Microwave reflective biosensor for glucose level detection in aqueous solutions

Journal Pre-proof Microwave reflective biosensor for glucose level detection in aqueous solutions Amir Ebrahimi, James Scott, Kamran Ghorbani

PII:

S0924-4247(19)31167-7

DOI:

https://doi.org/10.1016/j.sna.2019.111662

Reference:

SNA 111662

To appear in:

Sensors and Actuators: A. Physical

Received Date:

8 July 2019

Revised Date:

4 September 2019

Accepted Date:

7 October 2019

Please cite this article as: { doi: https://doi.org/ This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier.

Microwave Reflective Biosensor for Glucose Level Detection in Aqueous Solutions Amir Ebrahimia,∗ , James Scotta and Kamran Ghorbania a School

of Engineering, RMIT University, Melbourne, VIC 3001

ABSTRACT

Keywords: Microwave biosensor Glucose sensor Metamaterials Microfluidic sensor

This article presents the design and analysis of a real-time non-invasive microwave microfluidic sensor for measuring glucose concentration in aqueous solutions. The sensor is made of an open-ended microstrip transmission line loaded with a complementary split-ring resonator (CSRR). The CSRR shows a very intense electric field concentration at resonance, which is highly sensitive to the dielectric sample loading. A microfluidic channel is designed to deliver the glucose solutions to the sensitive area of the device. By applying liquid samples to the channel, a resonance frequency shift is detectable in the reflection coefficient (𝑆11 ) of the device. This in turn leads to a change in the |𝑆11 |. Both of the frequency shift and Δ|𝑆11 | can be used to measure the glucose level in the solution. Mathematical models are developed based on the measurement results of the glucose-water solutions using the resonance frequency shift and Δ|𝑆11 |. The developed sensing models are then used for detecting the glucose levels down to physiological values using the designed biosensor. The results prove the potential compatibility of the proposed biosensor for human glycaemia monitoring.

oo

f

ARTICLE INFO

cies has shown a great potential in non-invasive characterization of aqueous solutions [11, 12, 13, 14, 15, 16, 17, 18]. Biosensors are fundamental components in nowadays medThe operation principle in most of these devices is based on ical and biological experiments and diagnostics. A majorthe resonators. Applying dielectric liquid samples to the senity of this type sensors measure the dose of various biositive areas of the resonators modifies the electric and magchemicals species in aqueous solutions [1, 2, 3]. Example netic fields distributions around them [19, 20, 21, 22]. This are bacteria growth monitoring sensors [4], biosensors for causes a shift in the resonance frequency of the device from blood cholesterol monitoring [5], uric acid detection in urine which the dielectric properties of the materials under test [6], etc. These examples show a vast requirement in designare determined. The introduction of metamaterials in recent ing biosensors for medical and biological aqueous samples years paved the way for designing more compact resonancedosing. based microwave sensors with high sensitivity [23, 24, 25, The blood glucose level is one of the most critical health 26, 27, 28]. At resonance, the metamaterial-based partifactors. In normal health, the blood glucose level is automatcles such as split-ring resonators (SRRs) and their compleically controlled through the insulin hormone. If insulin promentary counterparts produce highly dense electromagnetic duction is low, blood glucose can reach to dangerous levels fields concentrations, which are very sensitive to the matecausing hyperglycaemia. This condition greatly increases rials loading or geometry alterations [29, 30, 31, 32]. The the risk of a range of diseases, including heart disease, blindefforts on the design of microwave biosensors for glucose deness, gangrene and kidney disease. Glucose sensors are imtection show a promising potential of this approach in labelportant devices helping in monitoring and controlling the free and non-invasive detection [33, 34, 35, 36, 37, 38, 39]. blood glucose levels. A majority of commercial glucose The devices in [33, 34] are based on three-dimensional resmonitoring sensors are based on electrochemical methods onators, which renders their application in the integrated lab[7, 8, 9, 10], which offer selective glucose detection by uson-a-chip platform. There is no channel in [35, 36, 39] sening a chemical mediator. In spite of their high accuracy, this sors for controlling the amount of the liquid sample applied type of sensors are invasive to the sample under test. This to the sensing area. This might cause added measurement erlimits their application since only a single measurement can ror due to cross-sensitivity to the amount of the liquid sample be performed on each sample under test. More importantly, under test. The sensitivity of the devices in [37, 38] is limthe commercial glucose test strips are not reusable resulting ited for low concentration measurements. In addition, all of in high expense in strips investment [1]. these sensors are two port devices relying on transmission Dielectric spectroscopy at RF and microwave frequencoefficient measurement requiring a two port measurement ⋆ system. Corresponding Author. ⋆⋆ This work was performed in part at the RMIT Micro Nano Research Here, we propose a single port microwave-based microfluFacility (MNRF) in the Victorian Node of the Australian National Fabricaidic biosensor using a complementary split-ring resonator tion Facility (ANFF). (CSRR) loaded on an open-ended microstrip transmission [email protected] (A. Ebrahimi); line. A notch appears in the reflection coefficient (𝑆11 ) of [email protected] (J. Scott); [email protected] (K. the sensor at the resonance frequency of the CSRR. A miGhorbani) ORCID (s): 0000-0002-1787-2230 (A. Ebrahimi); 0000-0003-3200-6821 crofluidic channel is attached to the sensor for delivering the

