concentration a

of the primary ion. This value refers

to the minimum detectable concentration of the

primary ion with respect to the distorted

environment, i.e. the concentration a

of the

interfering ion. The respective selectivity coefficient

is calculated as follows (Umezawa et al., 2000):























⁄

(3)

2.3 Automated Calibration and Data

Acquisition via Mobile Device

The printed sensor circuit is attached to a

multifunctional electronic box.

The differential sensor voltage is converted to a

ground related and filtered voltage by the analog

front end. This ground related voltage is then

acquired by the ADC of the MCU at a sampling rate

of 2 Sa/s. To lower the power consumption for data

transmission, five samples are accumulated on the

MCU’s RAM before sending the data via BLE to the

mobile device. Thus a transmission rate of 0.4 Hz is

used.

The data received by the Android mobile device

can be used for further data handling and evaluation

(see Figure 2). When a calibration for the sensor is

available, the app calculates the corresponding

concentration and visualizes a range in which the

concentration lies.

For the calibration, the sensor is submerged into

the different calibration solutions (pC 5 to pC 1) for

7 minutes each to get constant values over time

(Oertel et al., 2016). This covers the typical

ammonium range in sweat before and during

physical strain.

Figure 2: Sensor system with circuit board and data

acquisition/transfer via Bluetooth 4.0 for the concentration

of pC 3, pC 2 and pC 1.

3 RESULTS

Selectivity coefficients of the ammonium sensor

were evaluated by separate solution method (SSM)

and fixed interference method (FIM) for the

interfering ions sodium and potassium.

3.1 Selectivity against Sodium

In Figure 3 the resulting calibration curves of the

two methods are shown for sodium as disturbing ion.

In the case of SSM (Figure 3a), the EMFs for the

concentrations of 0.1 mol/l (pC 1), 0.01 mol/l (pC 2)

and 0.001 mol/l (pC 3) of both ions were used for

calculation of the selectivity coefficient according to

equation 2. The coefficient for FIM (Figure 3b) was

calculated from the interpolated minimum detectable

concentration of the primary ion with respect to the

distorted environment. This is derived from the

intersection of the previously described fits of the

constant branch related to the interfering ion and the

linear range, where the primary ion dominates (see

dotted lines in Figure 3b).

The mean selectivity coefficient obtained from

SSM K







,





 is 0.0034 ± 0.0003. The

coefficient at each concentration is also observable

in Figure 3a by the inverted triangles. It is shown

that the selectivity coefficient is nearly constant in

this region of concentration. Selectivity coefficient

determined by FIM K







,





 is 0.0095. The

coefficients are comparable to the results published

by Guinovart et al., (2003, K







,





 0.0013) and

Koncki et al., (1999, K







,





 0.0079).

BIODEVICES 2017 - 10th International Conference on Biomedical Electronics and Devices

In both measurements, the same sensor was used.

Despite numerous measurements at full range of

concentrations during SSM, the EMF of the constant

branch in FIM (Figure 3b) is equivalent to the EMF

of pC 2 of sodium measured by SSM. Therefore, no

additional calibration of the sensor is necessary

before FIM as there is no aging of the sensor visible.

Nevertheless, the selectivity derived by FIM shows a

slightly increased value with a difference of 0.0061.

Figure 3: Measurement of ammonia concentration upon

interference with sodium for the determination of the

selectivity coefficients using (a) SSM and (b) FIM.

3.2 Selectivity against Potassium

For the determination of the selectivity coefficient

towards potassium, the resulting calibration curves

are shown in Figures 4a and b.

With respect to the measurements shown for

sodium interference, the EMF values of standard

solutions of potassium are closer to those for

ammonium at each respective concentration. From

the SSM calibration curve (Figure 4a), a reduced

selectivity against potassium can be derived. This is

due to the ionic radius of potassium, which is closer

to that of ammonium.

The mean selectivity coefficient via SSM





SSM

 is 0.0832 ± 0.0086 (see inverted triangles

in Figure 4a) and is almost constant for the linear

range of the concentration as it is in the case of

sodium. Here, the sensor achieved a slightly increa-

sed selectivity compared to the result of Guinovart et

al., (2000) obtaining a selectivity coefficient of

0.0158 using SSM.

Figure 4: Calibration curves for determination of the

selectivity coefficient using (a) SSM and (b) FIM for the

interfering ion potassium.

Again, the EMF of the constant branch in the

FIM measurement corresponds to the potassium

concentration of pC 2 of the SSM calibration curve.

