A Novel Cortisol Immunosensor Based on a Hafnium Oxide/Silicon Structure for Heart Failure Diagnosis

Assessing cortisol levels in human bodies has become essential to diagnose heart failure (HF). In this work, we propose a salivary cortisol detection strategy as part of an easily integrable lab-on-a-chip for detection of HF biomarkers. Our developed capacitive immunosensor based on hafnium oxide (HfO2)/silicon structure showed good linearity between increasing cortisol concentration and the charge-transfer resistance/capacitance. Moreover, the developed biosensor was demonstrated to be highly selective toward cortisol compared to other HF biomarkers such as tumor necrosis factor (TNF-α) and N-terminal pro-brain natriuretic peptide (NT-proBNP). The precision of our developed biosensor was evaluated, and the difference between the determined cortisol concentration in saliva and its expected one is <18%.


Introduction
Heart failure (HF) is a cardiovascular chronic disease caused by structural abnormalities of the heart that make it unable to fill or pump out blood, resulting in lower delivery of oxygen. Many different factors such as cardiac abnormalities (e.g., structural abnormalities including stiffness of the ventricular chambers, functional abnormalities including mitral incompetence, etc.), coexisting conditions (e.g., diabetes, anemia, sleep apnea, etc.), or other factors (e.g., age, genetic influence, lifestyle, etc.) can be the cause of HF [1,2].
Cortisol (also called 11β, 17α, 21-trihydroxypregn-4-ene-3,20-dione) is a steroid hormone of the glucocorticoids family, whose chemical formula is C 21 H 30 O 5 ( Figure 1) and whose molar mass is 362.46 g/mol. Cortisol is a crucial glucocorticoid hormone that is produced by the zona fasciculate of the adrenal gland. It is well known as a "stress hormone", and it takes part in the regulation of various physiological functions including energy metabolism, electrolyte balance, blood pressure, and cognitive function [3]. In addition, it contributes to the homeostasis of the adrenal [4], cardiovascular [5,6], immune [7], and endocrine systems [8]. Moreover, it plays a key role in brain regions that are important for cognitive learning, retrieval, encoding, and memory consolidation [9]. This steroid hormone thus represents a potential biomarker for numerous pathological conditions and diseases [10], as well as a useful clinical indicator for relapse vulnerability in chronic alcohol use.
In humans, cortisol is secreted from the outside of the adrenal cortex of the adrenal gland. It is mainly released in response to stress and a low glucose concentration in the blood (hypoglycemia) [11][12][13]. Stress is a condition that releases glucocorticoids, such as cortisol, and catecholamines, such as adrenaline. These two types of molecules can have effects on memory. Indeed, short exposure improves memory, but prolonged exposure can damage the hippocampal cells, which plays an essential role in memory and space In humans, cortisol is secreted from the outside of the adrenal cortex of the adrena gland. It is mainly released in response to stress and a low glucose concentration in the blood (hypoglycemia) [11][12][13]. Stress is a condition that releases glucocorticoids, such as cortisol, and catecholamines, such as adrenaline. These two types of molecules can have effects on memory. Indeed, short exposure improves memory, but prolonged exposure can damage the hippocampal cells, which plays an essential role in memory and space navigation [14][15][16]. Stress can also lead to high production of cortisol and cause an allo static load, which can result in various modifications in the human body. Among these are, for example, mood disorders, anxiety disorders, diseases, obesity, or even atrophy o the nerve cells of the brain. An increase in reactive oxygen species (ROS) is linked to sev eral serious pathologies [17], for example, rheumatoid arthritis [18], neurodegenerative diseases [19], atherosclerosis [20], and heart diseases such as heart failure [17,21,22]. In addition, oxidative stress is an important pathophysiological pathway in the developmen and progression of heart failure [21,23,24], making cortisol a potential biomarker for hear failure. Cortisol plays an essential role in oxidative stress, which is the result of an imbal ance between the production of reactive (radical) oxygen species (ROS) and the body's defenses (in terms of antioxidant cellular capacities). Indeed, an increase in cortisol can cause an excess of oxidative stress in the human body and trigger the oxidation of many cells, making them less functional and leading to heart failure in a person [25]. Studies o However, the monitoring of cortisol levels must be carried out on the same person because the cortisol level is specific to each, depending on their circadian cycle [32], their diet [35], alcohol consumption [36], and quality of sleep, etc. There are different solutions to regulate the cortisol level in the human body. Thus, some of the factors reducing cortisol levels are, for example, taking magnesium after aerobic exercise [37], taking fish oil [38], music therapy [39], massage therapy [40], laughter and humor [41], and regular dancing [42], etc. Nowadays, cortisol is usually determined by enzyme immunoassay (EIA) [43,44], enzyme-linked immunoassay (ELISA), and radioimmunoassay (RIA) [45], as well as chromatographic techniques coupled with mass spectrometry (MS) or tandem MS/MS [46][47][48]. Commonly, the principal limitations of these steps lie in their high costs, the long run-time, and the requirement of sophisticated technical skills. Herein, a novel and highly selective approach is proposed for detecting cortisol. Similarly, Ben Halima's approach [49][50][51], a substrate based on hafnium oxide/silicon structure, was biofunctionalized by immobilizing an anti-cortisol antibody onto the ISFET surface after functionalization with 11-triethoxysilyl undecanal (TESUD) by a vapor-phase method in a saturated medium, exploiting the reaction between the aldehyde and the N-terminus of the antibodies. Electrochemical impedance spectroscopy (EIS) was used for analyte detection due to its ability to detect variations in resistance and capacitance induced by binding events, thus enhancing device sensitivity. The linear range, accuracy, precision, and limit of detection (LOD) of the biosensor were investigated to reach a preliminary validation of the device. Selectivity was confirmed by analyzing TNF-α and the gold standard HF biomarker N-terminal pro-brain natriuretic peptide (NT-proBNP) samples [52]. Finally, our cortisol concentration was quantified with our biosensor using the standard addition method (SAM) in real saliva samples. The obtained results prove our biosensor to be a promising tool for cortisol detection in saliva. To the best of our knowledge, this is the first biosensor that uses HfO 2 /silicon structures for cortisol detection. Due to its high K (dielectric constant), HfO 2 leakage is reduced, and the gate capacitance is enhanced when compared to SiO 2 . This type of structure has already been used for the sensitive detection of interleukin-10 [53]. The use of these materials from microelectronics will allow their integration on a silicon lab-on-a-chip.

