High stability resistive switching mechanism of a screen-printed electrode based on BOBZBT2 organic pentamer for creatinine detection

The resistive switching (RS) mechanism is resulted from the formation and dissolution of a conductive filament due to the electrochemical redox-reactions and can be identified with a pinched hysteresis loop on the I–V characteristic curve. In this work, the RS behaviour was demonstrated using a screen-printed electrode (SPE) and was utilized for creatinine sensing application. The working electrode (WE) of the SPE has been modified with a novel small organic molecule, 1,4-bis[2-(5-thiophene-2-yl)-1-benzothiopene]-2,5-dioctyloxybenzene (BOBzBT2). Its stability at room temperature and the presence of thiophene monomers were exploited to facilitate the cation transport and thus, affecting the high resistive state (HRS) and low resistive state (LRS) of the electrochemical cell. The sensor works based on the interference imposed by the interaction between the creatinine molecule and the radical cation of BOBzBT2 to the conductive filament during the Cyclic Voltammetry (CV) measurement. Different concentrations of BOBzBT2 dilution were evaluated using various concentrations of non-clinical creatinine samples to identify the optimised setup of the sensor. Enhanced sensitivity of the sensor was observed at a high concentration of BOBzBT2 over creatinine concentration between 0.4 and 1.6 mg dL−1—corresponding to the normal range of a healthy individual.

Resistive switching mechanism. The working principle of the sensor is by exploiting the possible polar interaction between creatinine molecule and BOBzBT 2 pentamer as shown in Fig. 2a. There are two partial negatives, δ − groups in BOBzBT 2 pentamer, namely at the electron-rich atom such as S at the thiophene unit and O at the alkyloxy side chain. Meanwhile, two partial positive polar groups, δ + in creatinine molecule, were provided by the NH and NH 2 groups. It is more likely that the polar interaction is formed between S atom at the thiophene or benzothiophene units in BOBzBT 2 with the NH group in creatinine molecule. This is supported by the fact that the larger red-shift was observed in the N-H···S hydrogen-bonded as compared to the red-shift obtained from the N-H···O hydrogen-bonded as demonstrated in 26 . Moreover, the binding energy of the N-H···S complex was also comparable to that of the indole·benzene complex.
Previously, various voltage ranges, ranging from ± 4 to ± 20 V, were utilized to induce the RS [27][28][29][30][31] . To identify the voltage range suitable for the setup, CV measurements on the sensor modified with the 3BOBzBT 2 :5CHCl 3 dilution were conducted using different voltage ranges. As shown in the inset of Fig. 2b, the response obtained is highly unstable at a 0-1.5 V voltage range. This is expected due to the simplified electrode preparation resulting in nonuniform BOBzBT 2 layer. By increasing the voltage ranges to ± 10 V and ± 15 V, the stability of the responses was improved significantly. However, there were no appearances of a pinched hysteresis loop, which is a fingerprint of the RS. Performing the CV measurement with the voltage range of ± 20 V, revealing the pinched hysteresis loop which indicates that the system exhibiting the RS.
The inferred switching mechanism of the electrochemical cell is the formation and dissolution of horizontal CF. During the application of the forward bias voltage, the oxidation at the Ag electrode produces the Ag + cations. The cations were then migrated towards the inert electrode during the state that is referred to as the HRS (as shown in II) in Fig. 2c. At the same time, due to the charge transfer between the BOBzBT 2 pentamer and the WE, the radical cations of the BOBzBT 2 were generated. The deposited Ag atoms on the inert electrode selfassemble into Ag CFs (as shown in III) in Fig. 2c resulting in the formation of a conductive route between CE and WE. As consequence, the cell experience an abrupt drop in resistance and the surge of the output current. Following that, the CF is then dissolved bringing the system back to the HRS as represented in (IV) of Fig. 2c. By introducing the creatinine molecules into the electrolyte, the formation and the dissolution of the CF were perturbed due to its interaction with the BOBzBT 2 radical cations. This is evident by a significant reduction in the LRS peak and the area of the hysteresis loop obtained from the cell with the creatinine solution as compared to the peak obtained from the blank solution (Fig. 2d).