Jo

ur n

al

Pr

e-

pr

1. Introduction

(J. Scott); 0000-0001-8767-0207 (K. Ghorbani)

Amir Ebrahimi et al.: Preprint submitted to Elsevier

Page 1 of 8

Microwave glucose biosensor

a

Top View

Bottom View

b

l1

d1 l2

d2

f

s

50 Ω

Z

g

Patch Resistor

S11

CSRR

CC

Microstrip Line

CR

SMA Port

e

d

V/m

32000

Outlet

20000

in CSRR

17 μm

Glucose Molecule

PDMS

8000

Cu

RO4350, 0.762 mm

16000 12000

CR

f

24000

Microfluidic channel

PDMS Cu

Inlet

28000

E-field Contour

RR

oo

c

LR

SMA Port

Channel

4000

pr

0

out of the ground plane of the sensor. The end section of the microstrip transmission line is extended to an square patch shape covering the CSRR in the ground plane for improving the coupling between the CSRR and the microstrip transmission line. A resistor is placed in the gap area of the CSRR resonator for impedance matching and having a high quality factor resonance, when the CSRR is excited using the microstrip transmission line. By applying a microwave signal to the input SMA connector, the CSRR will be excited through the electric field generated by the microstrip transmission line. At resonance, there is a high concentration of fringing electric field at the top side of the CSRR as shown in Fig. 1(c). This is the most sensitive area to the dielectric loading. Thus, the microfluidic channel is aligned to the top side of the CSRR as shown in Fig. 1(d). A crosssectional view of the microfluidic channel, when it is filled with the test liquid is shown in Fig. 1(e), when a test liquid is filling the channel. An equivalent circuit model of the sensor is presented in Fig. 1(f). In the circuit model, the 𝐿𝑅 , 𝐶𝑅 , and 𝑅𝑅 model the CSRR, which is etched out of the ground plane, the 50 Ω transmission line section models the microstrip transmission line and 𝐶𝐶 stands for the capacitive coupling between the microstrip transmission line and the CSRR. The resonance frequency of the device is

Jo

ur n

al

Pr

test aqueous samples to the sensing area and controlling the amount of sample loaded to the sensor. By applying liquid samples of various glucose concentrations to the channel, the resonance frequency is shifted with a change in |𝑆11 |. These parameter variations are used for determining the glucose concentration in the solution. The proposed device is a permittivity sensor. As demonstrated in [1, 35, 36], the permittivity of blood or blood serum is dependent on the glucose concentration. Thus the designed sensor can be used as a biosensor for glucose concentration measurements. Based on the results in [40, 41], temperature variation can cause errors in permittivity measurements of water solutions. The cross-sensitivity to the environmental factors and other solutes in the solution can be removed by using two sensors in a differential setup, where one acts as a reference. The designed sensor is a single port device that potentially requires a simpler measurement system if a differential measurement setup is required. In measurement using a two port system such as a vector network analyzer (VNA), an identical sensor can be used in port 2 as a reference for removing the crosssensitivity to the environmental factors such as temperature and humidity [20, 21].

e-

Figure 1: The designed microwave microfluidic biosensor. (a) Top view of the sensor. (b) Bottom view of the sensor. (c) Electric field distribution around the CSRR at resonance. (d) The sensor view, when the PDMS microfluidic part is attached. (e) Cross-sectional view of the microfluidic channel, when it is filled with a test solution. (f) Equivalent circuit model of the biosensor.

2. Materials and methods 2.1. Sensor design As shown in Fig. 1, the proposed microwave microfluidic sensor is made of an open-ended microstrip line loaded with a complementary split-ring resonator (CSRR) etched Amir Ebrahimi et al.: Preprint submitted to Elsevier

𝑓r =

1 . √ 2𝜋 𝐿𝑅 (𝐶𝑅 + 𝐶𝐶 )

(1)

At this frequency, the impedance seen from the combiPage 2 of 8

Microwave glucose biosensor

a

b Resistor

CSRR

Bottom view

Top view

c

oo

f

d VNA

PMMA Frame

Test samples

e-

Sensor

pr

Tubes

Pr

Figure 2: The fabricated microwave microfluidic biosensor. (a) Top view of the sensor. (b) Bottom view of the sensor. (c) The test setup for verification of the sensor performance. (d) The assembled sensor after attaching the PDMS microfluidic channel and the PMMA frame.