Fabrication Technology of HfO 2 /Silicon Structures
A thin-film layer of hafnium oxide (HfO 2 ) high-κ dielectric was deposited by atomic layer deposition (ALD) on top of a thin silicon dioxide (SiO 2 ) layer grown thermally on silicon wafers. The 4-inch <100> p-type silicon (p-Si) wafers were first cleaned with a 5% of hydrofluoric acid (HF) solution to remove any remaining organic compounds, right before a controlled dry thermal oxidation process at 800 • C to grow a thin interfacial 30 Å SiO 2 layer. Following the ALD process, a chemical vapor deposition (CVD) technique based on selfterminating gas-substrate reactions was carried out at 200 • C in a Savannah-200 apparatus from Cambridge NanoTech (Cambridge MA, USA), where the wafer surface was exposed to alternating precursor pulses of tetrakis (dimethylamido) hafnium (TDMAH/N 2 ) and water vapor (H 2 O/N 2 ), forming a single monoatomic layer of HfO 2 per cycle. Therefore, it was considered that the growth per cycle (GPC) reached a thickness of 20 nm. Finally, a post-deposition annealing process (PDA) was carried out by rapid thermal annealing at different temperatures of 400 • C, 500 • C, and 600 • C, studying the effect on the distribution of interface trap charges and its consequent effect in the dielectric constant by characterizing the sample surface roughness with atomic force microscopy ( Agilent 5500 AFM) (Agilent Technologies, Palo Alto, CA, USA) in tapping mode, as well as characterizing the electrical capacitive behavior. Homogenization of the HfO 2 surface was obtained at 600 • C, as shown in Figure 2.
(TDMAH/N2) and water vapor (H2O/N2), forming a single monoatomic layer of HfO2 per cycle. Therefore, it was considered that the growth per cycle (GPC) reached a thickness of 20 nm. Finally, a post-deposition annealing process (PDA) was carried out by rapid thermal annealing at different temperatures of 400 °C, 500 °C, and 600 °C, studying the effect on the distribution of interface trap charges and its consequent effect in the dielectric constant by characterizing the sample surface roughness with atomic force microscopy ( Agilent 5500 AFM) (Agilent Technologies, Palo Alto, CA, USA) in tapping mode, as well as characterizing the electrical capacitive behavior. Homogenization of the HfO2 surface was obtained at 600 °C, as shown in Figure 2.