The optimization of BOBzBT 2 :CHCl 3 dilutions. The stability of the sensor modified with different concentrations of BOBzBT 2 dilution was later evaluated to identify the most optimized setup for creatinine sensing www.nature.com/scientificreports/ application. By performing the room temperature CV measurements using a single concentration of creatinine (0.7 mg dL −1 ), 5 hysteresis loops were recorded successively after the electroforming process. At low concentrations of BOBzBT 2 namely 1BOBzBT 2 :4CHCl 3 and 1BOBzBT 2 :3CHCl 3 , the hysteresis loops noticeably fluctuated in-between the measurements as shown in Fig. 3a, b. The stability between the measurements was seen to have slightly improved at 1BOBzBT 2 :2CHCl 3 (see Fig. 3c) and with 3BOBzBT 2 :5CHCl 3 dilution, greater stability between the loops was observed (Fig. 3d). In addition, at all BOBzBT 2 concentrations, the SET voltage during the electroforming cycle is significantly higher than the SET voltage of the RS cycles. The electroforming process is a crucial step to activate the RS mechanism in the cell, namely by forming the initial path for CF 7 . On the contrary, despite having the highest concentration of BOBzBT 2 , 1BOBzBT 2 :1CHCl 3 -sensor showing moderate stability comparable to medium concentration 1BOBzBT 2 :2CHCl 3 dilution as shown in Fig. 3e. This may have been resulted from the possible interference with the solvent and surface fouling due to the excessive BOBzBT 2 pentamer. The precision in-between the measurements was improved significantly in the following order-1BOBzBT 2 :4CHCl 3 < 1BOBzBT 2 :3CHCl 3 < 1BOBzBT 2 :1CHCl 3 < 1BOBzBT 2 :2CHCl 3 < 3BOBzBT 2 :5CHCl 3 with the average RSD values of 20.6% < 17.7% < 15.1% < 10.0% < 5.2%, respectively. The overall stability responses obtained using high voltage cyclic voltammetry throughout the experiment show better film degradation resistance of BOBzBT 2 . It was shown that a large forming voltage is required when the distance between the electrodes is larger 32 . For the commercially available SPE as used in the present work, a relatively large voltage range of ± 20 V is required to induce the RS. However, as the most consequential changes in the structure occur during the forming process, applying such a high voltage can lead to either the soft breakdown or the permanent breakdown 33 . The latter occurs when multiple bridges are produced during the forming process  www.nature.com/scientificreports/   www.nature.com/scientificreports/ and can result in unstable RS systems. On the contrary, soft breakdown refers to the condition where only a single CF is formed. As shown in Fig. 4a, at a low concentration of BOBzBT 2 (1BOBzBT 2 :3CHCl 3 in this example), multiple bridges with thickness ranging from 30 to 50 µm can be observed connecting the active metal electrode and the carbon WE. This may explain the relative instability of the system at lower BOBzBT 2 concentrations as obtained in Fig. 3a-c. By referring to Fig. 4b, at a higher concentration of BOBzBT 2 (3BOBzBT 2 :3CHCl3 in this case), a single CF with an average thickness of 70 µm can be seen clearly. The observation implies that more BOBzBT 2 in the system, provides more control to the formation of the CF, hence contributing to the enhanced stability of the system as in Fig. 3d. AFM-3D images of the bare carbon WE (Fig. 4c), after drop casting BOBzBT 2 dilution onto the WE (Fig. 4d), 1BOBzBT 2 :3CHCl 3 -modified WE and 3BOBzBT 2 :3CHCl 3 -modified WE after performing the CV measurements ( Fig. 4e, f, respectively) show that morphological change at each process. The surface of the WE showed a significant increase in thickness after the addition of BOBzBT 2 dilution along with the reduction in the surface roughness parameter, Ra to 26 nm from the average of 44 nm on bare WE. Meanwhile, the value of Ra obtained from 1BOBzBT 2 :3CHCl 3 -modified WE after performing the measurement increased to the average of 120 nm compared to the value before the measurement. Conversely, the surface of 3BOBzBT 2 :3CHCl 3 -modified WE changed to a smooth roughness form with the presence of sparsely located spikes after the measurement. This may be due to coherent electrochemical polymerization onto the WE as a result of the presence of more BOBzBT 2 pentamers in the cell.
