Improvement of the detection limit of impedance spectroscopy for glucose detection by functionalization of electrode surfaces in aqueous NaCl solutions with boronic acid

(1)

Universiteit van Amsterdam Vrije Universiteit, Amsterdam April-July 2019

Improvement of the detection limit of impedance spectroscopy

for glucose detection by functionalization of electrode surfaces in

aqueous NaCl solutions with boronic acid

Fanja Bouts

11033592

Supervisor & examiner: Elizabeth von Hauff

Second examiner: Rudolf Sprik

Bachelor thesis, Physics and Astronomy

Universiteit van Amsterdam & Vrije Universiteit

Faculteit der Wis- en Natuurkune

This thesis was written as a part of the Bachelor of Science in Physics and astronomy at the UvA and VU. The experiments were done in the D-Lab

(2)

(3)

i

Abstract

Glucose measurements are relevant for diabetic care and are measured by taking in blood samples. The aim is to measure glucose concentrations continuously in other bodily fluids via non invasive measurements. such as impedance spectroscopy measurement in samples using sweat. In this thesis I investigated the relationship between impedance spectra and glucose concentrations. Electrodes were functionalized using boronic acid, known to selectively bind glucose, to improve the limit of detection (LOD). I found that the overall LOD and slope of the linear fit did not improve after adding boronic acid to the electrode surface, ranging from 284mg/dl to 735mg/dl. More measurements need to be carried out to see if this does help improve the detection of low concentrations of glucose.

(4)

ii Contents

List of Figures

2.1 In this figure the open chain structure of D-glucose is shown, with the numbered Carbon compounds on the right (1). . . 3 2.2 In figure 2.2, it is shown how α-D-glucose turns into β-D-glucose via the

linear chain-form γ-D-glucose (2) . . . 4 2.3 In this figure the conversion from Glucose into Gluconic acid via

Gluconolactone is shown, with GOX as a catalyser in the reaction (3). . . 5 2.4 In this figure the structure of boronic acid is shown, with B the boron

compound, R the alkyl group, and two OH groups. (4). . . 5 2.5 This figure shows a schematic illustration of the functionalization of an AU

surface with boronic acid. A) The molecules form a monolayer on the AU surface. B) The molecules in the solution bind to the monolayer on the AU-surface. (5). . . 6 2.6 In this figure a 3 point electrochemical cell is shown. The cell contains of a

working electrode, a counter electrode and a reference electrode (6). . . . 7 2.7 In this figure a phase shift in time of the applied voltage and the resulting

current is shown (7). . . 8 2.8 In this figure a Nyquist plot is shown (8). Here, the negative of the

imaginary part of the impedance is shown on the y-axis, and the positive real part of the impedance is outlined on the x-axis. . . 11 2.9 In this figure a Nyquist plot is shown divided by area’s RA to RB,

representing the internal resistance and RB to RC, representing the

charge-transfer resistance and the double layer resistance (9). . . 12 2.10 In this figure two different Bode plots are shown (10). The graph on the

top shows the log of the Impedance plotted against the frequency. In the lower graph the phase shit is plotted against the frequency. . . 13 2.11 This image shows an equivalent circuit model similar to the circuit used in

this research (11) . . . 15 2.12 This image shows an equivalent circuit model similar to the circuit used in

this research (11) . . . 15 2.13 This figure shows the electrode electrolyte interface. It shows the Stern

layer, the diffuse layer and then the Bulk electrolyte inbetween the detectors (9). . . 17 3.1 This figure shows the Dropsens Inter-digitated Electrode (IDE), consisting

of a carbon working and counter electrode, a silver reference electrode, and a 3 point connector printed on a ceramic substrate. In this research a gold working and counter electrode was used, rather than a carbon one. . . 18 3.2 In this figure connector type A is shown. On the left the connector is shown

from above, and on the right the connector can be seen from the front. The image on the right shows the opening in which the electrode is put. . . . 19 3.3 In this figure connector type B is shown. On the left the connector is shown

from above, and on the right the connector can be seen from the front. The image on the right shows the opening in which the electrode is put. On the top the screw can also be seen that has to be tightened once the electrode is put in this connector. . . 19

(6)

iv List of Figures

4.1 This image shows the Nyquist representation for a measurement set with detector A1. The graph shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graph at high frequencies is zoomed in to tell apart the different concentrations. On the right the concentrations of glucose in mg/dL are shown. . . 23 4.2 This image shows the Nyquist representations for two measurement sets

with detector A1. The top graph shows the measurements done without boronic acid. The bottom graph shows the measurements done with boronic acid. The graphs shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graphs at high frequencies are zoomed in to tell apart the different concentrations. On the right the concentrations of glucose in mg/dL are shown. . . 24 4.3 In this figure two graphs are shown with the real Z plotted against the

glucose concentration. The red line represents a linear fit through the data points. This data is measured without the addition of boronic acid. . . . 26 4.4 In this figure the real Z is plotted against the glucose concentration. The red

line represents a linear fit through the data points. This data is measured with boronic acid. . . 27 4.5 In this figure two graphs are shown with the real Z plotted against the

glucose concentration. The red line represents a linear fit through the data points. This data was measured without the addition of boronic acid. . . 28 4.6 In this figure two graphs are shown with the real Z plotted against the

glucose concentration. The red line in the top graph represents a linear fit through the data points. This data is measured without the addition of boronic acid. The graph on the bottom shows a zoomed in version with the limits -100 to 1100 on the x-axis and 68 to 71 on the y-axis. . . 30 4.7 In this figure two graphs are shown with the real Z plotted against the

glucose concentration. The red line in the top graph represents a linear fit through the data points. This data is measured with a layer of boronic acid on the detector. The graph on the bottom shows a zoomed in version with the limits -100 to 1100 on the x-axis and 66 to 69 on the y-axis. . . . 31 4.8 Here the Limit of detection, the Adjusted R2 and the slope are shown for

each detector, with and without boronic acid. . . 32 A1.1 Here the bode plots are shown for detector A1. The top graph shows the

phase against the frequency. The bottom graph shows the real impedance against the frequency. . . 38 A1.2 Here the bode plots are shown for detector A1. The top graph shows the

phase against the frequency. The bottom graph shows the real impedance against the frequency. The top graph is without boronic acid and the bottom graph is with boronic acid. . . 39

(7)

List of Figures v

A3.1 This image shows the Nyquist representations for two measurement sets with detector B1. Both of the measurements were done without boronic acid. The graphs shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graphs at high frequencies are zoomed in to tell apart the different concentrations 46 A3.2 This image shows the Nyquist representations for two measurement sets

with detector C1. The top graph is meaasured without boronic acid. The bottom graph is measured with boronic acid. The graphs shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graphs at high frequencies are zoomed in to tell apart the different concentrations . . . 47

(8)

1

1 Introduction

There have been drastic rises in the amount of people suffering from diabetes in the last few decades, and this number is expected to increase in the coming years (12).