Functionalization of HfO2 Surface with Antibodies
For the fabrication of the biosensing platform, HfO2 was first functionalized using silane aldehyde. For this purpose, bare HfO2 substrates were cleaned by sonication in acetone, followed by thorough rinsing in ultrapure water. Surface activation of the HfO2 substrates was performed using a UV/Ozone Procleaner TM in order to create -OH groups at the hafnium oxide surface for grafting silane aldehyde. Afterwards, the substrates were thoroughly rinsed and sonicated in ultrapure water. The active HfO2 substrates (with -OH) were functionalized using TESUD using the vapor-phase method [50,51,[53][54]. Then, the substrates were placed in an oven at 100 °C. After baking, they were rinsed with absolute ethanol and dried with nitrogen. Subsequently, the functionalized substrate surface was incubated with anti-cortisol antibody (10 µg/mL) already diluted in PBS. Finally, the remaining aldehyde groups were blocked by a treatment with ethanolamine (1% v/v) in PBS buffer. This step is crucial to prevent any nonspecific bonding phenomenon at the detection stage of cortisol (see Figure 3).

Magnetic Nanoparticles Biofunctionalization
The magnetic nanoparticles were prepared with a magnetic core of Fe3O4 surrounded by styrene/DVB/ACPA polymers that included COOH as the terminal functional group [55]. Initially, they were washed twice with PBS buffer (pH 7.4) and subsequently the COOH group was activated using a mixture of EDC/NHS at 100 mM in PBS. A magnetic

Functionalization of HfO 2 Surface with Antibodies
For the fabrication of the biosensing platform, HfO 2 was first functionalized using silane aldehyde. For this purpose, bare HfO 2 substrates were cleaned by sonication in acetone, followed by thorough rinsing in ultrapure water. Surface activation of the HfO 2 substrates was performed using a UV/Ozone Procleaner TM in order to create -OH groups at the hafnium oxide surface for grafting silane aldehyde. Afterwards, the substrates were thoroughly rinsed and sonicated in ultrapure water. The active HfO 2 substrates (with -OH) were functionalized using TESUD using the vapor-phase method [50,51,53,54]. Then, the substrates were placed in an oven at 100 • C. After baking, they were rinsed with absolute ethanol and dried with nitrogen. Subsequently, the functionalized substrate surface was incubated with anti-cortisol antibody (10 µg/mL) already diluted in PBS. Finally, the remaining aldehyde groups were blocked by a treatment with ethanolamine (1% v/v) in PBS buffer. This step is crucial to prevent any nonspecific bonding phenomenon at the detection stage of cortisol (see Figure 3). on the distribution of interface trap charges and its consequent effect in the dielectric constant by characterizing the sample surface roughness with atomic force microscopy ( Agilent 5500 AFM) (Agilent Technologies, Palo Alto, CA, USA) in tapping mode, as well as characterizing the electrical capacitive behavior. Homogenization of the HfO2 surface was obtained at 600 °C, as shown in Figure 2.

