Effect of Fluoride on the Morphology and Electrochemical Property of Co3O4 Nanostructures for Hydrazine Detection

In this paper, we systematically investigated the influence of fluoride on the morphology and electrochemical property of Co3O4 nanostructures for hydrazine detection. The results showed that with the introduction of NH4F during the synthesis process of Co3O4, both Co(CO3)0.5(OH)·0.11H2O and Co(OH)F precursors would be generated. To understand the influence of F on the morphology and electrochemical property of Co3O4, three Co3O4 nanostructures that were respectively obtained from bare Co(CO3)0.5(OH)·0.11H2O, Co(OH)F and Co(CO3)0.5(OH)·0.11H2O mixtures and bare Co(OH)F were successfully synthesized. The electrochemical tests revealed the sensing performance of prepared Co3O4 nanostructures decreased with the increase in the fluoride contents of precursors. The more that dosages of NH4F were used, the higher crystallinity and smaller specific surface area of Co3O4 was gained. Among these three Co3O4 nanostructures, the Co3O4 that was obtained from bare Co(CO3)0.5(OH)·0.11H2O-based hydrazine sensor displayed the best performances, which exhibited a great sensitivity (32.42 μA·mM−1), a low detection limit (9.7 μΜ), and a wide linear range (0.010–2.380 mM), together with good selectivity, great reproducibility and longtime stability. To the best of our knowledge, it was revealed for the first time that the sensing performance of prepared Co3O4 nanostructures decreased with the increase in fluoride contents of precursors.


Introduction
Hydrazine and based chemicals are water soluble volatile colorless liquids, and the simplest unique diamine in its class has aroused wide concern for its large number of applications in many spheres, for instance, corrosive inhibitors, fuel cells and so on [1][2][3][4]. The laboratory research and commercial application of hydrazine as a reducing agent and catalyst are commonly implemented. However, hydrazine and its derivatives do great harm to the body through the digestive system along with skin permeation [5]. Consequently, it is highly imperative to propose a sensitive, original and analytically credible tool for the effective detection of hydrazine. Recently, electroanalytical techniques have developed as a desirable method for detection of many chemicals, such as hydroquinone [6], acetone [7], herbicides [8], etc. due to their great sensitivity, better efficiency and low cost. It is promising to develop electrochemical methods for hydrazine detection. Nowadays, lots of semiconductor metal oxides have been used for hydrazine electrochemical sensing. Ahmad et al. have fabricated a ZnO nanorods-based hydrazine sensor and showed a low detection limit [9]. Wu et al.