𝐿𝑅 (𝐶𝐶 + 𝐶𝑅 ) 𝑅𝑅 𝐶𝐶2

,

(2)

ur n

𝑍𝑓r =

al

nation of 𝐿𝑅 , 𝐶𝑅 , 𝑅𝑅 , and 𝐶𝐶 is purely real and equal to

Jo

This causes a notch in the reflection coefficient (𝑆11 ) of the sensor at resonance. The notch depth and quality factor is improved if the impedance is closer to 50 Ω. It is evident from (3) that the purely real value of 𝑍 at the resonance frequency is controlled by 𝑅R if the geometry of the sensor is unchanged. Thus, the value of 𝑅R can be optimized using the full-wave simulations in the CST Microwave Studio to achieve the best matching and maximum 𝑆11 notch depth at the resonance frequency (𝑓r ), when the channel is filled with water. The optimized value of 𝑅𝑅 is 100 Ω. Applying a dielectric liquid to the microfluidic channel modifies the value of 𝐶𝑅 capacitor, which in turn shifts the resonance frequency. This frequency shift is measured and used for characterization of the liquid under test.

2.2. Fabrication of the sensor The designed sensor is fabricated using a printed circuit board (PCB) process. A commercial low dielectric RO4350 laminate with 𝜀r = 3.66 and tan 𝛿 = 0.0037 from Rogers Amir Ebrahimi et al.: Preprint submitted to Elsevier

corporation is used as a substrate for fabrication of the sensor. The substrate has a thickness of 0.762 mm and 0.17 𝜇m copper cladding on both sides. The top microstrip and the CSRR in the ground plane are etched using a precise laser milling process. The CSRR has a square shape with side lengths of 𝑙1 = 𝑙2 = 9 mm and a gap width of 𝑠 = 0.2 mm. The thin metallic path connecting the CSRR capacitive patch to the ground plane has a width of 𝑔 = 0.2 mm. The width of the microstrip transmission line is 1.67 mm, and the patch dimensions are 𝑙1 = 6 mm and 𝑙2 = 7 mm. A SMA connector is soldered to the input of the sensor for excitation with a microwave source and measurement purposes. The top and bottom views of the fabricated sensor PCB are presented in Fig. 2(a) and (b) respectively. The microfluidic channel is made of polydimethylsiloxane (PDMS). A mold is prepared for channel fabrication on a silicon substrate using a thick photo-resist mask followed by chemical etching. In the next step, the PDMS is deposited into the mold and cured at 80◦ C. The channel is then peeled off and attached to the sensor. The width, height, and length of the fabricated channel are 0.7 mm, 70 𝜇m, and 13 mm respectively. The channel is aligned to the upper side of the CSRR, where there is a highly dense concentration of electric field for maximum sensing. In order to keep the channel Page 3 of 8

Microwave glucose biosensor

a

0 -10

|S 11| (dB)

in its position and avoiding misalignment during the measurements, a frame is machined using PMMA material for pushing the PDMS against the substrate and fixing the channel position. Microfluidic tubes are attached to the in/outlet of the channel for filling and draining the fluid samples. An assembled sensor prototype after attaching the microfluidic channel and PMMA frame is shown in Fig. 2(d).

-20 DI water 10 mg/mL glucose

-30

30 mg/mL glucose 50 mg/mL glucose

-40 2.40

3. Results and Discussion

70 mg/mL glucose

2.45

3.1. Samples preparation and test setup

al

3.2. Calibration of the sensor

Jo

ur n

In order to calibrate the sensor and developing a mathematical model for measuring the glucose concentration in aqueous solutions, the sensor is first tested using the liquid samples with 0, 10, 30, 50, and 70 mg/mL glucose concentrations. The measurements are performed using the procedure described above and the reflection coefficient is recorded for each sample. Fig. 3(a) shows the measured reflection coefficients of the biosensor for this set of the test samples. In order to have a high frequency resolution, the measurements are performed over the frequency span of 2.4−2.6 GHz with 801 data points. This results in 0.25 MHz frequency resolution in the measurements. As seen in Fig. 3(a), the measured curves can be easily discriminated from each other in different measurements because of a high quality factor (𝑄) of the sensor at resonance. This offers high sensitivity in amplitude and frequency readings and results in high accuracy measurements. The resonance frequency of the sensor shifts up by increasing the concentration of glucose in the solution. The frequency shift also causes a change in the reflection coefficient (𝑆11 ) level at a fixed frequency of 2.48 GHz, which is the resonance frequency for distilled water. Therefore, both of the resonance frequency shift (Δ𝑓𝑟 ) Amir Ebrahimi et al.: Preprint submitted to Elsevier

Fitted curve

f

30 20 10

20 40 60 Clucose concentration (mg/mL)

80

pr

0

oo

Freq. Shift (MHz)