Functionalization of HfO2 Surface with Antibodies
For the fabrication of the biosensing platform, HfO2 was first functionalized using silane aldehyde. For this purpose, bare HfO2 substrates were cleaned by sonication in acetone, followed by thorough rinsing in ultrapure water. Surface activation of the HfO2 substrates was performed using a UV/Ozone Procleaner TM in order to create -OH groups at the hafnium oxide surface for grafting silane aldehyde. Afterwards, the substrates were thoroughly rinsed and sonicated in ultrapure water. The active HfO2 substrates (with -OH) were functionalized using TESUD using the vapor-phase method [50,51,[53][54]. Then, the substrates were placed in an oven at 100 °C. After baking, they were rinsed with absolute ethanol and dried with nitrogen. Subsequently, the functionalized substrate surface was incubated with anti-cortisol antibody (10 µg/mL) already diluted in PBS. Finally, the remaining aldehyde groups were blocked by a treatment with ethanolamine (1% v/v) in PBS buffer. This step is crucial to prevent any nonspecific bonding phenomenon at the detection stage of cortisol (see Figure 3).

Magnetic Nanoparticles Biofunctionalization
The magnetic nanoparticles were prepared with a magnetic core of Fe3O4 surrounded by styrene/DVB/ACPA polymers that included COOH as the terminal functional group [55]. Initially, they were washed twice with PBS buffer (pH 7.4) and subsequently the COOH group was activated using a mixture of EDC/NHS at 100 mM in PBS. A magnetic

Magnetic Nanoparticles Biofunctionalization
The magnetic nanoparticles were prepared with a magnetic core of Fe 3 O 4 surrounded by styrene/DVB/ACPA polymers that included COOH as the terminal functional group [55]. Initially, they were washed twice with PBS buffer (pH 7.4) and subsequently the COOH group was activated using a mixture of EDC/NHS at 100 mM in PBS. A magnetic field was used to separate the nanoparticles from the storage solution. Anti-cortisol antibody (10 µL) at 100 µg/mL was added to the mixture and incubated with slow stirring at room temperature for 90 min. The nonreacted active carboxylic acid groups were blocked with bovine serum albumin (BSA) (0.1%) in PBS buffer for 30 min. The antibody-coated magnetic nanoparticles were then separated from the mixture, resuspended in 1 mL of PBS buffer, and used for the incubation. The procedure described above is summarized in Figure 4. body (10 µL) at 100 µg/mL was added to the mixture and incubated with slow sti room temperature for 90 min. The nonreacted active carboxylic acid groups were with bovine serum albumin (BSA) (0.1%) in PBS buffer for 30 min. The antibody magnetic nanoparticles were then separated from the mixture, resuspended in 1 PBS buffer, and used for the incubation. The procedure described above is summ in Figure 4.

Characterization of HfO2 Surface Using Contact Angle Measurement (CAM)
The HfO2 substrate was characterized using the contact angle measurement to follow each functionalization step by controlling the surface's hydrophilicity. T tact angle measurements were performed using ultrapure water. The dosing volu 5 µL with a dosing rate of 5µL/s.

Electrochemical Measurements with HfO2/Silicon Structures
Electrochemical impedance spectroscopy (EIS) is an efficient technique for gating interfacial properties on surface-modified working electrodes. Measuremen made on the HfO2/silicon structure with an aluminum backside contact (500 nm t the working electrode, with a counter platinum electrode CE, and an Ag/AgCl re electrode ( Figure 5). All measurements were made with freshly prepared PBS s that required 8 mL to fill the cell, while the analysis was performed inside a Farad A VMP3 multichannel potentiostat purchased from Biologic-EC-Lab was used. T liminary plot established the required potential on the HfO2/silicon structure: −1.5 sus Ag/AgCl, in the accumulation range for silicon. Starting from the lowest, inc cortisol concentrations were added, incubated for 30 min at 4 °C before imped measurements with the following conditions: EIS frequency ranging from 5 Hz to 1 and a sinus amplitude of 25 mV with polarization potential of −1.5 V-. The scan ti 26 s/scan. Data acquisition and analysis were accomplished using EC-Lab s V11. 30. The obtained EIS data were modeled by the EC-Lab software using the R ize + Simplex method. Here, randomize was fixed at 5000 iterations and the fit 5000 iterations.
Capacitive measurements were performed at 862 Hz, in a potential range fr V (accumulation range) to +0.5 V (inversion range).