Structural and Morphological Studies
The composition of the prepared precursors was first examined by XRD characterization. As shown in Figure 1a,c, all of the diffraction peaks could be well indexed to Co(CO 3   , which indicates that the precursors had been thoroughly converted to Co 3 O 4 phase and there were no other impurities that could be detected. The crystallinity of the prepared products was evaluated via the Formula (1).
In Formula (1), Xc represents the crystallinity measured by X-ray diffractometry; Ic is the integral intensity of the crystal diffraction peak; and Ia is the integral intensity of the amorphous diffuse peak; K is the elative scattering factor.
The crystallinity of prepared Co 3 O 4 samples is calculated and shown in Table 1. It is obvious that Co 3 O 4 -1 possesses the lowest crystallinity (73.87%) and Co 3 O 4 -3 has the highest crystallinity 91.62%). To explore the influence of F on the crystallinity of Co 3 O 4 , different dosages of NH 4 F (from 0 to 20 mmol) had been added during the synthesis process of Co 3 O 4 . As shown in Table 1, the crystallinity of Co 3 O 4 presents an increase trend with increasing the amount of NH 4 F during the synthesis process of precursors, which indicates that the strong interaction between Co 2+ and F − can make the particles grow regularly.
The formation of the precursors and annealed products can be illustrated with the following steps. When CO(NH 2 ) 2 is used as a hydrolysis reagent, the formation process of Co(CO 3 ) 0.5 (OH)·0.11H 2 O can be depicted as Equations (2)- (4). While when HMT is used as a hydrolysis reagent, with the existence of F -, the formation of Co(OH)F can be possibly expressed by Equations (5)- (8). After calcination, both of the precursors can be converted into cobaltosic oxide by Equations (9) and (10).
Materials 2018, 11, x FOR PEER REVIEW 3 of 16 of precursors, which indicates that the strong interaction between Co 2+ and F − can make the particles grow regularly. The formation of the precursors and annealed products can be illustrated with the following steps. When CO(NH2)2 is used as a hydrolysis reagent, the formation process of Co(CO3)0.5(OH)·0.11H2O can be depicted as Equations (2)- (4). While when HMT is used as a hydrolysis reagent, with the existence of F -, the formation of Co(OH)F can be possibly expressed by Equations (5)- (8). After calcination, both of the precursors can be converted into cobaltosic oxide by Equations (9) and (10).
H2O + NH3 → OH − + NH4 + (7)    [28][29][30][31]. The peaks that appeared at 963 and 511 cm −1 are ascribed to δ(Co-OH) and ρ w (Co-OH) bending modes, respectively [28][29][30][31]. For the precursor of Co 3 O 4 -3, whose chemical composition is Co(OH)F, the peak at 3575 cm −1 is assigned to the O-H stretching mode too. The shoulder vibration at 3417 cm −1 corresponds to the O-H groups interacting with fluoride anions. In addition, no any other band related to carbonate anions can be found (Figure 2a). After annealing treatment, the bands of precursors disappear and two very strong peaks are centered at 665 and 543 cm −1 characteristic of Co 3 O 4 are noticed ( Figure 2b). The former peak at 665 cm −1 corresponds to Co 2+ -O bond [30]. The other peak at 543 cm −1 can be ascribed to Co 3+ -O bond [31]. The results verified the formation of Co 3 O 4 under thermal degradation, which coincides with the XRD patterns.
To get more information about the chemical states of elements in prepared Co 3 O 4 samples, XPS analysis was then performed. Figure 3a shows the Co 2p peaks of Co 3 O 4 -1, Co 3 O 4 -2, and Co 3 O 4 -3, respectively. The curve of Co 2p shows two spin-orbit doublets of Co 2p1/2 at 779.96 and 794.94 eV attributed to Co 3+ , and two spin-orbit doublets of Co 2p3/2 at 781.46 and 796.46 eV belonging to Co 2+ . The intensity of the peaks shows downward course from Co 3 O 4 -1 to Co 3 O 4 -3. Figure 3b depicts the O1s spectra of the prepared Co 3 O 4 samples, the large O 1s peak at 530.12 eV is attributed to the lattice oxygen (LO) in Co 3 O 4 crystals and the small O 1s peak at 531.47 eV represents the oxygen vacancies (OV) on the surface of Co 3 O 4 [32][33][34]. The intensity of the oxygen vacancy also shows a decreased trend from Co 3 O 4 -1 to Co 3 O 4 -3 ( Table 2). The more oxygen vacancies indicate a higher possibility for the exposure of active sites [35]. The general morphology of the precursors and the obtained Co 3 O 4 products was further explored using FE-SEM, as shown in Figure 4. The Co(CO 3 ) 0.5 (OH)·0.11H 2 O precursor exhibited a nanorod structure, with a smooth surface (Figure 4a). The average length of the nanorods is in the range of 3-5 µm. The Co(OH)F precursor showed gear-like nanosheets, with an uneven surface ( Figure 4e). The average thickness of Co(OH)F nanosheets is about 0.025 µm and the typical diameter of the nanosheets is in the range of 8-12 µm. Figure 4b depicts the SEM image of the mixture of Co(CO 3 ) 0.5 (OH)·0.11H 2 O nanorods and Co(OH)F nanosheets. When CO(NH 2 ) 2 was used as the hydrolysis reagent, both Co(CO 3 ) 0.5 (OH)·0.11H 2 O and Co(OH)F precursors were generated in the same time with the existence of F. However, from Figure 4c, it is noteworthy that the quantity of nanorods is far more than nanosheets, which indicates that the growth rate of Co(CO 3 ) 0.5 (OH)·0.11H 2 O is faster than the growth rate of Co(OH)F. The elemental mapping analysis in the selected area of Figure 4c further confirmed the generation of Co(OH)F. Figure 5 suggests that the F element was evenly dispersed within the Co(OH)F sample, with a calculated atomic amount of ca. 5.8%. After annealing the precursors, the morphology of obtained Co 3 O 4 -1, Co 3 O 4 -2, and Co 3 O 4 -3 samples were also characterized using FE-SEM, as shown in Figure 4b,d,f. It was worthwhile mentioning that the products still maintained the similar morphology with their precursors after the annealing treatment. Contrary to the precursors, the Co 3 O 4 displayed rough and porous surfaces, which might be due to the abscission of attached OH ions during the calcination process. The difference of morphology may lead to the difference of specific surface area of Co 3 O 4 . Based on the BET data, it could be found that the rod-like Co 3 O 4 possessed larger specific surface area than sheet-like Co 3 O 4 . The relatively large specific surface area of rod-like structure may be beneficial to the exposure of active sites and in the meantime facilitate the contact of Co 3 O 4 to the targeted chemicals. of the nanosheets is in the range of 8-12 μm. Figure 4b depicts the SEM image of the mixture of Co(CO3)0.5(OH)·0.11H2O nanorods and Co(OH)F nanosheets. When CO(NH2)2 was used as the hydrolysis reagent, both Co(CO3)0.5(OH)·0.11H2O and Co(OH)F precursors were generated in the same time with the existence of F. However, from Figure 4c, it is noteworthy that the quantity of nanorods is far more than nanosheets, which indicates that the growth rate of Co(CO3)0.5(OH)·0.11H2O is faster than the growth rate of Co(OH)F. The elemental mapping analysis in the selected area of Figure 4c further confirmed the generation of Co(OH)F. Figure 5 suggests that the F element was evenly dispersed within the Co(OH)F sample, with a calculated atomic amount of ca. 5.8%. After annealing the precursors, the morphology of obtained Co3O4-1, Co3O4-2, and Co3O4-3 samples were also characterized using FE-SEM, as shown in Figure 4b,d,f. It was worthwhile mentioning that the products still maintained the similar morphology with their precursors after the annealing treatment. Contrary to the precursors, the Co3O4 displayed rough and porous surfaces, which might be due to the abscission of attached OH ions during the calcination process. The difference of morphology may lead to the difference of specific surface area of Co3O4. Based on the BET data, it could be found that the rod-like Co3O4 possessed larger specific surface area than sheetlike Co3O4. The relatively large specific surface area of rod-like structure may be beneficial to the exposure of active sites and in the meantime facilitate the contact of Co3O4 to the targeted chemicals.