The most stable 3BOBzBT 2 :5CHCl 3 -sensor is then compared with the unmodified-sensor and the middle concentration, 1BOBzBT 2 :2CHCl 3 -sensor using 0.7 mg dL −1 creatinine solution. Hysteresis loops obtained from 1BOBzBT 2 :2CHCl 3 -sensor and 3BOBzBT 2 :5CHCl 3 -sensor were noticeably larger than the unmodified-sensor as shown in Fig. 5a, suggesting that the RS yield is improved by the presence of the BOBzBT 2 pentamer. The interaction between BOBzBT 2 radical cations and the creatinine molecules impedes the mobility of Ag + cations to form the CF. On the other hand, different concentrations of BOBzBT 2 , i.e. 1BOBzBT 2 :2CHCl 3 and 3BOBzBT 2 :5CHCl 3 , generate a dissimilar concentration of BOBzBT 2 radical cations; hence, producing different resistance states in the cell. This explained the larger loop obtained from 1BOBzBT 2 :2CHCl 3 -sensor compared to the loop obtained from 3BOBzBT 2 :5CHCl 3 -sensor. Despite that, the output currents obtained from 1BOBzBT 2 :2CHCl 3 reduced inconsistently as the concentration of creatinine increased (Fig. 5b). In comparison, the reduction of peak current in the 3BOBzBT 2 :5CHCl 3 -sensor varies consistently according to the change of creatinine concentration (Fig. 5c). The occurrence indicates that the stability in the 3BOBzBT 2 :5CHCl 3 -sensor is better compared to the 1BOBzBT 2 :2CHCl 3 -sensor.
Taking the absolute values of output current and observing the logarithmic dependence in respect to the applied voltage, reveals a voltage gap between current minima obtained from forward and reverse loops. By referring to Fig. 5d, as the concentration of BOBzBT 2 increased from 1BOBzBT 2 :2CHCl3 to 3BOBzBT 2 :5CHCl3, the gap was reduced. This is expected since more radical cations of BOBzBT 2 are available for the interaction with creatinine molecules hence, restricting the activities of residual charges. Besides that, a larger variation of current minima, I a was observed from the 3BOBzBT 2 :5CHCl 3 -sensor compared to the variation, I b obtained from 1BOBzBT 2 :2CHCl 3 -sensor. The variation may indicate an increase in analyte uptake as the consequence www.nature.com/scientificreports/ of increasing the BOBzBT 2 concentration in the system 34 . However, one should be aware that this may also be contributed by parasitic resistance present in the system.