For individuals that suffer from diabetes, their body is incapable of regulating insulin on its own. As a consequence, the glucose level in their blood does not stay stable like it should be, but fluctuates between concentrations that are either too high or too low. To ensure their glucose level stays within this range, it must be measured manually. This can be done by pricking the tip of the index finger in order to draw blood, a measurement of the glucose concentration concentration can be taken of the resulting blood droplet. However, this method is invasive and uncomfortable for the patient. In addition, this method is wasteful as new test strips have to be used each time. These test strips can also vary in quality, with the risk of false or inaccurate reading of glucose concentrations as a serious consequence (2). Therefore, the development of a different glucose detection method is very desirable.

The key reason why blood is used as a bodily fluid for glucose detection, is its high concentration of glucose. Higher concentrations of glucose are easier to detect then lower concentrations and are therefore more favourable for testing. Another reason why blood is used is because enzymatic activity, the technology for glucose measurement that is currently used, is only stable in blood. Blood also has a constant temperature, pH and composition. In other bodily fluids such as saliva or sweat the enzymatic activity is unstable and factors such as temperature, pH and composition is varied. furthermore, the glucose concentration is much lower and as a result the detection of glucose is more challenging and the technique is more complicated. Nevertheless, there are many possible techniques that can be developed and are being researched (13). One of these methods is impedance spectroscopy, measuring the current response over a range of frequencies. It has been shown that a difference in glucose concentrations in a solution cause significant changes in the impedance spectrum. In this research the impedance response to different glucose concentrations in NaCl solutions are determined and analyzed. I will also look into what processes occur in the solutions and on the electrode surface that can influence the impedance outcome. In addition to this, I will add boronic acid to functionalize the electrode surface to improve the impedance spectroscopy glucose measurement. The layer of boronic acid that forms could bind the glucose molecules in the solution to the surface of the electrode. The accumulation of the glucose molecules on the electrode surface can improve the limit of detection.

(9)

2

This has led to the following research question:

Does the detection limit of impedance spectroscopy improve by functionalizing the electrode surface with boronic acid when testing for physiologically relevant glucose concentrations in a frequency range of 1Hz-1MHz?

(10)

3

2 Theoretical Background

2.1 The glucose molecule

Glucose is an organic compound that belongs to the carbohydrates as a monosaccharide, a simple sugar. Glucose is the source of energy in the cell and is the main free sugar circulating in our blood. Glucose comes in the form of D-glucose and L-glucose. These two forms of glucose are stereoisomers. This means that they only differ from one another by the chirality. A chiral center is a molecule that has 3-dimensional asymmetry (14). From these two forms of glucose, D-glucose is much more abundant. In this experiment D-glucose was used. Therefore when mentioning glucose in this thesis, D-glucose is referred to. The full name of D-(+)-glucose is 2,3,4,5,6-pentahydroxyhexanal, or C6H12O6 (1).

The open chain structure of D-glucose is shown in image 2.1.

Figure 2.1: In this figure the open chain structure of D-glucose is shown, with the numbered Carbon compounds on the right (1).

D-glucose comes in different 3D configurations, resulting in different forms of the molecule. There are two cyclic forms of D-glucose, α-D-glucose and β-D-glucose. These two forms of D-glucose only differ from one another by one OH-group in the molecule. Once these two forms are put into an aqueous solution, they come to an equilibrium. The reaction in which α-D-glucose turns into β-D-glucose happens via a linear chain form of D-glucose, γ-D-glucose. This linear chain form serves as an intermediate between the two cyclic forms of D-glucose. This reaction is shown in figure 2.2.

(11)

4 2.2 Glucose detection

Figure 2.2: In figure 2.2, it is shown how α-D-glucose turns into β-D-glucose via the linear chain-form γ-D-glucose (2)

It is unsure if the interchanging inbetween the different forms of glucose has an affect on the detection of glucose. According to van Enter & von Hauff (2), it probably does not have an effect on the detection of glucose. However, according to Bruen (3), the interchanging of the different forms could influence the detection of glucose.

Glucose is soluble in water. The solubility of glucose in water is 133 mg/ml (15). When glucose is dissolved in water, the ion concentraion is not increased. It is a stable molecule. This causes glucose not to partake in charge transfer. Glucose as a molecule has a larger dipole moment than water molecules. This dipole moment changes when the electric field in the solution rapidly changes. This reorientation of dipole moments and the OH-groups in glucose causes dielectric polarization. This change in polarization and therefore resistance of the dielectric material can be measured by electrochemical impedance spectroscopy. This depends on the input frequency, and if it’s similar to the relaxation frequency of the glucose.

2.2 Glucose detection

There are many different techniques that exist to measure concentrations of glucose in a solution. From these existing techniques, most of them are based on electrochemical reactions with the use of enzymes. Enzymes that are often used for the purpose of glucose detection, are glucose oxidase (GOx) or glucose dehydrogenase, with GOx the most widely used (16). The main reason for its wide use is its high sensitivity for glucose, on top of the fact that it can withstand large changes in pH, temperature and ionic strength (3). Ionic strength is the amount of ions in the solution. Adding this to a glucose solution causes the glucose to oxidise. The GOx reacts with β-D-glucose, forming gluconolactone, and then gluconic acid. This process is shown in image 2.1.

(12)

2.3 Boronic acid 5

Figure 2.3: In this figure the conversion from Glucose into Gluconic acid via Gluconolactone is shown, with GOX as a catalyser in the reaction (3).

The glucose concentration is then measured by a glucometer out of the reaction flux for the reaction into Gluconic acid (2).

Besides the use of enzymes for glucose detection, there are other possible detection methods. Since glucose is a neutral, electrochemically inactive molecule, using conductivity for detection can be difficult. As it is chemically inactive, detection via an electrochemical reaction is also challenging. For this reason detecting the molecule physically rather than chemically could be an outcome. One way to do this is to measure a voltage as a result of putting a current through a glucose solution. The impedance response can then be measured. Since Impedance does not just measure resistance of a system, but the resistance and the reactance, this measurement method contains more information.

2.3 Boronic acid

Boronic acids are organic compounds that consist of an electron deficient boron compound, possessing one alkyl element and two hydroxyl groups (4). The structure of boronic acid is shown in figure 2.2.