The increased minimum detectable concentration of

the primary ion (pC 2.65) is confirming the

observation of a reduced selectivity coefficient

derived from SSM method.

The resulting selectivity determined by FIM

K







,





 is 0.0912. Again, a slightly increased

Determination of the Selectivity of Printed Wearable Sweat Sensors

coefficient for FIM can be recognized when

compared to SSM (with a difference of 0.008).

Compared to the published work by Koncki et al.,

(1999) with a selectivity coefficient of 0.0398, our

sensor achieved a slightly increased selectivity.

3.3 Artificial Sweat

Two compositions of artificial sweat were measured

to test the selectivity of the sensor in more realistic

environments. After recording the ammonium

calibration curve, the sensor was dipped into artifi-

cial sweat with specific ammonium concentrations

of 0.065 mol/l (1020-1-D) and 0.1 mol/l (1006-1-B).

The measured EMF values of both artificial sweat

samples as well as the EMFs of 0.065 mol/l and 0.1

mol/l ammonium standard solutions are listed in

Table 1. The shift of the EMF for the ammonium

standard solution compared to the calibration curves

of the SSM is due to the long shelf time between the

measurements. However, the slope of the calibration

curve remains constant. For the measurement of the

artificial sweat, a slight shift of 40 mV (1020-1-D)

and 32 mV (1006-1-B) compared to ammonium

standard solution is observable, which is due to the

influence of the contained ions, especially the high

amount of sodium as well as the influence of the pH.

Table 1: Resulted EMFs for ammonium standard solution

and artificial sweat (1020-1-D and 1006-1-B).

EMF in mV

c in

mol/l

Ammonium

Standard

Solution

1020-1-D

1006-1-B

0.1 422 454

0.065 400 440

4 DISCUSSION

Compared to already published selectivity

coefficients of printed ammonium-selective sensors

(Guinovart et al., 2013 and Koncki et al., 1999), our

fully-printed sensor shows similar selectivity and for

the first time compares FIM to SSM selectivity data

using same devices for both techniques.

The comparison of the selectivity coefficients

determined via SSM and FIM shows for both,

sodium and potassium, a higher value using FIM

technique. This is contradictory to the theory,

because the determination of the selectivity

coefficient based on ideal Nernstian behavior should

be independent on the method (Bakker et al., 2000).

However, Figures 3a and 4a already feature a

non-ideal behavior expressed by an increase of the

EMF that is roughly +35 mV/pC higher than the

expected Nernstian slope of 59 mV in the

ammonium, sodium, and potassium standard solu-

tions. According to Bakker et al., (2000), possible

reasons for this deviation can be long-term degrada-

tion of the selective membranes or electrodes or

sample-to-sample variations.

The printed sensor in this work was tested over a

period of a few months. The repeated measurement

of calibration curves after treatment in different ion

environment showed the same slope and potentials

without shift within one month, so the sensors show

sufficient long-term stability without aging effects.

Considering the reproducibility of the calibration

curve, the fabrication of sensors after three printing

runs yields sensors with equivalent behavior. The

sensors showed also no hysteresis in the calibrations

curves of ammonium ions from low to high and

from high to low concentrations. Furthermore, the

utilization of a single sensor for both extraction

techniques rules out sample variations in this work.

Besides, further work is in progress testing the

reproducibility of the sensor selectivity prepared

with the same composition of the ion-selective

membrane. So far, the sensors show promising

results.

The stability of the reference electrode was

proven by means of cyclic voltammetry (not shown

here). By comparing cyclic voltammetry data of

fresh and used reference electrodes, no aging of the

reference electrodes could be observed.

Since most of the possible reasons for the deviant

value of the Nernstian slope are already disproved,

other reasons like effects caused by the solid-state

multilayer construction of the sensor and enhanced

diffusion of ammonium through the membrane

(Nery et al., 2016) need to be analyzed in greater

detail.

However, the measurement via FIM represents in

general more realistic conditions when aiming at

human sweat than the measurement via SSM

(Spichiger-Keller, 1998).

5 CONCLUSIONS

In this work, we reported data on the selectivity of a

fully-screen printed ammonium sensor for analysis

of human sweat against the main interfering ions,

sodium and potassium.

Within physiologically reasonable concentrations

of the species under consideration, the sensor

BIODEVICES 2017 - 10th International Conference on Biomedical Electronics and Devices

delivers excellent selectivity against the secondary

ions. The more application-related fixed-interference

method (FIM) yields selectivity coefficients of

0.0095 and 0.0912 against sodium and potassium,

respectively.