Measurement

40

0

c

2.60

50

b

25

Measurement

Fitted curve

e-

Δ|S11|, dB

20 15 10

Pr

In order to verify the proposed sensing principle, distilled water-glucose test samples are prepared by mixing D(+)-Glucose powders from Sigma-Aldrich with distilled water in different concentrations. Nine sets of samples are prepared with concentrations ranging from 5 − 80 mg/mL. A photograph of the test setup for measurement and verification is presented in Fig. 2(c), where the sensor is connected to a vector network analyzer (VNA) for measuring the reflection coefficient (𝑆11 ). The measurements are preceded by a SOL calibration process on port 1 of the VNA, which sets the measurement reference plane at the input of the SMA connector. The glucose-water samples are then transferred to the microfluidic channel using a syringe attached to the tube connected to the in/outlet of the microfluidic channel. The measurements are performed using a flow-stop procedure, where the channel is first filled with the liquid test sample then, the flow is stopped for recording the associated 𝑆11 . Next, the channel is drained by pumping out the test liquid using the syringe. A very low pressure is applied to the syringe in each measurement to avoid channel deformation. Each measurement step is followed by filling and draining the channel with distilled water for washing and cleaning purposes.

2.50 2.55 Frequency (GHz)

5 0

0

20 40 60 Glucose concentration (mg/mL)

80

Figure 3: (a) The measured reflection coefficients of the biosensor for 0, 10, 30, 50, and 70 mg/mL concentrations of the glucose in the solution. (b) Resonance frequency shift versus the glucose concentration in the solution. (c) Reflection level variation (Δ|𝑆11 |) as a function of the glucose concentration.

and the 𝑆11 level can be used for sensing. The resonance frequency shift versus the glucose concentration is plotted in Fig. 3(b), where the distilled water is considered as a reference sample. In addition, the change of the 𝑆11 level with respect to the distilled water at 2.48 GHz versus the glucose concentration is shown in Fig. 3(c). The results show a linear relationship between the resonance frequency shift and the glucose concentration in the solution. However, the 𝑆11 level variation is a nonlinear function of the glucose concentration for the considered concentration ranges. The points in Fig. 3(b) and (c) are obtained by repeating each measurement for 10 times. Mathematical sensing models are developed for the designed biosensor using the regression analysis relating the glucose concentration to the measured resonance frequency shift and 𝑆11 level. The obtained sensing model

Page 4 of 8

Microwave glucose biosensor Table 1 A comparison between the actual concentrations and the measured glucose concentration using the biosensor.

5 20 40 60 80

Measurement using (3) Meas. 𝑓𝑟 Δ𝑓𝑟 Meas. (GHz) (MHz) Concentration (mg/mL) 2.4824 2.45 4.9 2.4902 10.2 20.4 2.4990 19.5 39 2.5110 31 62 2.5194 39.4 78.8

Error (%)

Concentration (mg/mL)

2.0 2.0 2.5 3.3 0.25

5 20 40 60 80

-20 5 mg/mL glucose 20 mg/mL glucose 40 mg/mL glucose

12 2.55 1.75 0.36 0.12

in (3) and (4). A comparison between the measured concentrations using the biosensor and the actual ones is presented in Table 1. The results in the table verify the accuracy of the designed sensor in detecting the glucose concentrations in aqueous solutions. Generally, the measurements based on the resonance frequency shift and (3) show a better accuracy for lower values of glucose concentrations. However, as the glucose concentration increases, the measurement based on Δ𝑆11 and (4) becomes more accurate. The reason is for low glucose concentrations, the amplitude variation is very sharp as shown in Fig. 3(c). This makes the amplitude measurements more vulnerable with respect to the cross factors such as temperature and increases the measurement error. But, as the concentration increases the slope of amplitude variation decreases giving a better stability and accuracy in the measurements. The maximum measurement error using (3) is 3.3% for 60 mg/mL concentration, whereas the maximum error using (4) is 12% for 5 mg/ml concentration. The measurement accuracy might be improved by using both of the frequency shift and Δ𝑆11 in a single high-order mathematical model. Further measurements have been carried out to examine the compatibility of the designed biosensor for hyperglycaemia screening applications. To this aim, measurements are performed using glucose concentrations at physiological levels of 1, 2, 3, 4, and 5 mg/mL. Fig. 5 presents the measured resonance frequency shift and Δ𝑆11 for these samples. Likewise, the measurements are repeated 10 times for each sample and the results are presented with error bars. As seen in Fig. 5(a) and (b), both of the frequency shift and Δ𝑆11 measurements show a linear trend for physiological glucose concentrations. The describing equation for the fitted dashed curve in Fig. 5(a) is (3), and the describing equation for the dashed line in Fig. 5(b) is

pr

60 mg/mL glucose

-40 2.40

Error (%)

oo

|S 11| (dB)