Figure 2.4: In this figure the structure of boronic acid is shown, with B the boron compound, R the alkyl group, and two OH groups. (4).

This causes the boronic acid to be Lewis acidic (17). According to Sivaev & Bregadze (18) a Lewis acid is any kind of molecule or ion that is able to accept a pair of electrons. A Lewis

(13)

6 2.4 Electrode surface functionalization

acid based reaction occurs when a base donates a pair of electrons to an acid. Detecting carbohydrates, such as glucose, is based on the change in Lewis acidity as a result of this reaction. As shown in subsection 2.1, glucose consists of various hydroxyl (OH) groups. The Lewis acidity of the Boronic acid in combination with the electronegative hydroxyl group on the glucose molecules, causes the molecules to be able to form a strong bond. This strong bond is why boronic acid has been chosen in this research to functionalyze the gold electrode surfaces. The electrode surface functionalization is explained in more detail in the paragraph below.

2.4 Electrode surface functionalization

Because glucose concentrations are much lower in body fluids such as sweat and since the interaction of glucose with metal detectors is poor, it could help to funcionalize them. This will cause the sensors to have a higher sensitivity to the glucose molecules, giving a better chance of detection. A molecular monolayer on a surface area is formed when the molecule binds to the substrate surface. Adding a monolayer to the surface of the electrode can lead to its functionalization. When adding boronic acid to the glucose solution, boronic acid molecules can form a bond with the glucose molecules. When the boronic acid molecules have formed a layer on the electrode, the glucose molecules can bind and accumulate at the surface of the electrode.

Figure 2.5: This figure shows a schematic illustration of the functionalization of an AU surface with boronic acid. A) The molecules form a monolayer on the AU surface. B) The molecules in the solution bind to the monolayer on the AU-surface. (5).

(14)

2.5 Electrochemical cell 7

electrodes rather than in the bulk of the solution. This should make the glucose molecules easier to detect, as they are in more direct contact with the detector. Another reason why creating a boronic acid monolayer on the electrode surface could possibly lead to a better detection of glucose in the solution, is because it will stop interchanging between α-D-glucose and β-D-glucose once it binds to the boronic acid molecules. As mentioned before however, it is unsure if this actually adds to a better detection of glucose (2). Therefore, the main reason for the better detection of glucose is assumed to be the glucose being closer to the electrode surface.

2.5 Electrochemical cell

The electrochemical cell used in this experiment is the 3 point electrode. The 3 point electrochemical cell consists of a working electrode (WE), a counter electrode (CE) and a reference electrode (RE). This cell is shown in image 2.3 below.

Figure 2.6: In this figure a 3 point electrochemical cell is shown. The cell contains of a working electrode, a counter electrode and a reference electrode (6).

The working electrode is the main electrode that processes the electrochemical response of the solution, this is where the cell reaction appears. The counter electrode is used in the electrochemical cell to close the current circuit. The current flows between the working electrode and the counter electrode. There is a controlled potential difference between the working electrode and the counter electrode. This difference is measured between the Reference electrode and s, the sense electrode. The reference electrode is used as a point of reference in the way that it has a well known potential. The current flow through the reference electrode is (ideally) kept at zero (6).

(15)

8 2.6 Impedance Spectroscopy

2.6 Impedance Spectroscopy

The electrochemical impedance can be measured by applying an AC voltage V on a system at a certain frequency (ω). The voltage applied on the system changes over time and can be expressed as a function of t. This is shown by function 2.1, where V0 is the amplitude

of the signal, ω the angular frequency of the voltage signal in radians per second and t the time.

V (t) = V0· sin(ωt) (2.1)

As a response to the applied voltage on the system, a current will start to to flow. The function for the response current in shown in 2.2, where IA is the current amplitude.

Function 2.2 also demonstrates the phase shift between the applied voltage and the resulting current, indicated by φ. When there is a voltage applied to a system, there will be a current response with the same frequency, but with a shift in phase. This effect is also shown in figure 2.7.

Figure 2.7: In this figure a phase shift in time of the applied voltage and the resulting current is shown (7).

I(t) = IA· sin(ωt + φ) (2.2)

To determine the complex impedance of the system, Ohm’s law must be applied. This is shown in function 2.3. Here, ZA is defined as amplitude of the voltage signal divided by

the amplitude of the resulting current. With ZA = |Z|.

Z∗ = V (t) I(t) = VAsin(ωt IAsin(ωt + φ) = ZA sin(ωt) sin(ωt + φ) (2.3)

(16)

2.7 Impedance parameters 9

Now, Euler’s relationship, as shown in function 2.4, can be used to rewrite the voltage and current in terms of complex components. The resulting functions for the voltage and current are shown in 2.5 and 2.6.

ejφ = cos(φ) + jsin(φ) (2.4)

V (t) = VA· ejωt (2.5)

I(t) = IA· ejωt−jφ (2.6)

From these new expressions for V(t) and I(t), a new expression for the Impedance can be worked out. What follows is a function for the impedance Z that can be expressed in a real part (ZREAL) and an imaginary part (ZIM) of the impedance. This relation is

shown in function 2.7. The real part of the function represents the in-phase part, where the imaginary part represents the out-of-phase part.

Z∗ = V

I = ZA· (cosφ + jsinφ) = ZREAL+ jZIM (2.7)

2.7 Impedance parameters

Besides the alternating current voltage (AC), there is a direct voltage modulation (DC) in impedance spectroscopy, the modulus spectroscopy. There are two characteristics of the electrode material that are described by dielectric analysis:

• Permittivity

The permittivity of a material shows the ability to store electrical energy.

• Conductivity σ (resistivity ρ alternatively) The conductivity of a material indicates the ability to transfer electric charge.

These two parameters, permittivity and conductivity, are influenced by molecular activity. They are also dependent on the temperature of the material, the frequency of the alternating current, and the applied voltage.

Like the concept of complex impedance |Z|, which is a combination of the real part and the imaginary part of the impedance, there is also a complex permittivity ∗_.

(17)

10 2.7 Impedance parameters

This too, is a combination of the real part ’, or simply the permittivity or dielectric constant, and the imaginary part, ”, known as the loss factor. This relation is given by function 2.11.

∗ = 0 − j00 (2.8)

Here, 0 _{represents the energy storage, or the alignment of the dipoles of atoms. The}

inverse of 0 _{is equivalent to Z}

IM. 00 on the other hand is the inverse of ZREAL, and

represents the conduction component.