Characterization of HfO 2 Surface Using Contact Angle Measurement (CAM)
The HfO 2 substrate was characterized using the contact angle measurement system to follow each functionalization step by controlling the surface's hydrophilicity. The contact angle measurements were performed using ultrapure water. The dosing volume was 5 µL with a dosing rate of 5 µL/s.

Electrochemical Measurements with HfO 2 /Silicon Structures
Electrochemical impedance spectroscopy (EIS) is an efficient technique for investigating interfacial properties on surface-modified working electrodes. Measurements were made on the HfO 2 /silicon structure with an aluminum backside contact (500 nm thick) as the working electrode, with a counter platinum electrode CE, and an Ag/AgCl reference electrode ( Figure 5). All measurements were made with freshly prepared PBS solution that required 8 mL to fill the cell, while the analysis was performed inside a Faraday cage. A VMP3 multichannel potentiostat purchased from Biologic-EC-Lab was used. The preliminary plot established the required potential on the HfO 2 /silicon structure: −1.5 V versus Ag/AgCl, in the accumulation range for silicon. Starting from the lowest, increasing cortisol concentrations were added, incubated for 30 min at 4 • C before impedimetric measurements with the following conditions: EIS frequency ranging from 5 Hz to 100 kHz and a sinus amplitude of 25 mV with polarization potential of −1.5 V-. The scan time was 26 s/scan. Data acquisition and analysis were accomplished using EC-Lab software V11.30. The obtained EIS data were modeled by the EC-Lab software using the Randomize + Simplex method. Here, randomize was fixed at 5000 iterations and the fit fixed at 5000 iterations.
Capacitive measurements were performed at 862 Hz, in a potential range from −0.8 V (accumulation range) to +0.5 V (inversion range).

Surface Characterization
To assess the effectiveness of the functionalization, contact angle measurements wer performed on bare HfO2, after activation by UV/Ozone, and after functionalization wit TESUD. A contact angle of 53.9 ± 3° revealed a slightly hydrophilic nature on bare HfO which agrees with Lee's values [53]. After surface oxidation with UV/ozone activation, th HfO2 surface became highly hydrophilic at 12.7 ± 5° due to the high amount of hydroxy groups on the surface. After the functionalization process with the TESUD, the contac angle increased again to 80.3 ± 2°. This hydrophobic character can be explained by th presence of TESUD hydrocarbon chains (see Figure 6).

Biosensor Calibration by EIS Measurements
Bare HfO2 substrates were cleaned and activated with TESUD, as previously de scribed. The anti-cortisol antibody was immobilized in a conventional three-electrode ce overnight at 4 °C, then rinsed with PBS. Following the immobilization of anti-cortisol an tibody onto the substrate, the antibody-modified HfO2/silicon structure was incubate with cortisol at different increasing concentrations.
Impedance results were presented as a Nyquist plot, as shown in Figure 7A.

Surface Characterization
To assess the effectiveness of the functionalization, contact angle measurements were performed on bare HfO 2 , after activation by UV/Ozone, and after functionalization with TESUD. A contact angle of 53.9 ± 3 • revealed a slightly hydrophilic nature on bare HfO 2 , which agrees with Lee's values [53]. After surface oxidation with UV/ozone activation, the HfO 2 surface became highly hydrophilic at 12.7 ± 5 • due to the high amount of hydroxyl groups on the surface. After the functionalization process with the TESUD, the contact angle increased again to 80.3 ± 2 • . This hydrophobic character can be explained by the presence of TESUD hydrocarbon chains (see Figure 6).

Surface Characterization
To assess the effectiveness of the functionalization, contact angle measurements were performed on bare HfO2, after activation by UV/Ozone, and after functionalization with TESUD. A contact angle of 53.9 ± 3° revealed a slightly hydrophilic nature on bare HfO2 which agrees with Lee's values [53]. After surface oxidation with UV/ozone activation, the HfO2 surface became highly hydrophilic at 12.7 ± 5° due to the high amount of hydroxyl groups on the surface. After the functionalization process with the TESUD, the contact angle increased again to 80.3 ± 2°. This hydrophobic character can be explained by the presence of TESUD hydrocarbon chains (see Figure 6).