-10

-30

(4) Meas. Concentration (mg/mL) 4.37 19.49 39.3 60.22 79.9

f

0

Measurement using Meas. 𝑆11 Δ𝑆11 @ 2.48 GHz (dB) (dB) -29.63 2.98 -22.09 10.52 -16.54 16.07 -13.56 19.05 -11.65 20.47

80 mg/mL glucose

2.45

2.50 2.55 Frequency (GHz)

2.60

Pr

Figure 4: The measured reflection coefficients of the biosensor for 5, 20, 40, 60, and 80 mg/mL concentrations of the glucose in the solution.

e-

Concentration (mg/mL)

based on the frequency shift is 𝜌 (mg∕mL) = 2Δ𝑓𝑟 (MHz),

al

(3)

ur n

where 𝜌 is the glucose concentration in the solution and Δ𝑓𝑟 is the resonance frequency shift with respect to the reference sample that is the distilled water. This shows a linear relation with (𝑅2 = 0.995) between the resonance frequency shift and the glucose concentration. Furthermore, the obtained sensing model based on the 𝑆11 level and using a nonlinear regression analysis is

Jo

( ) Δ𝑆11 𝜌 (mg∕mL) = −29.976 ln 1 − , 22

(4)

with 𝑅2 = 0.9985.

3.3. Validation of the sensing models In order to verify the developed sensing model in (3) and (4), the measurements are performed with another set of test solutions with 5, 20, 40, 60, and 80 mg/mL glucose concentration. The reflection coefficients of the biosensor obtained from these measurements are plotted versus frequency in Fig. 4. The resonance frequency and 𝑆11 level at 2.48 GHz for each sample are recorded for calculating the glucose concentration based on the developed sensing model Amir Ebrahimi et al.: Preprint submitted to Elsevier

𝜌 (mg∕mL) = 1.666Δ𝑆11 (dB),

(5)

which is a linear approximation to (4) for small values of glucose concentrations. For evaluating the biosensor reliability for hyperglycaemia monitoring applications, the measured glucose concentrations are presented on Clarke error grid versus the actual concentration values. This grid is developed by Clarke on 1987 to quantify the clinical accuracy of the methods and Page 5 of 8

Microwave glucose biosensor Table 2 Comparisons with the sate-of-art microwave glucose sensors Ref.

Sensor Structure

Sensing method

Norm. 𝑓r Sensitivity |𝑆| sensitivity Conc. Range No. Ports per (mg/mL) per (mg/mL) (mg/mL)

[33] [34] [35] [36] [37] [38] This Work

Dielectric Resonator 3D SRR Resonator LC Resonator LC Resonator LC Resonator CSRR Resonator CSRR Resonator

𝑓r and |𝑆11 | 𝑓r 𝑓r 𝑓r |𝑆21 | |𝑆21 | 𝑓r and |𝑆11 |

2.0 × 10−5 12.4 × 10−3 26 × 10−3 16.7 × 10−3 N.A. N.A. 0.5 × 10−3

0 0

1 2 3 4 Glucose concentration (mg/mL)

5

5

A E

3

2

1

0 1 2 3 4 Glucose concentration (mg/mL)

5

al

0

ur n

Figure 5: (a) The measured resonance frequencies and (b) Δ𝑆11 for water-glucose solutions with physiological concentrations.

Jo

systems in monitoring the blood glucose level [42]. The 𝑥axis in the grid represents the actual value of the glucose concentration in the solutions under test, and 𝑦-axis is the measured concentrations values using the sensor. The measurement plane is divided into five regions. A is the optimal region with measurement error of less than 20% that is defined as an acceptable error in human glycaemia monitoring. In region B, the measurement error is more than 20% but it does not lead to inappropriate treatment. Regions C, D, and E are the points with high measurement errors that might endanger patient’s health. The results in Fig. 6 verify the potential of the proposed biosensor for glycaemia monitoring as all the measured values fall within region A of the Clarke grid for both of the sensing principles based on the resonance frequency shift and Δ𝑆11 .

Amir Ebrahimi et al.: Preprint submitted to Elsevier

A

B

2

D

1

B C

0

4

E

A

C B

D B C

0 0

5

A

B

2 1

D

E

5

3

B

1 2 3 4 Actual glucose concentration (mg/mL)

e-

b

C

f

4

0

Pr

|ΔS11|, dB

b3

2 2 2 2 2 2 1

oo

1

0 − 300 0 − 36 0 − 0.72 0−5 0−5 0 − 400 0−5

pr

2

Meas. concentration (mg/mL)

a

Meas. concentration (mg/mL)

Freq. Shift (MHz)

a3

0.004 dB N.A. N.A. N.A. 0.008 dB 0.009 dB 0.5 dB

D E

1 2 3 4 Actual glucose concentration (mg/mL)

5

Figure 6: Clarke error grid obtained by measuring solutions with 1, 2, 3, 4, and 5 mg/mL glucose concentrations. (a) Measurements based on the resonance frequency shift. (b) Measurements based on |Δ𝑆11 | .