Hydrazine Chemical Sensor Studies of Co3O4 Modified Electrodes
To prepare the hydrazine sensor, Co3O4 nanomaterials were coated on the surface of GCE. The electrocatalytic activity of Co3O4 nanomaterials towards hydrazine was firstly investigated by cyclic voltammetry technique. Figure 6a shows the cyclic voltammogram (CV) of bare GCE and Co3O4/GCE in the presence of 1 mM hydrazine in 0.1 M NaOH electrolyte at a scan rate of 0.02 V·s −1 . It is apparent that no matter whether hydrazine exists, the bare GCE does not exhibit any redox peak in 0 to 0.6 V, just the current elevated when hydrazine added. This result indicates the low catalytic activity of bare GCE. However, for these three kinds of Co3O4 modified electrodes, there are significant differences among them. For Co3O4-3 (obtained from Co(OH)F) modified electrode, no peak could be observed when it was tested. However, for Co3O4-2 (obtained from the mixture) and Co3O4-1 (obtained from Co(CO3)0.5(OH)·0.11H2O) modified electrodes, an oxidation peak (I) apparently emerged at around 0.50 V. The observed CV response also exhibited reversible nature as it showed a reduction peak (II) at 0.46 V during the reverse sweep. Among them, Co3O4-1 possesses the best electrochemical activity towards hydrazine oxidation while Co3O4-3 has the poorest electrochemical activity towards hydrazine. This result suggests that F has a negative effect on the