However, further investigations are necessary to

understand the non-ideal Nernstian behavior, which

is revealed by the difference of the selectivity

coefficients derived from FIM and the

complementary separate solution method.

Furthermore, the proper detection of ammonium

activity in artificial sweat is reported giving reason

for a viability of the approach even in harsh but

realistic environments. These results encourage

further research on the integration of sensors and

read-out electronics with textiles for wearable

functional sports clothing. Moreover, on body tests

are of particular interest for future work. Here,

considering possible motion artefacts are important.

ACKNOWLEDGEMENTS

This contribution was supported by the Bavarian

Ministry of Economic Affairs and Media, Energy

and Technology as a part of the Bavarian project

”Leistungszentrum Elektroniksysteme (LZE)”.

REFERENCES

Dam, V. A. T., Zevenbergen, M. A. G., van Schaijk, R.

2015. Flexible Chloride Sensor for Sweat Analysis.

Procedia Eng., 120, 237-240.

Matzeu, G., O'Quigley, C., McNamara, E., Zuliani, C.,

Fay, C., Glennon, T., Diamond, D. 2016. An integra-

ted sensing and wireless communications platform for

sensing sodium in sweat. Anal. Methods, 8 (1), 64-71.

Oertel, S., Jank, M. P.M., Frey, L., Hofmann, C., Lang,

N., Struck, M. 2016. Screen-printed biochemical

sensors for detection of ammonia levels in sweat –

towards integration with vital parameter monitoring

sports gear. Proceedings of the 9th International Joint

Conference on Biomedical Engineering Systems and

Technologies (BIOSTEC 2016), 1, 160-165.

Ament, W., Huizengau, J. R., Mook, G. A., Gips, C. H.,

Verkerke, G. J. 1997. Lactate and ammonia concentra-

tion in blood and sweat during incremental cycle

ergometer exercise. Int. J. Sports Med., 18, 35-39.

Alvear-Ordenes, I., Garcia-Lopez, D., De Paz, J.,

Gonzalez-Gallego, J. 2005. Sweat lactate, ammonia,

and urea in rugby players. Int. J. Sports Med., 26, 632-

637.

Meyer, F., Laitano, O., Bar-Or, O., Mc Dougall, D.,

Heigenhauser, G. J. F. 2007. Effect of age and gender

on sweat lactate and ammonia concentrations during

exercise in the heat. Braz. J. Med. Biol. Res., 40(1),

135-143.

Bakker, E., Pretsch, E., Bühlmann, P. 2000. Selectivity of

Potentiometric Ion Sensors. Anal. Chem., 72(6), 1127-

1133.

Egorov, V. V., Zdrachek, E. A., Nazarov, V. A. 2014.

Potentiometric Selectivity Coefficients: Problems of

Experimental Determination. J. Anal. Chem., 69 (6),

535-541.

Guinovart, T., Bandokar, A. J., Windmiller, J. R.,

Andrade, F. J., Wang, J. 2013. A potentiometric tattoo

sensor for monitoring ammonium in sweat. Analyst,

138, 7031-7038.

Koncki, R., Glab, S., Dziwulska, J., Palchetti, I., Mascini,

M. 1999. Disposable strip potentiometric electrodes

with solvent-polymeric ion-selective membranes

fabricated using screen-printing technology. Anal.

Chim. Acta, 385, 451-549.

Spichiger-Keller, U. E. 1998. Chemical sensors and

biosensors for medical and biological applications.

Wiley-VCH. 3-527-28855-4.

Umezawa, Y., Bühlmann, P., Umesawa, K., Thoda K.,

Amemiyay, S. 2000. Potentiometric selectivity

coefficients of ion-selective electrodes. Part I.

Inorganic cations (technical report). Pure and Applied

Chemistry, 72 (10), 1851-2082.

Czarnowski, D., Gorski, J., Jozwiuk, J. and Boron-

Kaczmarska, A. 1992. Plasma ammonia is the princi-

pal source of ammonia in sweat. Eur. J. Appl. Physiol.,

65, 135-137.

Nery, E. W., Kubota, L. 2016. Integrated, paper-based

potentiometric electronic tongue for the analysis of

beer and wine. Anal. Chim. Acta, 918, 60-68.

Determination of the Selectivity of Printed Wearable Sweat Sensors