4. Comparisons with other sensors In order to have a better insight to the performance of the proposed glucose biosensor, comparisons are performed in Table 2 with the state-of-art microwave-based glucose sensors. For having a fair comparison, the 𝑓r sensitivity is normalized to the resonance frequency of a bare sensor in each case. The measured concentration range for each sensor is presented in Table 2. The calculated sensitivity values in the table are the average values across the concentration range. The sensing method in some of the sensors is based on the resonance frequency, and the others use 𝑆11 or 𝑆21 variations for sensing. Furthermore, the sensor in this work and [33] use both 𝑓r and 𝑆11 for sensing. The sensor in [33] is based on a dielectric resonator for sensing, that shows the lowest sensitivity in terms of 𝑓r and |𝑆| among the others. Page 6 of 8

Microwave glucose biosensor

ur n

al

f

oo

Pr

In summary, we demonstrated a glucose concentration measurement method using a reflective microwave biosensor. The sensor offers a non-invasive characterization technique based on the dielectric properties of the test aqueous solutions. Applying the solutions with various glucose concentrations into the microfluidic channel integrated to the biosensor, modifies the resonance frequency and |𝑆11 |. The shift in these parameters is detected by measuring the reflection coefficient 𝑆11 for detecting the glucose level in the solution. Mathematical sensing models are developed for glucose concentration detection using these parameters. The biosensor is also used for measuring the glucose concentrations down to physiological levels and interpretation of the measured results on the Clarke grid proves its potential for blood glycaemia monitoring. The one port structure of the sensor allows a simple measurement setup and provides the possibility of using a second sensor in two port measurement systems as a reference for differential measurements and improving the accuracy.

1. Chien, J.C., Yeh, E.C., Lee, L.P., Anwar, M., Niknejad, A.M.. A microwave reconfigurable dielectric-based glucose sensor with 20 mg/dL sensitivity at sub-nL sensing volume in CMOS. In: IEEE MTT-S International Microwave Symposium. 2015:doi:10.1109/ MWSYM.2015.7167141. 2. Chretiennot, T., Dubuc, D., Grenier, K.. Double stub resonant biosensor for glucose concentrations quantification of multiple aqueous solutions. In: IEEE MTT-S International Microwave Symposium (IMS2014). 2014:doi:10.1109/MWSYM.2014.6848570. 3. Ebrahimi, A., Withayachumnankul, W., Al-Sarawi, S.F., Abbott, D.. Microwave microfluidic sensor for determination of glucose concentration in water. In: IEEE 15th Mediterranean Microwave Symposium (MMS). 2015:doi:10.1109/MMS.2015.7375441. 4. Ghafar-Zadeh, E., Sawan, M., Chodavarapu, V.P., Hosseini-Nia, T.. Bacteria growth monitoring through a differential cmos capacitive sensor. IEEE transactions on biomedical circuits and systems 2010;4(4):232–238. 5. Li, G., Liao, J., Hu, G., Ma, N., Wu, P.. Study of carbon nanotube modified biosensor for monitoring total cholesterol in blood. Biosensors and Bioelectronics 2005;20(10):2140–2144. 6. Kanyong, P., Pemberton, R., Jackson, S., Hart, J.P.. Development of a sandwich format, amperometric screen-printed uric acid biosensor for urine analysis. Analytical biochemistry 2012;428(1):39–43. 7. Liu, M., Liu, R., Chen, W.. Graphene wrapped Cu2O nanocubes: non-enzymatic electrochemical sensors for the detection of glucose and hydrogen peroxide with enhanced stability. Biosensors and Bioelectronics 2013;45:206–212. 8. Qiu, J.D., Zhou, W.M., Guo, J., Wang, R., Liang, R.P.. Amperometric sensor based on ferrocene-modified multiwalled carbon nanotube nanocomposites as electron mediator for the determination of glucose. Analytical biochemistry 2009;385(2):264–269. 9. Tang, H., Chen, J., Yao, S., Nie, L., Deng, G., Kuang, Y.. Amperometric glucose biosensor based on adsorption of glucose oxidase at platinum nanoparticle-modified carbon nanotube electrode. Analytical Biochemistry 2004;331(1):89–97. 10. Yan, Q., Peng, B., Su, G., Cohan, B.E., Major, T.C., Meyerhoff, M.E.. Measurement of tear glucose levels with amperometric glucose biosensor/capillary tube configuration. Analytical chemistry 2011;83(21):8341–8346. 11. Withayachumnankul, W., Jaruwongrungsee, K., Tuantranont, A., Fumeaux, C., Abbott, D.. Metamaterial-based microfluidic sensor for dielectric characterization. Sensors and Actuators A: Physical 2013;189:233–237. 12. Ebrahimi, A., Withayachumnankul, W., Al-Sarawi, S., Abbott, D.. High-sensitivity metamaterial-inspired sensor for microfluidic dielectric characterization. IEEE Sensors Journal 2014;14(5):1345–1351. 13. Grenier, K., Dubuc, D., Poleni, P.E., Kumemura, M., Toshiyoshi, H., Fujii, T., Fujita, H.. Integrated broadband microwave and microfluidic sensor dedicated to bioengineering. IEEE Transactions on microwave theory and techniques 2009;57(12):3246–3253. 14. Velez, P., Grenier, K., Mata-Contreras, J., Dubuc, D., Martín, F.. Highly-sensitive microwave sensors based on open complementary split ring resonators (OCSRRs) for dielectric characterization and solute concentration measurement in liquids. IEEE Access 2018;6:48324–48338. 15. Chahadih, A., Cresson, P.Y., Hamouda, Z., Gu, S., Mismer, C., Lasri, T.. Microwave/microfluidic sensor fabricated on a flexible kapton substrate for complex permittivity characterization of liquids. Sensors and Actuators A: Physical 2015;229:128–135. 16. Vélez, P., Muñoz-Enano, J., Grenier, K., Mata-Contreras, J., Dubuc, D., Martín, F.. Split ring resonator-based microwave fluidic sensors for electrolyte concentration measurements. IEEE Sensors Journal 2018;19(7):2562–2569. 17. Ebrahimi, A., Scott, J., Ghorbani, K.. Ultra-high sensitivity microwave sensor for microfluidic complex permittivity measurement. IEEE Transactions on Microwave Theory and Techniques 2019;doi:10.1109/TMTT.2019.2932737.