Response time and the dynamics of the system. One of the characteristics of a sensor is the time it takes to reach the steady-state response. The response time curve can be used as a reference time frame at which the measurement of the analyte can be made. To observe the response time of 3BOBzBT 2 :5CHCl 3 -sensor, consecutive CV measurements were conducted with 5 s time intervals in-between the measurement. The current output values at 5 V, 10 V, and 15 V were then extracted and plotted with respect to time. As a result of electroforming process, the output current demonstrates a sharp increase and reached a peak at t = 10 s for creatinine concentration of 0.7 mg dL −1 (Fig. 6a). Meanwhile, by using a creatinine concentration of 1.2 mg dL −1 i.e. with the increase of creatinine molecules in the system, the steady-state response was achieved at a much later time i.e. at t = 20 s (Fig. 6b). This suggests the electroforming process was interrupted with more creatinine-BOBzBT 2 radical cations interactions. A noticeable spike of output was observed at approximately t = 60 s for 3BOBzBT 2 :5CHCl 3 -sensor in 0.7 mg dL −1 and 1.2 mg dL −1 of creatinine concentrations (dashed arrows in Fig. 6a, b). Interestingly, it was preserved throughout all tested voltages. To investigate this occurrence, the following Cottrell equations were employed as below; where C O and C R are the concentration of oxidised (O) and reduced (R) forms of electroactive species, respectively. Meanwhile, D + and D − are the diffusion coefficients during the forward bias, calculated using maximum current, i p,+ and the reverse bias, calculated using modulus of minimum current, i p,− , respectively. The process of estimating the value of D + and D − can be found in the Supplementary Information. By solving both equations for t → ∞ and plotting the results against the normalised output current for four different concentrations of www.nature.com/scientificreports/ creatinine, it was demonstrated that there is an intersection between C O and C R in the range of 58 s-62 s, indicating an equal presence of O and R in the system (Fig. 6c).
Sensor performance for quantifying creatinine concentration. To investigate the performance of 3BOBzBT 2 :5CHCl 3 -sensor, a calibration curve was established by taking the output current at voltage value of 15 V measured at each creatinine concentration. This value is then subtracted by the output measured from performing the measurement using a blank solution (0.0 mg dL −1 ) to obtained, I . By referring to Fig. 7a, the working range of the 3BOBzBT 2 :5CHCl 3 -sensor was 0.4-1.6 mg dL −1 of creatinine. Despite the subtracted output current, I rises steadily with a concentration in this range, two linear regressions can be drawn from the curve namely, one for the low concentration range of 0.4 mg dL −1 ≤ x ≤ 0.9 mg dL −1 (region x in Fig. 7a) and one for  www.nature.com/scientificreports/ the high concentration range of 0.9 mg dL −1 ≤ y ≤ 1.6 mg dL −1 (region y in Fig. 7a). These regions x and y were fitted to the data with correlation coefficients R 2 of 0.977 and 0.992, respectively. Despite larger R 2 , the sensitivity, taken from the slope of the linear regression line; for the region y shows a 70% reduction at 0.665 A dL g −1 as compared to 2.204 A dL g −1 obtained from the region x . The finding is similar to the observation reported in 35 which suggested that two linear ranges with lower sensitivity at higher concentration regions may have been due to the accumulation of the analyte on the modified electrode. The value of the limit of detection (LOD) region x and y were determined at 0.20 mg dL −1 and 0.18 mg dL −1 , respectively while the limit of quantification (LOQ), was calculated at 0.63 mg dL −1 for region x and 0.55 mg dL −1 region y . Both the LOD and LOQ were calculated based on the ICH Topic Q2 (R1) Validation of Analytical Procedures: Text and Methodology, namely clauses 6 and 7, respectively. Meanwhile, by transforming both axes of the calibration curve (Fig. 7b) to the logarithmic scale, regions x and y can be effectively combined into a single pseudo-linear regression line with a relatively high value of R 2 = 0.978.
To further demonstrate the improvement made by the BOBzBT 2 on the sensor performance, the calibration curves obtained from the unmodified-sensor, 1BOBzBT 2 :2CHCl 3 -sensor, and 3BOBzBT 2 :5CHCl 3 -sensor were compared (Fig. 8a). No linear operating range can be established from the unmodified-sensor due to the fluctuation of the output currents for the concentrations. This further suggests that, in the absence of BOBzBT 2 in the cell, the CF was not modulated, resulting in the incoherence formation and dissolution process of the CF. Moreover, as previously demonstrated in Fig. 5a, the narrower loop obtained from the unmodified sensor indicates that the cell response approaches that of the ohmic-like behaviour. The addition of BOBzBT 2 enhanced the retention of the CF similar to the observation reported in 33 , where the addition of defective graphene into Ag/SiO 2 /Pt cell led to the highly stable and robust CF during the HRS phase. On the other hand, it was observed that the sensor with 1BOBzBT 2 :2CHCl3 dilution exhibits a limited detection range of 0.7 mg dL −1 and 1.1 mg dL −1 compared to the 3BOBzBT 2 :5CHCl 3 -sensor. This may have been caused by branching and the formation of multiple CFs as discussed in the previous section, predominantly at the lower (1.1 mg dL −1 ) concentrations of creatinine. Modest performance for the sensor with 1BOBzBT 2 :2CHCl 3 dilution was also obtained with larger LOD and LOQ of 0.36 mg dL −1 and 1.08 mg dL −1 , respectively.