Hydrazine Chemical Sensor Studies of Co3O4 Modified Electrodes
To prepare the hydrazine sensor, Co3O4 nanomaterials were coated on the surface of GCE. The electrocatalytic activity of Co3O4 nanomaterials towards hydrazine was firstly investigated by cyclic voltammetry technique. Figure 6a shows the cyclic voltammogram (CV) of bare GCE and Co3O4/GCE in the presence of 1 mM hydrazine in 0.1 M NaOH electrolyte at a scan rate of 0.02 V·s −1 . It is apparent that no matter whether hydrazine exists, the bare GCE does not exhibit any redox peak in 0 to 0.6 V, just the current elevated when hydrazine added. This result indicates the low catalytic activity of bare GCE. However, for these three kinds of Co3O4 modified electrodes, there are significant differences among them. For Co3O4-3 (obtained from Co(OH)F) modified electrode, no peak could be observed when it was tested. However, for Co3O4-2 (obtained from the mixture) and Co3O4-1 (obtained from Co(CO3)0.5(OH)·0.11H2O) modified electrodes, an oxidation peak (I) apparently emerged at around 0.50 V. The observed CV response also exhibited reversible nature as it showed a reduction peak (II) at 0.46 V during the reverse sweep. Among them, Co3O4-1 possesses the best electrochemical activity towards hydrazine oxidation while Co3O4-3 has the poorest electrochemical activity towards hydrazine. This result suggests that F has a negative effect on the