Another property of the material that can be calculated from these parameters, is the Polarization density. The Polarization density gives an indication on the amount of aligned dipoles in the material. The function that describes the Polarization density is known as the Debye Equation, shown in equation 2.9.

P = ( − 1)0V (2.9)

This polarization can be either • electronic and atomic

The electronic and atomic polarization is given by a very small translational displacement of the electronic cloud. This happens at a frequency range of THz. • Orientational or dipolar

This is the rotational moment that permanently polar molecules experience. This happens in a frequency range of kHz-MHz.

• ionic

The ionic polarization represents the displacement of the ions with respect to each other. This happens at a frequency range of Hz-kHz.

The capacitance C of a material is given by equation 2.10.

C =

0 0A

d (2.10)

Finally, the complex modulus can be defined as the inverse of the complex permittivity, as shown in equation 2.11.

M∗ = 1 ∗ = M

0_{− jM}00

(18)

2.8 Graphical representation 11

2.8 Graphical representation

As impedance is measured at different frequencies, different plots can be made to represent this data. From the data a Nyquist representation and a Bode representation can be made, as shown below.

2.8.1 Nyquist representation

The impedance measurement can be visualized in a Nyquist plot, shown in figure 2.8. In a Nyquist plot the negative imaginary part of the impedance is plotted against the real part of the impedance. A single point on this plot represents one frequency. On the left hand side of the plot the highest frequencies are shown, and on the right hand side the lowest frequencies are shown (10). A shortcoming of the Nyquist plot is that it does not clearly show the frequency.

Figure 2.8: In this figure a Nyquist plot is shown (8). Here, the negative of the imaginary part of the impedance is shown on the y-axis, and the positive real part of the impedance is outlined on the x-axis.

The image below shows how to interpret the Nyquist plot further. The previous section discusses the many different processes that occur at the electrode-electrolyte interface. These processes can be seen in the Nyquist plots. The Nyquist graph in figure 2.11 is divided by area RA to RB and RB to RC. Although there are contradicting theories on

what processes are shown exactly in a Nyquist plot, a theory posed by (9) can be used to descibe the different area’s of the Nyquist plot. The process that is descibed by the semi-circle, RA to RB, is mostly the internal resistance, or the resistance of the electrolyte.

RB to RC represents different processes occurring on the electrode surface such as the

double layer resistance, and with it the electrode polarization, and the charge transfer resistance.

(19)

12 2.8 Graphical representation

Figure 2.9: In this figure a Nyquist plot is shown divided by area’s RAto RB, representing

the internal resistance and RB to RC, representing the charge-transfer resistance and the

double layer resistance (9).

2.8.2 Bode representation

Another method of presenting impedance data is via a Bode plot. In a Bode plot the log of the frequency is shown on the x-axis, and the log of the absolute value of the impedance on the y-axis. Another version of the Bode plot is with the log of the phase shift on the y-axis, rather than the impedance. These Bode plots are shown in figure 2.9. The Bode plots are useful additions to the Nyquist representation as the frequency can be clearly seen in the graph, whereas this is not possible with the Nyquist.

(20)

2.8 Graphical representation 13

Figure 2.10: In this figure two different Bode plots are shown (10). The graph on the top shows the log of the Impedance plotted against the frequency. In the lower graph the phase shit is plotted against the frequency.

2.8.3 Equivalent circuit modelling

When impedance of a system is measured, the out coming data and graphs can be analyzed. One way to analyze the plots is to fit an equivalent circuit model to it. The circuit that is created for the fit should contain elements similar to the system that is actually measured. The equivalent system must be based on an understanding of the setup and the properties of the system. The circuit can contain various elements that can be used to get a proper fit to the impedance plots. The elements, each with their own properties, can be combined into a circuit by putting them either in series or parallel with respect to each other.

A number of circuit elements relevant for this system are outlined below. • Resistor

To model just the resistive processes in a system, a resistor can be used as a circuit element. Here the impedance is given by just the resistance, so Z = R.

• Capacitor

If one wants to model just the capacitative processes in a system, a capacitor can be used, given by equation 2.10.

• Constant phase element (CPE)

(21)

14 2.8 Graphical representation

with glucose, a redistribution of charges appears resulting in an interface (19). That interface can be modelled by a constant phase element (CPE). This constant phase element is often used as a system shows non ideal behaviour. In many experiments the existing capacitors do not behave ideally, but rather behave as a constant phase element. A CPE is not very realiable at extremely low or extremely high frequencies (20). This is outside the range that is used in this experiment, so that is not applicable here. The Constant Phase Element is a non-ideal circuit element that takes into account the deviations from the ideal resistor or capacitor. Examples of these deviations are, as mentioned before, no uniform potential distribution at the surface of the electrode, or uneven adsorption etc. (21). An expression for the CPE is given by equation 2.12.

ZCP E =

1

Q(jω)n (2.12)

The value of n can vary between -1 and 1. When n=1, the CPE element becomes an ideal capacitor, when n=0 Q becomes 1

R with R an ideal resistor (22).

• Warburg element (Warburg Short)

The Warburg curcuit element can be added to a circuit to take into account the diffusion in the system to and from the electrode surface. It takes into account infinite diffusion. The Warburg impedance is given by equation 5.13.

ZW =

AW

ω0.5(1 − j) (2.13)

In this equation, AW represents the Warburg coefficient, which depends on the

diffusion coefficient D of the particles in the solution, and their concentration on the electrode surface. ω is the angular frequency of the voltage caused by the alternating current (22).

• Open Warburg element (Warburg Open)

An open Warburg element is the impedance of the finite diffusion, which is therefore a more accurate diffusion element in this system than the Warburg element that represents the infinite diffusion.

In images 6.1 and 6.2 2 models are shown. These models were created by N. Narain (11) based on models in literature based on similar research.

(22)

2.9 Electrode-electrolyte interface 15

Figure 2.11: This image shows an equivalent circuit model similar to the circuit used in this research (11)

Figure 2.12: This image shows an equivalent circuit model similar to the circuit used in this research (11)

These models consist of different equivalent circuit elements. These models were fitted to the data from this research to get a better understanding of the processes occurring in the system. Both circuit models came close to the nyquist graphs when they were fitted. This has led to the belief that the processes represented by the circuit elements are processes that occur in this system. These processes from model 1 are the resistance of the electrode interface, represented by circuit element R1. Secondly, the electrode

polarization and the double layer capacitance, posed by the CPE-1 element. Lastly, the resistance of the electrolyte, represented by R2. Model 2 poses that there is an overall

capacitance in the system, C1, an electrolyte resistance, R1, a charge transfer resistance,

R2, and a double layer capacitance C2. Finally, the open Warburg element represents a

finite diffusion of the system. All these phenomena in the system influence the impedance outcome.