The analysis of variance (ANOVA) was later conducted to analyze the statistical significance of the variance within the group as represented by the error bar and between the groups, i.e. between the BOBzBT 2 -and unmodified-sensors. The null hypothesis is set to no difference between all three sensors. The Sum of Squares Within Groups (SSW), Sum of Squares between Group (SSG), and Total Sum of Squares (SST) for the sensors were determined to obtain the value of F-ratio at each creatinine concentration. Using the confidence level of 95% and the numerator and the denominator degrees of freedom are 2 and 9, respectively, the critical value of F (2,9) is given at 4.256. It was calculated that the value of F is larger than the critical value at all tested creatinine concentrations as listed in Table S.1 in the Supplementary Information. A larger calculated F value as compared to the critical value indicates that the variation among group means is significant. This is thus concluded that the null hypothesis can be rejected hence the difference between 3BOBzBT 2 :5CHCl 3 -sensor, 1BOBzBT 2 :2CHCl 3 -sensor and unmodified-sensor are statistically significant.
The selectivity of the modified-SPE was evaluated by conducting CV measurements using different concentrations of human serum albumin (HSA) solution. The HSA was chosen since it has the net negative charge 36 i.e., a direct opposition from the inferred interaction between the partial negative group in the BOBzBT 2 and the partial positive group in the creatinine molecule. As demonstrated in Fig. S1 in the Supplementary Information, only Ohmic-like responses were recorded at different HSA concentrations, except for 0.8 mg dL −1 of HSA. Despite the visible pinched hysteresis loop at this concentration, the area of the loop is extremely narrow compared to the loop obtained from using the creatinine solutions. This implies that the interaction between HSA molecules and the BOBzBT 2 radical cations is inadequate to modulate the resistance state of the cell; thus, the efficiency www.nature.com/scientificreports/ of the cell to induce the resistive switching is low. Moreover, no working range could be established (inset of Fig. S1 in the Supplementary Information) suggesting the selectivity of modified-SPE towards the creatinine molecules over the HSA molecules. The utilization of a large voltage range to induce the RS mechanism comes with the expense of deteriorating the electrodes. As a result, the sensor strip is a single use which can be demonstrated by performing repeatability testing. The test was conducted by rinsing the strip with DI water after each measurement. The rinsed strip is then left dried before being drop-casted with 3BOBzBT 2 :5CHCl 3 dilution. As shown in Fig. S2 in the Supplementary Information, after the measurement using the fresh strip, there were no usable signals that can be recorded from the subsequent measurements. The reproducibility of the sensor was later investigated by conducting CV measurement for a blank solution followed by a measurement using 1.2 mg dL −1 creatinine concentration. The step was repeated five times and the results have demonstrated that the sensor response is consistent throughout the cycle, indicating its high reproducibility (Fig. 8b). On the other hand, the comparison between the LOD and the detection range of 1BOBzBT 2 :2CHCl 3 -and 3BOBzBT 2 :5CHCl 3 -sensors with other methods proposed in the literature are summarized in Table S

Conclusions