Hydrazine Chemical Sensor Studies of Co 3 O 4 Modified Electrodes
To prepare the hydrazine sensor, Co 3 O 4 nanomaterials were coated on the surface of GCE. The electrocatalytic activity of Co 3 O 4 nanomaterials towards hydrazine was firstly investigated by cyclic voltammetry technique. Figure 6a shows the cyclic voltammogram (CV) of bare GCE and Co 3 O 4 /GCE in the presence of 1 mM hydrazine in 0.1 M NaOH electrolyte at a scan rate of 0.02 V·s −1 . It is apparent that no matter whether hydrazine exists, the bare GCE does not exhibit any redox peak in 0 to 0.6 V, just the current elevated when hydrazine added. This result indicates the low catalytic activity of bare GCE. However, for these three kinds of Co 3 O 4 modified electrodes, there are significant differences among them. For Co 3 O 4 -3 (obtained from Co(OH)F) modified electrode, no peak could be observed when it was tested. However, for Co 3 O 4 -2 (obtained from the mixture) and Co 3 O 4 -1 (obtained from Co(CO 3 ) 0.5 (OH)·0.11H 2 O) modified electrodes, an oxidation peak (I) apparently emerged at around 0.50 V. The observed CV response also exhibited reversible nature as it showed a reduction peak (II) at 0.46 V during the reverse sweep. Among them, Co 3 O 4 -1 possesses the best electrochemical activity towards hydrazine oxidation while Co 3 O 4 -3 has the poorest electrochemical activity towards hydrazine. This result suggests that F has a negative effect on the performance of Co 3 O 4 . We believe that the performance of Co 3 O 4 should be related to their crystallinity and specific surface area. According to the XPS analysis, with increasing the amount of F in the synthesis process, the oxygen vacancy of their final product (Co 3 O 4 ) shows a decreased trend. The addition of F allows the precursor to grow more regularly but cause the decrease of the specific surface area, resulting in relatively less active sites and worse performance of Co 3 O 4 [36]. The possible reactions on Co 3 O 4 electrode can be expressed as the following Equations (11)- (13). performance of Co3O4. We believe that the performance of Co3O4 should be related to their crystallinity and specific surface area. According to the XPS analysis, with increasing the amount of F in the synthesis process, the oxygen vacancy of their final product (Co3O4) shows a decreased trend. The addition of F allows the precursor to grow more regularly but cause the decrease of the specific surface area, resulting in relatively less active sites and worse performance of Co3O4 [36]. The possible reactions on Co3O4 electrode can be expressed as the following Equations (11)- (13).
4CoO2 + N2H4 → 4CoOOH + N2 (13) The influence of hydrazine concentration and the scan rates on the performance of modified electrode was then investigated using the Co3O4-1 sample. Figure 6b exhibits the cyclic voltammograms of Co3O4-1 modified GCE with different hydrazine concentrations at a scan rate of 0.02 V·s −1 . With increasing the hydrazine concentration from 0.5 to 5 mM, the current displays a growth trend. The simultaneous response reveals that the fabricated Co3O4-1-based sensor can be used for the effective determination of hydrazine. Figure 6c depicts the cyclic voltammograms of Co3O4-1 modified GCE with 1 mM hydrazine at different scan rates ranging from 0.01 to 0.08 V·s −1 .
The inset of Figure 6c shows that the peak current (Ip) also increases synchronously with the scan rate. The relationship between Ip and the scan rate was further calculated based on the Randles-Sevcik equation [37]. The equation can be expressed as Ip = 261.3 ν 1/2 − 14.4 (R 2 = 0.997). The negative intercept may be due to the adsorption of the N2H4 occurred on the electrode surface, which indicates that the electrode reaction is not a single diffusion-controlled process [38]. Moreover, it is notable that The influence of hydrazine concentration and the scan rates on the performance of modified electrode was then investigated using the Co 3 O 4 -1 sample. Figure 6b exhibits the cyclic voltammograms of Co 3 O 4 -1 modified GCE with different hydrazine concentrations at a scan rate of 0.02 V·s −1 . With increasing the hydrazine concentration from 0.5 to 5 mM, the current displays a growth trend. The simultaneous response reveals that the fabricated Co 3 O 4 -1-based sensor can be used for the effective determination of hydrazine. Figure 6c depicts the cyclic voltammograms of Co 3 O 4 -1 modified GCE with 1 mM hydrazine at different scan rates ranging from 0.01 to 0.08 V·s −1 .
The inset of Figure 6c shows that the peak current (Ip) also increases synchronously with the scan rate. The relationship between Ip and the scan rate was further calculated based on the Randles-Sevcik equation [37]. The equation can be expressed as Ip = 261.3 ν 1/2 − 14.4 (R 2 = 0.997). The negative intercept may be due to the adsorption of the N 2 H 4 occurred on the electrode surface, which indicates that the electrode reaction is not a single diffusion-controlled process [38]. Moreover, it is notable that Materials 2018, 11, 207 9 of 16 the peak potential shift towards positive potential with increasing the scan rate. Figure 6d exhibits the linear relation between the peak potential (Ep) and log (ν), implying the irreversible oxidation of hydrazine at the surface of Co 3 O 4 -1/GCE. Figure 7a displays the chronoamperometric response of Co 3 O 4 -1/GCE with different concentration of hydrazine. The transient currents decayed with prolonging the time, also revealing the diffusion-controlled process of hydrazine electrooxidation. The peak current exhibited linear relationship with t −1/2 (Figure 7b). In addition, the slope of the line increased with increasing the hydrazine concentration (Figure 7c). Thus, the diffusion coefficient of hydrazine (D) could be calculated via Cottrell's equation: In Equation (14), n, F (C·mol −1 ), A (cm 2 ), C (mol·cm −3 ) respectively represents the number of involved electron transfer, the Faraday constant (96,485), the surface area of GCE, and the dosage of hydrazine. The slopes of the obtained linear lines were plotted against the hydrazine concentrations (Figure 7c). Based on this plot, D was determined to be 1.66 × 10 −5 cm 2 ·s −1 , which is consistent with the previous report [39].
Materials 2018, 11, x FOR PEER REVIEW 9 of 16 the peak potential shift towards positive potential with increasing the scan rate. Figure 6d exhibits the linear relation between the peak potential (Ep) and log (ν), implying the irreversible oxidation of hydrazine at the surface of Co3O4-1/GCE. Figure 7a displays the chronoamperometric response of Co3O4-1/GCE with different concentration of hydrazine. The transient currents decayed with prolonging the time, also revealing the diffusion-controlled process of hydrazine electrooxidation. The peak current exhibited linear relationship with t −1/2 (Figure 7b). In addition, the slope of the line increased with increasing the hydrazine concentration (Figure 7c). Thus, the diffusion coefficient of hydrazine (D) could be calculated via Cottrell's equation: In Equation (14), n, F (C·mol −1 ), A (cm 2 ), C (mol·cm −3 ) respectively represents the number of involved electron transfer, the Faraday constant (96,485), the surface area of GCE, and the dosage of hydrazine. The slopes of the obtained linear lines were plotted against the hydrazine concentrations (Figure 7c). Based on this plot, D was determined to be 1.66 × 10 −5 cm 2 ·s −1 , which is consistent with the previous report [39].