Biosensor Calibration by EIS Measurements
Bare HfO2 substrates were cleaned and activated with TESUD, as previously described. The anti-cortisol antibody was immobilized in a conventional three-electrode cel overnight at 4 °C, then rinsed with PBS. Following the immobilization of anti-cortisol antibody onto the substrate, the antibody-modified HfO2/silicon structure was incubated with cortisol at different increasing concentrations.
Impedance results were presented as a Nyquist plot, as shown in Figure 7A.

Biosensor Calibration by EIS Measurements
Bare HfO 2 substrates were cleaned and activated with TESUD, as previously described. The anti-cortisol antibody was immobilized in a conventional three-electrode cell overnight at 4 • C, then rinsed with PBS. Following the immobilization of anti-cortisol antibody onto the substrate, the antibody-modified HfO 2 /silicon structure was incubated with cortisol at different increasing concentrations.
Impedance results were presented as a Nyquist plot, as shown in Figure 7A. An attempt at signal enhancement was made by an additional incubation of the electrode with the MNPs functionalized with the complementary antibody. An SEM image, obtained using FEI Quanta FEG 250 (Thermo Fisher Scientific Inc, Waltham, MA USA) (Figure 8), shows a homogenous distribution of MNPs on the functionalized HfO 2 surface. Nevertheless, the effect of the enhancement of the signal was very disappointing.
The Nyquist plots of HfO 2 modified with anti-cortisol antibody, followed by increasing cortisol concentrations, were fitted with a Randles equivalent circuit (insert in Figure 7A). Here, the components are as follows: Rs corresponds to the resistance of the electrolyte solution; C dl is the double-layer capacitance that is in parallel with R ct , which is the charge-transfer resistance; and Zw is the Warburg impedance. The values of the extracted parameters are presented in Table 1.

Biosensor Calibration by EIS Measurements
Bare HfO2 substrates were cleaned and activated with TESUD, as previously described. The anti-cortisol antibody was immobilized in a conventional three-electrode cell overnight at 4 °C, then rinsed with PBS. Following the immobilization of anti-cortisol antibody onto the substrate, the antibody-modified HfO2/silicon structure was incubated with cortisol at different increasing concentrations.
Impedance results were presented as a Nyquist plot, as shown in Figure 7A.  The Nyquist plots of HfO2 modified with anti-cortisol antibody, followed by increas ing cortisol concentrations, were fitted with a Randles equivalent circuit (insert in Figure  7A). Here, the components are as follows: Rs corresponds to the resistance of the electro lyte solution; Cdl is the double-layer capacitance that is in parallel with Rct, which is the charge-transfer resistance; and Zw is the Warburg impedance. The values of the extracted parameters are presented in Table 1. An increase in Rct (charge-transfer resistance) can be observed from R0 (no cortisol at 206 kΩ to 302 kΩ for a cortisol concentration of 2 ng/mL. The change in Rct demonstrate the biorecognition of the cortisol by the anti-cortisol antibody grafted on the HfO2 surface The Rct increased: 370 kΩ for 10 ng/mL; 450 kΩ for 15 ng/mL; 620 kΩ for 50 ng/mL. The ∼ 104 kΩ variations between the anti-cortisol antibody modified electrode and the first con  An increase in R ct (charge-transfer resistance) can be observed from R 0 (no cortisol) at 206 kΩ to 302 kΩ for a cortisol concentration of 2 ng/mL. The change in R ct demonstrates the biorecognition of the cortisol by the anti-cortisol antibody grafted on the HfO 2 surface. The R ct increased: 370 kΩ for 10 ng/mL; 450 kΩ for 15 ng/mL; 620 kΩ for 50 ng/mL. The ∼104 kΩ variations between the anti-cortisol antibody modified electrode and the first concentration of cortisol at 2 ng/mL demonstrate that the quantification limit of the immunosensor is 2 ng/mL. In Figure 7B (black), the relative variation of Rct versus the cortisol concentration plot produced a linear relationship ranging from 2 ng/mL to 50 ng/mL, with R 2 = 0.9722 with a slope of 0.03 (ng/mL) −1 . A cross-selectivity study was performed to assess the level of nonspecific binding. For this purpose, two other HF biomarkers were used, namely TNF-α and NT-proBNP, using the same conditions and concentrations as for cortisol. From Figure 7B, it can be seen that the biosensor is demonstrated to be highly selective towards cortisol when compared to both TNF-α and NT-proBNP. The sensitivity for TNF-α (green) is 0.0005 (ng/mL) −1 , and for NT-proBNP (Red) the sensitivity is 0.0026 (ng/mL) −1 . The immunosensor is 200 times more sensitive to cortisol than TNF-α, and 12 times more sensitive to cortisol than NT-proBNP.