pr

5. Conclusion

References

e-

The sensor in [34] uses a 3D split-ring resonator (SRR) for detection and shows a high 𝑓r sensitivity however, the 3D structure limits its potential application in integrated lab-ona-chip applications. The sensors in [35, 36, 37] are based on semi-lumped LC resonators, where the one in [35] shows the highest 𝑓r among the others. There is no microfluidic channel in [35, 36], which might potentially increase the cross-sensitivity of the device to the volume of the sample under test. The device in [38] uses a microstrip-line-coupled CSRR integrated with a microfluidic channel for glucose detection. The sensing in [38] is based on a two port |𝑆21 | measurement showing a moderate sensitivity to the variations in the glucose concentration. The designed sensor in this work is the only device offering a single port measuring capability. Although, it is the third best device among the other in terms of 𝑓r sensitivity, it offers the highest |𝑆| sensitivity in comparison with the state-of-art sensors making it a very competitive choice compared with the microwave glucose biosensors proposed in the literature.

Jo

Acknowledgements

The authors would like to thank Dr Khashayar Khoshmanesh for the useful discussions and comments. We also would like to thank Mr David Welch for his helps is the fabrication and assembling the final sensor prototype.

CRediT authorship contribution statement Amir Ebrahimi: Design, simulation, fabrication, and measurement of the sensor. A. Ebrahimi prepared the first draft of the manuscript. James Scott: Analyzed the data and contributed to the general concept of the sensor. Kamran Ghorbani: Analyzed the data, supervised the research and contributed to the general concept. Amir Ebrahimi et al.: Preprint submitted to Elsevier

Page 7 of 8

Microwave glucose biosensor

37.

38.

39.

40.

e-

pr

42.

oo

f

41.

robust radio frequency biosensor using microwave resonator by integrated passive device technology for quantitative detection of glucose level. Biosensors and Bioelectronics 2015;67:687–693. Chretiennot, T., Dubuc, D., Grenier, K.. Microwave-based microfluidic sensor for non-destructive and quantitative glucose monitoring in aqueous solution. Sensors 2016;16. doi:10.3390/s16101733. Jang, C., Park, J.K., Lee, H.J., Yun, G.H., Yook, J.G.. Temperaturecorrected fluidic glucose sensor based on microwave resonator. Sensors 2018;18(11). Adhikari, K.K., Kim, N.Y.. Ultrahigh-sensitivity mediator-free biosensor based on a microfabricated microwave resonator for the detection of micromolar glucose concentrations. IEEE Transactions on Microwave Theory and Techniques 2016;64(1):319–327. Abduljabar, A.A., Hamzah, H., Porch, A.. Double microstrip microfluidic sensor for temperature correction of liquid characterization. IEEE Microwave and Wireless Components Letters 2018;28(8):735– 737. doi:10.1109/LMWC.2018.2849218. Abduljabar, A.A., Clark, N., Lees, J., Porch, A.. Dual mode microwave microfluidic sensor for temperature variant liquid characterization. IEEE Transactions on Microwave Theory and Techniques 2017;65(7):2572–2582. doi:10.1109/TMTT.2016.2647249. Clarke, W.L., Cox, D., Gonder-Frederick, L.A., Carter, W., Pohl, S.L.. Evaluating clinical accuracy of systems for self-monitoring of blood glucose. Diabetes care 1987;10(5):622–628.