2.9 Electrode-electrolyte interface

Conductivity may result due to ionic or electrical transport. In a person’s body, charge is carried by ions. In an electrode however, charge is carried by electrons. At the electrode-electrolyte interface charge is transferred from the one carrier of charge to the

(23)

16 2.9 Electrode-electrolyte interface

other (23). This charge carrier shift happens in an electrochemical reaction at each electrode (20). A consequence of this is accumulation or depletion of matter and charge next to the electrodes. This results in an additional resistance, known as the charge transfer resistance, RCT.

Another consequence of this accumulation and depletion is that additional ions are released from or into the bulk solution. This causes the concentration to differ close to the electrodes compared to that in the bulk solution. An uneven concentration throughout the solution is then created. As a solution always tends to go back to an equal concentration distribution, diffusion occurs from the lower concentration to the higher concentration. This diffusion mass transport that happens because of the difference in concentration (the concentration gradient ∆C∗ _{(20), occurs in a diffusion layer. The most}

abundant particle in the solution starts partaking in this diffusion. This diffusion layer causes a complex diffusion impedance element Zdif f.

Particles in the solutions that don’t partake in the chemical reactions or diffusion at the surface of the electrode are also present. These particles form a layer on the electrode surface, causing a capacitance, the ’double layer capacitance’. Other occurrences that should be taken into account in this system, is adsorption and non-uniform distribution. By adsorption is meant that the particles in the solution can start to bind to the electrode surface, resulting in a layer on the electrode surface called the ’Stern layer’. The non uniform distribution of charges happens at all interfaces, and results in an electrochemical potential gradient. The electrode-electrolyte surface interface and the layer formations are shown in image 2.9.

Conducting electrodes can produce a polarization effect when in contact with an ionic solution. This occurs at low frequencies. This happens as an electrochemical double layer is formed because of the distribution of the ions in the liquid and the solid (the sample and the electrode surface) in an electric field. This influences the ions in the solution and could attract them towards the electrode surface, therefore inducing an additional current. This electrode polarization has an influence on the current response in frequencies under 1MHz. (24).

(24)

2.9 Electrode-electrolyte interface 17

Figure 2.13: This figure shows the electrode electrolyte interface. It shows the Stern layer, the diffuse layer and then the Bulk electrolyte inbetween the detectors (9).

Sweat and other body fluids consist of NaCl ions. These ions from NaCl must be taken into account when measuring glucose. Adding NaCl leads to more ions in the solution. This causes less resistance and the creating of a dielectric. NaCl was added to the solutions. When NaCl is added to the solution, the ion concentration increases. Because of this, more electrons can be transferred. Adding NaCl ions to the solutions makes the artificial sweat more similar to actual human sweat.

(25)

18

3 Methodology and material

and materials In this experiment, the measurements were performed by the Autolab PGSTAT302N. This system consists of a potentiostat that controls the electrode cell. A potentiostat maintains a potential on the working electrode over the reference electrode. The Autolab was then connected to the computer. The measurements were converted and processed by the software system NOVA (2.1.2). In this system the FRA and the OCP were measured and shown.

3.1 Dropsens Inter-digitated Electrode

In this setup the 3 point electrode mode of the connector was used on the Dropsens Inter-digitated Electrode, the IDE. This IDE consist of a gold Working Electrode with a 4mm diameter, gold Counter Electrode and a silver Reference Electrode. These were printed on the one end of a ceramic substrate (5mm by), with the other end 3 printed silver electric contacts (25). This electrode is shown in figure 3.1.

Figure 3.1: This figure shows the Dropsens Inter-digitated Electrode (IDE), consisting of a carbon working and counter electrode, a silver reference electrode, and a 3 point connector printed on a ceramic substrate. In this research a gold working and counter electrode was used, rather than a carbon one.

3.2 Connector

There were two different types of connectors used in the execution of this experiment. At the beginning of the experiment connector type A was used. Here, the electrode is slid into the connector and pressed against the surface by putting a piece of plastic in the space that is left over on the bottom of the electrode. In this way the working surface of

(26)

3.2 Connector 19

three point electrode touches the connector resulting in a current.

After, connector type B was used. In this connector, the electrode is slid in and then from the bottom of the electrode screwed tight onto the conducting points of the connector.

Figure 3.2: In this figure connector type A is shown. On the left the connector is shown from above, and on the right the connector can be seen from the front. The image on the right shows the opening in which the electrode is put.

Figure 3.3: In this figure connector type B is shown. On the left the connector is shown from above, and on the right the connector can be seen from the front. The image on the right shows the opening in which the electrode is put. On the top the screw can also be seen that has to be tightened once the electrode is put in this connector.

The main difference between connector type A and connector type B, is the fact that with connector B the conducting pins of the connector are screwed on tight to the 3 point silver contacts of the electrode. With connector A however, the conducting surfaces are just touching. This results in less of a resistance using connector B. For this reason, connector B was used throughout the rest of the experiment. Another reason that connector B

(27)

20 3.3 Solutions

showed better results was the fact that connector A was possibly oxidized. On top of this it was also not in direct connection with the Autolab.

3.3 Solutions

To be able to measure the difference in Impedance of various glucose concentrations, a number of solution samples were prepared. First, a solution of NaCl in MiliQ water was created by adding 0.5g NaCl to a flask filled up with 100ml of MiliQ water, resulting in a 100ml 5wt% solution. This NaCl solution was then added to 10ml test tubes containing different concentrations of glucose. The MiliQ water was obtained from the Milli-Q Reference Water Purification System from Millipore. This water is purified by first reaching a resistivity of 18.2 Mω .cm at a temperature of 25 C with a TOC value below 5ppb. The water was then once more sent through a loop to remove all contaminants (26).

To start, 0.00, 0.01, 0.05, 0.1, 0.2, 0.4 and 0.8g of D-glucose were added to each tube and then filled up with 10ml of the NaCl stock. They were then labelled and kept closed off with a cap at all times to minimalize the chance of contamination, except for during the measurement. Later, different solutions were made to look at even lower concentrations of glucose. Again, a 5wt% NaCl solution stock in MiliQ water was created as described above. Then to 10ml tubes, 0.00, 0.01, 0.02, 0.05, 0.07, 0.1, 0.2, 0.4, 0.8g of D-glucose were added and filled with 10ml of NaCl solution.

In this experiment there was chosen for a stock solution of 5wt% NaCl in miliQ water. As discussed in section 2.11, this is done to increase the amount of free ions in the solution.