Biosensor Calibration by Capacitive Measurements
Additionally, capacitive measurements were performed to characterize the semiconducting behavior of HfO 2 /silicon structures for each cortisol concentration. The biosensor was maintained in the electrochemical cell and incubated in PBS containing cortisol at different concentrations. The biosensor was then rinsed with PBS to remove any adsorbed proteins and analyzed afterwards by capacitance measurements using PBS as an electrolyte. This procedure of immunosensor incubation was carried out for all cortisol concentrations (2 ng/mL to 50 ng/mL). The detection of cortisol at various concentrations is shown in Figure 9A. Here, the Cs/Cmax-voltage curves show a shift in the positive direction with increasing cortisol concentrations, which confirms a flat-band voltage variation due to the increase in electrical charge at the surface of the HfO 2 /silicon structure. machines 2022, 13, x FOR PEER REVIEW 8 of for TNF-α (green) is 0.0005 (ng/mL) −1 , and for NT-proBNP (Red) the sensitivity is 0.002 (ng/mL) −1 . The immunosensor is 200 times more sensitive to cortisol than TNF-α, and 1 times more sensitive to cortisol than NT-proBNP.

Biosensor Calibration by Capacitive Measurements
Additionally, capacitive measurements were performed to characterize the semi-con ducting behavior of HfO2/silicon structures for each cortisol concentration. The biosenso was maintained in the electrochemical cell and incubated in PBS containing cortisol different concentrations. The biosensor was then rinsed with PBS to remove any adsorbe proteins and analyzed afterwards by capacitance measurements using PBS as an electr lyte. This procedure of immunosensor incubation was carried out for all cortisol concen trations (2 ng/mL to 50 ng/mL). The detection of cortisol at various concentrations shown in Figure 9A. Here, the Cs/Cmax-voltage curves show a shift in the positive dire tion with increasing cortisol concentrations, which confirms a flat-band voltage variatio due to the increase in electrical charge at the surface of the HfO2/silicon structure.  Figure 9B shows the calibration curve of the HfO2-based capacitive immunosensor a linear range from 2 ng/mL to 50 ng/mL of cortisol. The calibration curve has been pr sented as the absolute value of potential shift |E-E0| as a function of cortisol concentratio where E is the potential for different cortisol concentrations and E0 is the potential wit no cortisol in the solution. The slope of the straight line is 0.0023 (ng/mL) −1 .
The specificity of the biosensor prepared was studied using other biomarkers of hea failure, such as TNF-α and NT-proBNP. Capacitance measurements were performed u ing the same experimental process ( Figure 10A,B). TNF-α and NT-proBNP were detecte in the same linear range from 2 to 50 ng/mL with sensitivities of 0.0001 (ng/mL) −1 an 0.000001 (ng/mL) −1 , respectively. In this configuration of measurements, the immunosen sor is 23 times more sensitive to cortisol than TNF-α, and 2300 times more sensitive cortisol than NT-proBNP.  Figure 9B shows the calibration curve of the HfO 2 -based capacitive immunosensor in a linear range from 2 ng/mL to 50 ng/mL of cortisol. The calibration curve has been presented as the absolute value of potential shift |E-E 0 | as a function of cortisol concentration, where E is the potential for different cortisol concentrations and E 0 is the potential with no cortisol in the solution. The slope of the straight line is 0.0023 (ng/mL) −1 .
The specificity of the biosensor prepared was studied using other biomarkers of heart failure, such as TNF-α and NT-proBNP. Capacitance measurements were performed using the same experimental process ( Figure 10A,B). TNF-α and NT-proBNP were detected in the same linear range from 2 to 50 ng/mL with sensitivities of 0.0001 (ng/mL) −1 and 0.000001 (ng/mL) −1 , respectively. In this configuration of measurements, the immunosensor is 23 times more sensitive to cortisol than TNF-α, and 2300 times more sensitive to cortisol than NT-proBNP. The analytical performance of the HfO2/silicon-based immunosensor was compared to those of the published electrochemical immunosensors for the detection of cortisol (Table 2). The detection limit of the HfO2/silicon-based immunosensor is in the lower range, and its detection range falls in the range of concentrations of cortisol for a healthy person and for a person with HF.