Jo

ur n

al

Pr

18. Jafari, F.S., Ahmadi-Shokouh, J.. Reconfigurable microwave SIW sensor based on pbg structure for high accuracy permittivity characterization of industrial liquids. Sensors and Actuators A: Physical 2018;283:386–395. 19. Ebrahimi, A., Scott, J., Ghorbani, K.. Transmission lines terminated with LC resonators for differential permittivity sensing. IEEE Microwave and Wireless Components Letters 2018;28(12):1149–1151. 20. Ebrahimi, A., Scott, J., Ghorbani, K.. Differential sensors using microstrip lines loaded with two split-ring resonators. IEEE Sensors Journal 2018;18(14):5786–5793. 21. Su, L., Mata-Contreras, J., Vélez, P., Martín, F.. Splitter/combiner microstrip sections loaded with pairs of complementary split ring resonators (CSRRs): Modeling and optimization for differential sensing applications. IEEE Transactions on Microwave Theory and Techniques 2016;64(12):4362–4370. 22. Naqui, J., Damm, C., Wiens, A., Jakoby, R., Su, L., MataContreras, J., Martín, F.. Transmission lines loaded with pairs of stepped impedance resonators: Modeling and application to differential permittivity measurements. IEEE Transactions on Microwave Theory and Techniques 2016;64(11):3864–3877. 23. Withayachumnankul, W., Jaruwongrungsee, K., Fumeaux, C., Abbott, D.. Metamaterial-inspired multichannel thin-film sensor. IEEE Sensors Journal 2012;12(5):1455–1458. 24. Ebrahimi, A., Withayachumnankul, W., Al-Sarawi, S.F., Abbott, D.. Metamaterial-inspired rotation sensor with wide dynamic range. IEEE Sensors Journal 2014;14(8):2609–2614. 25. Zarifi, M.H., Deif, S., Daneshmand, M.. Wireless passive RFID sensor for pipeline integrity monitoring. Sensors and Actuators A: Physical 2017;261:24–29. 26. Fayaz, M., Lashaki, M.J., Abdolrazzaghi, M., Zarifi, M.H., Hashisho, Z., Daneshmand, M., Anderson, J.E., Nichols, M.. Monitoring the residual capacity of activated carbon in an emission abatement system using a non-contact, high resolution microwave resonator sensor. Sensors and Actuators B: Chemical 2019;282:218– 224. 27. Zarifi, M.H., Thundat, T., Daneshmand, M.. High resolution microwave microstrip resonator for sensing applications. Sensors and Actuators A: Physical 2015;233:224–230. 28. Hamzah, H., Lees, J., Porch, A.. Split ring resonator with optimised sensitivity for microfluidic sensing. Sensors and Actuators A: Physical 2018;276:1–10. 29. Velez, P., Grenier, K., Mata-Contreras, J., Dubuc, D., Martín, F.. Highly-sensitive microwave sensors based on open complementary split ring resonators (OCSRRs) for dielectric characterization and solute concentration measurement in liquids. IEEE Access 2018;6:48324–48338. 30. Horestani, A.K., Abbott, D., Fumeaux, C.. Rotation sensor based on horn-shaped split ring resonator. IEEE Sensors Journal 2013;13(8):3014–3015. 31. Horestani, A.K., Naqui, J., Shaterian, Z., Abbott, D., Fumeaux, C., Martín, F.. Two-dimensional alignment and displacement sensor based on movable broadside-coupled split ring resonators. Sensors and Actuators A: Physical 2014;210:18–24. 32. Ebrahimi, A., Ahmed, A., Mapleback, B., Scott, J., Ghorbani, K.. Microstrip lines loaded with bandstop resonators for high resolution permittivity sensing. In: Asia-Pacific Microwave Conference (APMC). 2018:926–928. doi:10.23919/APMC.2018.8617344. 33. Kim, J., Babajanyan, A., Hovsepyan, A., Lee, K., Friedman, B.. Microwave dielectric resonator biosensor for aqueous glucose solution. Review of Scientific Instruments 2008;79(8):086107. 34. Choi, H., Naylon, J., Luzio, S., Beutler, J., Birchall, J., Martin, C., Porch, A.. Design and in vitro interference test of microwave noninvasive blood glucose monitoring sensor. IEEE transactions on microwave theory and techniques 2015;63(10):3016–3025. 35. Park, H., Yoon, H.S., Patil, U., Anoop, R., Lee, J., Lim, J., Lee, W., Jun, S.C.. Radio frequency based label-free detection of glucose. Biosensors and Bioelectronics 2014;54:141–145. 36. Kim, N., Dhakal, R., Adhikari, K., Kim, E., Wang, C.. A reusable

Amir Ebrahimi et al.: Preprint submitted to Elsevier

Page 8 of 8