Amperometric Detection of Hydrazine Using Co 3 O 4 Modified Electrodes
The Co 3 O 4 -1 and Co 3 O 4 -2 modified electrodes were then used as a sensor for detection of hydrazine. The work potential was set at 0.50 V. For comparison, the amperometric responses of Co 3 O 4 -1 and Co 3 O 4 -2 electrodes are displayed. From Figure 8a,c, it is apparent that with the successive addition of hydrazine to a stirred solution, the anodic current increases gradually. When an aliquot of hydrazine was dropped into the stirred NaOH solution, the amperometric responses of the Co 3 O 4 -1 modified electrode achieved a steady state within 2 s, which is faster than the Co 3 O 4 -2 modified electrode. On the other hand, the magnitudes of the response current of the Co 3 O 4 -1 modified electrode is also larger than the Co 3 O 4 -2 modified electrode at the same condition. These results suggest that the Co 3 O 4 -1 modified electrode has better electrochemical performance than the Co 3 O 4 -2 modified electrode. When the hydrazine concentration exceeds a certain range, the response currents will no longer increase, but turn to be a declining trend. This phenomenon indicates that the hydrazine concentration exceeds the critical value of linear range. In order to further distinguish the difference between the two electrodes, mathematic fitting was utilized to calculate the sensitivity, linear response range and the detection limit. Figure 8b

Amperometric Detection of Hydrazine Using Co3O4 Modified Electrodes
The Co3O4-1 and Co3O4-2 modified electrodes were then used as a sensor for detection of hydrazine. The work potential was set at 0.50 V. For comparison, the amperometric responses of Co3O4-1 and Co3O4-2 electrodes are displayed. From Figure 8a,c, it is apparent that with the successive addition of hydrazine to a stirred solution, the anodic current increases gradually. When an aliquot of hydrazine was dropped into the stirred NaOH solution, the amperometric responses of the Co3O4-1 modified electrode achieved a steady state within 2 s, which is faster than the Co3O4-2 modified electrode. On the other hand, the magnitudes of the response current of the Co3O4-1 modified electrode is also larger than the Co3O4-2 modified electrode at the same condition. These results suggest that the Co3O4-1 modified electrode has better electrochemical performance than the Co3O4-2 modified electrode. When the hydrazine concentration exceeds a certain range, the response currents will no longer increase, but turn to be a declining trend. This phenomenon indicates that the hydrazine concentration exceeds the critical value of linear range. In order to further distinguish the difference between the two electrodes, mathematic fitting was utilized to calculate the sensitivity, linear response range and the detection limit. Figure 8b Table 3 summarizes the electrochemical parameters of some reported N2H4 sensors. Compared with them, the Co3O4-1/GCE and Co3O4-2/GCE exhibit rather high sensitivity and wide linear range. The performance of Co3O4-1/GCE is better than Co3O4-2/GCE. These results reveal that the existence   Table 4. The Co 3 O 4 -1/GCE and Co 3 O 4 -2/GCE exhibit no significant difference in surface area normalized current and surface area normalized sensitivity, which implies that the specific surface area is one of crucial factors for the electrochemical performances of Co 3 O 4 . As above-mentioned, the amount of NH 4 F is inversely related to the specific surface area. The dosage of NH 4 F during the hydrothermal process affects the specific surface area of the products directly and therefore causes the difference in their electrochemical performances. To obtain the highly active Co 3 O 4 -based modified electrodes for hydrazine detection, F should be avoided, although in many cases the F species cannot be detected by the XRD and FTIR analysis. However, the mechanism on how the remaining F affects the electrochemical activity of Co 3 O 4 is still under investigation.

Selectivity, Reproducibility and Stability Tests
Selectivity and stability are two of key parameters to evaluate performance of chemical sensors. Thus, the selectivity and stability of the Co 3 O 4 -1-based hydrazine sensor were also explored. Figure 9a exhibits the i-t curve response of hydrazine and interferent (Cl − , CO 3 2− , NO 3 − , NO 2 − , CH 3 COO − , K + , Na + , tap water, and humic acid). When 0.1 M N 2 H 4 (10 µL) was injected to the NaOH solution, a quick response can be detected. However, when the same dosage of interfering species is added to the electrolyte, no obvious current response could be observed, suggesting the good selectivity of Co 3 O 4 -1/GCE for N 2 H 4 detection. To evaluate the reproducibility, seven different glassy carbon electrodes were prepared via the same modification step. The relative standard deviation value of peak current towards 1 mM hydrazine was found to be 7.23% (Figure 9b). To test the stability, the electrode was stored for five days in ambient conditions. Figure 9c displays the peak current of Co 3 O 4 -1/GCE within five days. The value of peak current shows a declining trend with prolonging the time, but the peak current can still reach 86% of its initial response after being stored for five days. The obtained result suggests the long time stability of Co 3 O 4 -1/GCE.