Cortisol Quantification in Saliva Samples by Standard Addition Method (SAM)
To simulate the saliva sample analysis with an unknown cortisol concentration, three aliquots (450 µL) of the pooled saliva sample (PSS) were spiked with different volumes of 100 ng mL − 1 cortisol standard solution, obtaining a final concentration of 2, 5, and 7 ng mL − 1 of cortisol ,and then named "unknown sample" A, B, and C. SAM was carried out by first preparing SAM samples, where a constant volume (50 µL) of the "unknown sample" was added to each of a quartet of 1.5 mL Eppendorf Lo-bind centrifuge tubes (Sigma-Aldrich, Saint-Quentin-Fallavier, France). A total of 950 µL of PBS was added to the first The analytical performance of the HfO 2 /silicon-based immunosensor was compared to those of the published electrochemical immunosensors for the detection of cortisol ( Table 2). The detection limit of the HfO 2 /silicon-based immunosensor is in the lower range, and its detection range falls in the range of concentrations of cortisol for a healthy person and for a person with HF.

Cortisol Quantification in Saliva Samples by Standard Addition Method (SAM)
To simulate the saliva sample analysis with an unknown cortisol concentration, three aliquots (450 µL) of the pooled saliva sample (PSS) were spiked with different volumes of 100 ng mL −1 cortisol standard solution, obtaining a final concentration of 2, 5, and 7 ng mL −1 of cortisol, and then named "unknown sample" A, B, and C. SAM was carried out by first preparing SAM samples, where a constant volume (50 µL) of the "unknown sample" was added to each of a quartet of 1.5 mL Eppendorf Lo-bind centrifuge tubes (Sigma-Aldrich, Saint-Quentin-Fallavier, France). A total of 950 µL of PBS was added to the first tube up to obtain sample C0. Then, an increasing volume of the 100 ng mL −1 cortisol standard solution was added to each subsequent tube, before rounding the solutions in each tube to 1 mL with PBS, thus obtaining three samples, C1, C2, and C3, with known SAM concentrations, where they were analyzed by EIS.
Data obtained from SAM analysis are summarized in Table 3. The precision of our method was calculated by estimating the difference between the cortisol concentrations determined using SAM and their expected values. Based on these results, our biosensor shows good precision to determine the unknown concentration of cortisol in real saliva. These results confirmed that our biosensor was highly sensitive to the slight variation of the unknown concentration of cortisol.

Conclusions
In this study, we present the development of a label-free, highly sensitive, accurate, fast biosensor using EIS/capacitive measurement for cortisol detection in PBS/saliva while also showing good selectivity toward cortisol in the presence of other HF biomarkers (TNF-α and NT-proBNP). Tests performed in PBS showed a linear relationship with the increase in cortisol concentration and the resistance/capacitance of our transistor (R 2 was always > 0.97), showcasing the capability of our biosensor to quantitatively detect cortisol with a detection limit of 0.66 ng.mL −1 and a dynamic range from 2 to 50 ng.mL −1 . In addition, our biosensor showed high precision for detecting cortisol in unknown samples using the standard addition method and high sensitivity to small variations of cortisol concentration. This biosensor thus represents a promising bioanalytical tool for accurate quantification of cortisol in saliva to monitor symptoms of inflammation in HF patients. Moreover, the use of materials from microelectronics will allow their integration into silicon lab-on-a-chip devices.