3.4 Boronic acid monolayer application

A boronic acid monolayer was applied on various electrodes. The impedance spectroscopy was first measured without the boronic acid layer.

The surface area of two gold electrodes were put in a solution with bis-boronic acid. This was done in a controlled environment inside the glove box in the chemical lab. These electrodes were left in the solution overnight for approximately 24 hours. This was done to make sure that the electrode surface would be fully covered with a monolayer of boronic acid. After these 24 hours, the electrodes were taken out of the boronic acid solution.

(28)

3.5 Measurements 21

The electrode was then cleaned by submerging it in a petri dish filled with methanol and leaving it in for 5 minutes. This process was repeated for a petri dish filled with ethanol. Finally, the electrode was washed down with MiliQ water and dried afterwards.

3.5 Measurements

The measurements for this experiment were performed in the D-lab. One set of measurements with the a single electrode was performed on the same day. First, the electrode was dipped into the tube with a certain glucose concentration. The electrode was then put into the tube sufficiently deep, so that the entire working electrode surface is in the solution. Before the actual measurements were done, a test measurement with a dummy cell was performed. This was done to check if the cables were in order and properly connected. Because the dummy cell has known values, it can easily be determined if there is anything wrong with the equipment by comparison to the outcome of the measurements. This was not the case.

After this the measurement started, where first the Open circuit potential (OCP) and then the Frequency Response Analyzer (FRA) was measured. The OCP is a measurement in which the counter electrode is bypassed. It is therefore a passive measurement. The OCP measures the difference in resting electrical potential between the reference electrode and the working electrode. In this way the stability of the system can be determined. As the name suggests, the FRA analyzes the frequency response. The FRA is an active measurement that measures the gain and phase of a system in reaction to applied frequencies.

The different glucose concentrations in the solutions were measured in a random order, to prevent creating and seeing a trend because of carrying out the measurements in order of concentration. For each measurement the temperature was written down. There was little fluctuation in the temperature during the whole set of one measurement.

In NOVA 2.1.1 the measuring of the solutions is done in various steps that can be chosen and added yourself. The impendance is measured in a frequency regime up to 1 MHz. The OCP and FRA measurements were done twice in a row for each concentration. The average value was taken after.

(29)

22

4 Results and discussion

In the first section the models are shown that were used to fit to the data to get a better understanding of the occurrences influencing the impedance outcome. The Nyquist profile obtained from the measurements with the first detector are shown in section 2. The following two detectors had similar results which can be found in the appendix. In the second section the graphs are shown for the linear fits of the data. Section three outlines the different values that have been obtained from the research and from the different graphs. The values and the possible justifications for these values are discussed here as well.

4.1 Nyquist profile with and without boronic acid

In this section the Nyquist representations of the data sets of the same detector is shown. The Nyquist graphs of the other used detectors can be found in the Appendix.

Figure 6.3 shows the Nyquist representation for a measurement set with detector A1. The detector was not covered with boronic acid in this measurement set.

(30)

4.1 Nyquist profile with and without boronic acid 23

Figure 4.1: This image shows the Nyquist representation for a measurement set with detector A1. The graph shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graph at high frequencies is zoomed in to tell apart the different concentrations. On the right the concentrations of glucose in mg/dL are shown.

Figure 6.4 shows the Nyquist representation for detector A1 without boronic acid for the top graph and with boronic acid for the bottom graph.

(31)

24 4.1 Nyquist profile with and without boronic acid

Figure 4.2: This image shows the Nyquist representations for two measurement sets with detector A1. The top graph shows the measurements done without boronic acid. The bottom graph shows the measurements done with boronic acid. The graphs shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graphs at high frequencies are zoomed in to tell apart the different concentrations. On the right the concentrations of glucose in mg/dL are shown.

(32)

4.2 Linear fit 25

The nyquist plot starts at a high frequency showing the end of the semi-circle. As the frequency goes down the curve reaches a minimum. Looking closely at the graph you can then see a loop before the imaginary impedance increases again. It is unclear where this non-linear behaviour came from. It could be because of processes and electrochemical reactions occurring due to corrosion of the electrode. The graph then continues to go up linearly, eventually decreasing in slope ending up at the same value. In the nyquist plot there is no visible semi circle that is shown in section 3.4. This is because the frequency only goes up to 1MHz and stops there. There is an increase in Real Impedance and slightly in imaginary impedance. This shows that the glucose does influence the impedance. The semicircle represents the internal resistance, the resistance of the electrolyte. After the minimum the values go up representing the double layer resistance and the charge transfer resistance. For both the semicircle and after the minimum the Impedance increases with the concentration of glucose, especially for higher concentrations. The nyquist profiles can be looked at to see if there is a trend in the impedance for the increase of glucose concentration. When looking at the higher concentrations of glucose, 2000-8000, a clear trend can be seen. In each graph a clear shift in impedance is shown for a higher glucose level. However, looking at the lower concentrations of glucose, mainly 0-1000, no real trend is visible. For different levels of glucose, the impedance values in the graphs seem to be randomly distributed. The Autolab is unable to tell apart glucose in such low concentrations in these data sets.

In the appendix the Bode plots can be found.

4.2 Linear fit

In this section the graphs of the real part of Z plotted against the glucose concentration of each data set are shown. Each graph was linearly fitted. The script that was used to analyze the data in MATLAB and find the best linear fit can be found in the appendix. The graphs that came out of this script to find the frequency with the highest adjusted R2 are also added to the appendix. The corresponding slope, adjusted R2 and calculated

limit of detection are mentioned with each linear fit.

4.3 Linear fit comparison with and without boronic acid

In figure 4.4 the linear fits through the measured data are shown. These measurements were both done without boronic acid. For the first measurement set, the upper graph, the

(33)

26 4.3 Linear fit comparison with and without boronic acid

adjusted R2 _{is 0.94. The limit of detection is 728.3. As for the second set of measurements,}

the lower graph in figure 4.4, the adjusted R2 _{is 0.88. The limit of detection is 735.4. The}

slope of the two graphs in figure 4.4 are 0.0016 and 0.0015, respectively. The slope of the linear fit in the graph of figure 4.5 is 0.0016.

Figure 4.3: In this figure two graphs are shown with the real Z plotted against the glucose concentration. The red line represents a linear fit through the data points. This data is measured without the addition of boronic acid.

(34)

4.4 Linear fit comparison without boronic acid 27

with a boronic acid monolayer. Here, the adjusted R2 _{is 0.96 and the limit of detection}

465.6.