Real Sample Test
In order to evaluate the validity of the proposed method, the Co3O4-1/GCE was applied for the detection of hydrazine in different water samples which prepared by adding known amounts of hydrazine in water samples, the results are listed in Table 5. When a known amount of hydrazine was added to distilled water, tap water, and river water, quantitative recoveries of 99.77-102.79%,

Real Sample Test
In order to evaluate the validity of the proposed method, the Co 3 O 4 -1/GCE was applied for the detection of hydrazine in different water samples which prepared by adding known amounts of hydrazine in water samples, the results are listed in Table 5. When a known amount of hydrazine was added to distilled water, tap water, and river water, quantitative recoveries of 99.77-102.79%, 98.33-101.63%, 98.63-99.30% were obtained respectively. All the results revealed the feasibility of the proposed electrode in the determination of hydrazine in water samples.

Synthesis of Co 3 O 4 Nanostructures
All the precursors were synthesized using a straightforward hydrothermal process. The detailed synthesis procedures were described as follows. 1.455 g (5 mmol) of Co(NO 3 ) 2 ·6H 2 O and 0.601 g (10 mmol) of CO(NH 2 ) 2 were mixed in 50 mL DI water and stirred continuously for 10 min. The obtained mixture solution was moved into a 100 mL autoclave and then hydrothermally at 95 • C for 24 h. The attained precipitates were concentrated via centrifugation, and repeatedly rinsed with absolute ethanol and distilled water and subsequently dried at 60 • C. The dried powder was the precursor for Co 3 O 4 -1, being Co(CO 3 ) 0.5 (OH)·0.11H 2 O. The precursor for Co 3 O 4 -2 was prepared using the similar way except extra addition 0.370 g (10 mmol) of NH 4 F, being Co(CO 3 ) 0.5 (OH)·0.11H 2 O and Co(OH)F mixture. The precursor for Co 3 O 4 -3 was synthesized similarly to the precursor of Co 3 O 4 -2 but changing the urea to hexamethylenetetramine (C 6 H 12 N 4 , HMT), being pure Co(OH)F. The detailed synthesized conditions are summarized in Table 1. To obtain the Co 3 O 4 products, all the precursors were annealed at 400 • C for 4 h. For ease of description, the products obtained from bare Co(CO 3 ) 0.5 (OH)·0.11H 2 O, pure Co(OH)F, and their mixture were designated as Co 3 O 4 -1, Co 3 O 4 -3, and Co 3 O 4 -2, respectively.

Electrode Modification
Before modification, the prepared glassy carbon electrode (GCE) was respectively polished with 1.0, 0.3, 0.05 µm alumina powder for 10 min, and then rinsed with distilled water followed by drying under ambient conditions. The obtained homogeneous slurries containing 5 mg Co 3 O 4 , 50 µL of Nafion solution (5 wt. %, DuPont 520, Wilmington, DE, USA), and 1 mL of ethanol was the mixture of all chemical together via sonication for 30 min. The Co 3 O 4 modified GCE was produced by transferring 5 µL of the above attained homogeneous slurry on the GCE, and followed dried at ambient temperature.

Conclusions
In summary, three types of Co 3 O 4 samples were prepared and utilized as electrode materials for hydrazine detection. XRD analyses demonstrated that the precursors for Co 3 O 4 -1, Co 3 O 4 -2, and Co 3 O 4 -3 were Co(CO 3 ) 0.5 (OH)·0.11H 2 O, the mixture of Co(OH)F and (Co(CO 3 ) 0.5 (OH)·0.11H 2 O, and Co(OH)F, respectively. SEM analyses showed that these three Co 3 O 4 samples possess different morphologies. The existence of F in their precursors was confirmed using SEM-EDS elemental mapping. Cyclic voltammetry results revealed that the electrochemical activity of Co 3 O 4 decreased with the increase of F content in precursors. Furthermore, the prepared Co 3 O 4 -1 and Co 3 O 4 -2 were used to fabricate hydrazine chemical sensor. The results indicated that the Co 3 O 4 -1-based hydrazine sensor possessed a high sensitivity of 32.42 µA·mM −1 , a low detection limit of 9.7 µM (S/N = 3), and a wide linear range from 0.010 to 2.380 mM. All these observed parameters were much better than those of the Co 3 O 4 -2 or Co 3 O 4 -3-based hydrazine sensors. The obtained results show that the fabricated hydrazine sensor also has good selectivity, great reproducibility and longtime stability.