We demonstrated a low-cost, RS-based creatinine sensor using disposable screen-printed electrode (SPE). The working electrode (WE) of the SPE was modified with a small organic molecule, 1,4-bis[2-(5-thiophene-2-yl)-1-benzothiopene]-2,5-dioctyloxybenzene (BOBzBT 2 ). The polar interaction occurred in the electrolyte, namely, between the radical cations of BOBzBT 2 pentamer and creatinine molecules influencing the resistance state of the cell. As a result, the RS responses i.e. the pinched hysteresis loops varied with respect to the creatinine concentrations. Accordingly, the quantification of creatinine concentration was carried out based on the reduction of output current as creatinine concentration increased. BOBzBT 2 dilution with the ratio of 3BOBzBT 2 :5CHCl 3 has improved the overall performance of the sensor with the sensitivity of 2.204 A dL g −1 and 0.665 A dL g −1 for region x and region y , respectively. The sensor responded linearly over the concentration range 0.4 mg dL −1 to 1.6 mg dL −1 , corresponding to the creatinine level in serum for healthy individuals. Meanwhile, the limit of detection (LOD) of the sensor is as low as 0.18 mg dL −1 . Minimal fabrication procedures offered by the setup along with a good sensing performance may pave a way for the development of a RS-based SPE sensor for quantifying creatinine concentration. Moreover, by functionalizing BOBzBT 2 with a particular receptor, the selectivity and the specificity of the sensor can be further improved hence enabling the real sample measurement.

Preparation of creatinine stock solution.
Creatinine stock solution (5 mg dL −1 ) was prepared by dissolving 12.5 mg of creatinine with deionised water in a 250 ml (2.5 dL) volumetric flask. A series of creatinine dilutions was then prepared from a stock solution ranging from 0.4 to 1.6 mg dL −1 and used as a standard working solution. The range of concentration was chosen based on the standard concentration level of creatinine in serum of healthy individuals 13 .

UV-VIS spectrophotometry.
A Hitachi U-3900H Spectrophotometer was used to measure the absorption spectra for each ratio of BOBzBT 2 :CHCl 3 dilution and 3BOBzBT 2 :5CHCl 3 dilution with the addition of different concentrations of creatinine. The sample was prepared by dropping 2.5µL of BOBzBT 2 :CHCl 3 dilution onto the glass slide and was left for 5 min drying at room temperature. Meanwhile, a single concentration of dilution, namely 3BOBzBT 2 :5CHCl 3 was randomly selected to observed the UV-VIS absorbance spectra in the www.nature.com/scientificreports/ presence of creatinine solutions. A droplet of creatinine solution at ten different concentrations was added on top of 10 glass slides with each containing 2.5µL of 3BOBzBT 2 :5CHCl 3 that was left dried for 5 min. The spectrophotometry scanning was performed at a 300nm/min scan rate with a 300-700nm wavelength range.
Electrochemical measurement. The sensor used a three-electrode screen-printed strip (Zimmer Peacok; "SPE") with a carbon working electrode (WE) (0.038cm −2 of area) and Ag/AgCl electrode for both reference (RE) and counter electrodes (CE) (Fig. 1). Amperometric measurements were performed using a Keithley 2400 series 200 V, 1A, 20 W SMU. The potential sweeping is applied by the voltage source (V S ) between the WE and the CE. The potential across the RE and WE are measured with the voltmeter, V m and to maintain the intended potential at WE with respect to RE, V s is adjusted accordingly. Subsequently, the resulting current flowing to or from WE is measured with the ammeter (A M ).
Electrode preparation. Before the measurement, 2.5 µL of BOBzBT 2 dilution was drop-casted on WE (Area 1 in Fig. 1) and left for several minutes to dry at room temperature. Next, a droplet of creatinine solution was added in the vicinity of Area 1, which covered all three electrodes and left for another few minutes, allowing the chemical relaxation in the system to occur. The cyclic voltammetry measurement was performed, and the obtained output current was recorded. The procedure is repeated for all five ratios of BOBzBT 2 : CHCl 3 dilution, at which ten different concentrations of creatinine solutions were used to identify the most optimised dilution for the sensor.