Figure 4.4: In this figure the real Z is plotted against the glucose concentration. The red line represents a linear fit through the data points. This data is measured with boronic acid.

4.4 Linear fit comparison without boronic acid

In figure 4.5 the linear fits through the data measured with a detector where both measurements were done without boronic acid are shown. For the first measurement set, the upper graph, the adjusted R2 _{is 0.94. The limit of detection is 701.5. As for the}

second set of measurements, the lower graph in figure 4.4, the adjusted R2 _{is 0.96. The}

limit of detection is 782.1. The slope of the two graphs in figure 4.4 are 0.0019 and 0.0016, respectively. The slope of the linear fit in the graph of figure 4.5 is 0.0016.

(35)

28 4.4 Linear fit comparison without boronic acid

Figure 4.5: In this figure two graphs are shown with the real Z plotted against the glucose concentration. The red line represents a linear fit through the data points. This data was measured without the addition of boronic acid.

4.4.1 Linear fit comparison with and without boronic acid

In this section the linear fits through the real part of Z are shown against the glucose concentration. The graphs in 2.7 show the measurements done with without boronic acid. The graphs in figure 2.8 show the measurements done after the detector was covered with

(36)

boronic acid. The limit of detection for the graph shown in 2.7 is 182.06 and it has an adjusted R2 _{of 0.993. Finally, it has a slope of 0.0016. The limit of detection for the graph}

shown in figure 2.8 is 284.25 and the adjusted R2 _{is 0.985. The slope of the linear fit is}

0.00155. In figure 2.7 and 2.8, the lower graph represents the graph zoomed in with limits 68 to 71 on the y-axis and -100 to 1100 on the x-axis.

(37)

30 4.4 Linear fit comparison without boronic acid

Figure 4.6: In this figure two graphs are shown with the real Z plotted against the glucose concentration. The red line in the top graph represents a linear fit through the data points. This data is measured without the addition of boronic acid. The graph on the bottom shows a zoomed in version with the limits -100 to 1100 on the x-axis and 68 to 71 on the y-axis.

(38)

Figure 4.7: In this figure two graphs are shown with the real Z plotted against the glucose concentration. The red line in the top graph represents a linear fit through the data points. This data is measured with a layer of boronic acid on the detector. The graph on the bottom shows a zoomed in version with the limits -100 to 1100 on the x-axis and 66 to 69 on the y-axis.

(39)

32 4.5 Comparison of impedance signatures from bare electrodes and boronic-acidfunctionalized electrodes

4.5 Comparison of impedance signatures from bare electrodes

and boronic-acid functionalized electrodes

The table below shows all the values corresponding to the measurements of each detector, with and without boronic acid.

Figure 4.8: Here the Limit of detection, the Adjusted R2 and the slope are shown for each detector, with and without boronic acid.

In the table a clear overview of the different values are given. Here, the values of the measurements with and without the addition of boronic acid can be compared. For detector A1, one can see that the limit of detection decreases in the measurement where the detector is covered with boronic acid. The slope however stays similar in all data sets, with and without boronic acid. For detector B1 no boronic acid is added, yet the LOD varies greatly, as does the slope. Lastly, for detector C1, the LOD increases. The slope stays the same in each measurement. When taking a look at all the detectors together, it is clear that the slope of the linear fit is similar in each case.

Looking at these values, it seems that the addition boronic acid does not affect or improve the measurements. The change in LOD seems random or at least not any more different with the boronic acid than just with another measurement at a different time without the boronic acid. However, only three detectors where two were covered later covered in boronic acid are used here. It could be a possibility that these fluctuations in LOD seem random because there are not enough measurements done for a trend to be seen. To research this more measurements should be executed and analyzed.

Looking just at the slopes though, the values suggest that the boronic acid on the detectors has little effect or the effect is not seen. There can be different explanations for this.

• The detector surface was not completely or evenly covered in boronic acid. As a consequence of this the glucose would not have been evenly bound to the boronic acid on the surface and therefore not measured any more than when the glucose

(40)

33

was in the solution and not on the surface.

• The glucose did not properly bind to the boronic acid on the surface.

• Not enough measurements were carried out. It can be possible that more measurements have to be carried out for there to be an improvement visible after the addition of boronic acid. Maybe when more measurements are carried out, an improvement after adding boronic acid can be seen.

• The electrode was functionalized and the glucose did bind to the boronic acid. This did not show in an improvement of LOD because at low concentrations the glucose cannot be measured due to interferance of other processes that show in the impedance plots. For the higher concentrations of glucose (>2000mg/dL) the concentrations are so high that functionalization did not make a difference.

5 Outlook

There are various things that can be done in a future research to possibly improve the detection of glucose in solutions. First of all, more measurements can be carried out. Also, it would be good to do measurements at a frequency higher than 1MHz. This can for example be done with the MFIA rather than with the Autolab.

Besides this we must also look at possible explanations for the lack of trend or improvement after boronic acid addition mentioned before. First off, if it was the case that the detector was not fully covered in boronic acid, maybe the surface was not dipped in the solution long enough. In the future this time in which the surface is sitting in the solution could be increased to for example 48 hours instead of 24 hours. Lastly, more measurements have to be carried out to see if there is in fact an improvement that is not shown in this research due to a too low amount of measurements.

(41)

34

6 Conclusion

From this research impedance spectroscopy alone cannot be used to detect glucose concentrations 0-1000mg/dl in the frequency range 1Hz-1MHz. In the higher concentrations of glucose, 2000-8000mg/dL, a clear trend in impedance response can be seen. In the lower concentrations however the impedance response seems to be randomly distributed. There is a correlation between the concentration of glucose and the impedance response, granted that the concentration is sufficiently high.

In this research the detection of glucose is not improved by the addition of boronic acid. Each set of measurements show a different result with and without boronic acid on the electrode surface. The LOD for detctor A1 decreases, whereas the LOD for detector C1 increases after functionalizing the detector surface with boronic acid. The slope of the linear fit in the case without boronic acid and the case with boronic acid stay the same value of 0.0016. Therefore it seems to be the case here that boronic acid has no effect on the detection.

More measurements need to executed to see if there is in fact a correlation there that becomes clear with more data sets. Also more measurements need to be done comparing the solutions with added boronic acid.

(42)

References 35

(0)

References

[1] K. C. D. Shendurse, A., The Encyclopedia of Food and Health, Glucose: Properties and analysis, vol. 1. 2016.

[2] J. v. H. E. van Enter, B., “Challenges and perspectives in continuous glucose monitoring,” The Royal Society of Chemistry, no. 23, pp. 5032–5045, 2018.

[3] e. a. Bruen, D., “Glucose sensing for diabetes monitoring: Recent developments,” Sensors, 2017.

[4] G. Hall, D., “Structure, properties, and preparation of boronic acid derivatives. overview of their reactions and applications,” Boronic Acids, 2005.

[5] V. Yegnaraman, “Molecular-level functionalization of electrode surfaces. an overview,” Electrodics and Electrocatalysis Division, 1996.

[6] M. A. B.V., “Basic overview of the working principle of a potentiostat/galvanostat (pgstat) – electrochemical cell setup.”

[7] Gamry, “Introduction to electrochemical impedance spectroscopy.” Lecture, 2018. [8] C. Breitkopf, “Impedance spectroscopy.” University Lecture, 2012.

[9] e. a. Mei, B., “Physical interpretations of nyquist plots for edlc electrodes and devices,” The journal of physics and chemistry C., 2017.

[10] Gamry instruments, Basics of Electrochemical Impedance Spectroscopy. Gamry instruments, Warminster, Pennsylvania, 2010.

[11] N. Narain, “Evaluating impedance spectroscopy as a method for quantifying d-glucose concentrations in artificial sweat.” 2019.

[12] S. Vashist, “Non-invasive glucose monitoring technology in diabetes management: A review,” Analytica Chimica Acta, vol. 750, pp. 16–27, 2017.

[13] e. a. Geng, Z., “Noninvasive continuous glucose monitoring using a multisensor-based glucometer and time series analysis,” Natureresearch: Scientific reports, 2017. [14] T. Karabencheva, Advances in Protein Chemistry and Structural Biology, Chapter 3

- Mechanisms of protein circular dichroism: insights from computational modeling, vol. 80. Elsevier B.V., 2010.

(43)

36 References

[15] Sigma-Aldrich, “Product information sheet for d-()-glucose.” Sigma-Aldrich publications, 2019.

[16] S. J. T. D. L. Y. e. a. Fossey, J., “A bis-boronic acid modified electrode for the sensitive and selective determination of glucose concentrations,” Analyst, vol. 138, no. 23, pp. 7041—-7266, 2013.

[17] e. a. Daikuzono, “Impedance spectroscopy for monosaccharides detection using responsive hydrogel modified paper-based electrodes,” The Royal Society of Chemistry, no. 142, pp. 1133–1139, 2017.

[18] B. B. V. I. Sivaev, I., “Lewis acidity of boron compounds,” Elsevier, no. 270-271, pp. 75–88, 2014.

[19] B. K. Olarte, O., “Fractional models in electrical impedance spectroscopy data for glucose detection,” Elsevier, 2017.

[20] F. Lvovich, V., Impedance spectroscopy, vol. 112. John Wiley & sons, inc., 2012. [21] R. Macdonald, J., Impedance Spectroscopy Theory, Experiment, and Applications,

vol. 2. John Wiley & sons, inc., 2005.

[22] M. e. a. Singh, “Double layer capacitance of ionic liquids for electrolyte gating of zno thin film transistors and effect of gate electrode.,” Journal of Materials Chemistry, 2017.

[23] G. Webster, J., Encyclopedia of Medical Devices and Instrumentation, vol. 1. John Wiley & sons, inc., 2006.

[24] F. C. J. Gómez-Sánchez, J. A., “Description of corrections on electrode polarization impedance using isopotential interface factor,” Journal of Eelectrical Bioimpedance, 2012.

[25] Metrohm, “Screen-printed carbon electrodes.”

[26] Millipore, Milli-Q® Reference Water Purification System. Millipore Corporation, Billerica, MA., 2009.

[27] R Core Team, R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, 2017.

[28] WHO, “Global report on diabetes,” 2018.

[29] S. S. B. S. C. S. P. G. Turaa, A., “Impedance spectroscopy of solutions at physiological glucose concentrations,” Biophysical Chemistry, vol. 129, no. 4, pp. 235–241, 2007.

(44)

References 37

[30] W. A. Y. J. Scott, D., “Determining detection limits of aqueous anions using electrochemical impedance spectroscopy,” Journal of Analytical Science and Technology, 2017.

[31] H. K. C. S. C. B. C. H. K. Y. Park, J., “The correlation of the complex dielectric constant and blood glucose at low frequency,” Biosens. Bioelectron, vol. 19, no. 4, pp. 321–324, 2003.

[32] J. Anzai, “Recent progress in electrochemical biosensors based on phenylboronic acid and derivatives,” Elsevier, no. 67, pp. 737—-746, 2016.

[33] T. A. R. P. Sbrignadello, S., “Electroimpedance spectroscopy for the measurement of the dielectric properties of sodium chloride solutions at different glucose concentrations.,” Journal of Spectroscopy, 2012.

[34] W. e. a. Nery, E., “Electrochemical glucose sensing: Is there still room for improvement?,” Analytical Chemistry, 2016.

[35] e. a. Todd, C., “Towards non-invasive extraction and determination of blood glucose levels,” Bioengineering, 2017.

[36] G. Yoon, “Biosensors and bioelectronics: Dielectric properties of glucose in bulk aqueous solutions: Influence of electrode polarization and modeling,” Biosensors and Bioelectronics, 2011.

(45)

38

Appendix

A1

Bode plots

Figure A1.1: Here the bode plots are shown for detector A1. The top graph shows the phase against the frequency. The bottom graph shows the real impedance against the frequency.

(46)

A2 MATLAB 39

Figure A1.2: Here the bode plots are shown for detector A1. The top graph shows the phase against the frequency. The bottom graph shows the real impedance against the frequency. The top graph is without boronic acid and the bottom graph is with boronic acid.

A2

MATLAB

(47)

(48)

(49)

(50)

A2 MATLAB 43

(51)

44 A2 MATLAB

(52)

A2 MATLAB 45

(53)

46 A3 Nyquist plots

A3

Nyquist plots

A3.1

Detector B1

Figure A3.1: This image shows the Nyquist representations for two measurement sets with detector B1. Both of the measurements were done without boronic acid. The graphs shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graphs at high frequencies are zoomed in to tell apart the different concentrations

(54)

A3 Nyquist plots 47

A3.2

Detector C1

Figure A3.2: This image shows the Nyquist representations for two measurement sets with detector C1. The top graph is meaasured without boronic acid. The bottom graph is measured with boronic acid. The graphs shows the real part of the impedance of the x-axis, and the negative imaginary part of the impedance on the y-axis. The frequency left to right goes from high value to low value. One section of the graphs at high frequencies are zoomed in to tell apart the different concentrations