Effects of N-alkylation on anticorrosion performance of doped polyaniline/epoxy coating

1. Introduction

Al alloys are widely applied in marine environments. The aggressive Cl^- existing in marine environments induces pitting corrosion on Al alloys [[1], [2], [3], [4], [5]]. Generally, organic coatings are the most preferable method to protect Al alloys. Functional fillers are often added into traditional organic coatings, such as epoxy or polyurethane-based organic coatings, to achieve a better anti-corrosion performance. Among these fillers, polyaniline (PANI), as an organic filler, has been receiving increasing attention, because of its ability to passivate metal surfaces [[6], [7], [8], [9], [10], [11], [12]].

The PANI additives can not only enhance the corrosion resistance of the organic coatings but also improve its mechanical properties. Gupta et al. [8] investigated the corrosion protection of polyaniline-lignosulfonate/epoxy (PANI-LGS) coating for AA2024-T3 Al alloy for 30 d immersion period in 0.6 M NaCl solution. It is stated that the coating with a 5 wt% PANI-LGS additive revealed the highest impedance value, which was up to 10⁸ Ω cm² after 30 d immersion time; corrosion protection properties were improved by the formation of a constantly thickening oxide film. Bandeira et al. [9] studied the anticorrosion performance of PANI/polyvinyl chloride blended coating for AA7075-T6 Al alloy in 3.5% NaCl solution. The low water uptake, high pore and charge transfer resistances indicate the superior corrosion protection performance of the coating. Jia et al. [10] investigated the role of DBSA-PANI (dodecylbenesulfonic acid doped PANI) on the properties of epoxy coatings, it was found that both mechanical properties and the curing reaction of epoxy coatings were improved due to the mutual effects of the two ingredients. Therefore, it can be concluded that PANI-filled organic coatings exhibit superior corrosion resistance in corrosive aqueous NaCl solution and enhanced mechanical properties, especially after incorporation into traditional organic coatings like epoxy coating. Our previous researches [13,14] also focused on the protective properties of sulfosalicylic acid-doped polyaniline/epoxy coating (PANI-SSA/epoxy coating) and the capacity of PANI to passivate metals. It was found that PANI-SSA/epoxy coating can play a protective role for Al alloys by forming a complete oxide layer on the Al alloy matrix, whereas the barrier properties of the coating are not good because of the poor dispersion of PANI-SSA in epoxy matrix. Holes and aggregation defects caused by PANI-SSA addition in the coatings always deteriorate the coatings’ barrier properties. Consequently, the coatings cannot satisfy the need in actual application. Thus, investigation of effective dispersion of PANI is important for PANI/epoxy composite coating.

In fact, PANI has a rigid structure which makes it difficult to dissolve and disperse in organic solvents as well as in organic coating matrix [15,16]. Generally, PANI is modified by N-alkyl substitution with alkyl bromide to improve its solubility. The solubility and process-ability of PANI can be promoted by the incorporation of flexible chains [[16], [17], [18]]. Zhao et al. [18] indicated that the N-alkylation reactions of PANI with various alkyl halides follow the nucleophilic substitution mechanism. Nucleophilic substitution reaction includes aromatic nucleophilic substitution reaction and aliphatic nucleophilic substitution reaction, N-alkylation substitution reaction of PANI follows the aliphatic nucleophilic substitution reaction mechanism. The chain length of alkyl halides, solvents types, reaction temperature and time would influence the reaction rate and N-alkylation degree. Hwang et al. [19] conducted the N-alkylation of PANI (emeraldine base form) by alkyl halides of different chain length. When number of carbon atoms is more than 12, the polymer film becomes flexible. In addition, the solubility of PANI-EB in organic solvents is improved after modification with alkyl halides. Furthermore, studies have shown that alkyl chains linked to fillers will facilitate the curing performance and barrier property of epoxy coating [[20], [21], [22]]. Therefore, we believe that the anticorrosion property of the PANI/epoxy composite coating could be enhanced by N-alkylation of PANI. However, little research has been reported on the direct N-alkylation of doped PANI [15]. Thus, it is unclear whether the N-alkylation of acid-doped PANI is successful, and the corrosion protection property of N-alkylated PANI/epoxy coatings is also unknown yet.

In this work, N-alkylated PANI-SSA has been investigated, in which two alkyl bromides were used to modify the PANI-SSA. The characteristics of the N-alkylated PANI-SSA are presented and compared. The performance of N-alkylated PANI-SSA/epoxy coating is examined.

2. Experimental

2.1. Modification of polyaniline

Two alkyl bromides, bromopentane (C₅H₁₁Br) and bromododecane (C₁₂H₂₅Br), of different chain lengths were used. The alkyl halides were diluted to a concentration of 10 vol.% in the polar solvents DMF and IPA. A water bath was used to maintain the solutions at 80 ℃ constant temperature. PANI-SSA fillers (bought from Taizhou yongjia trading company, the average diameter is 7-10 μm and the conductivity is about 2 S/cm) were modified in the heated mixed solution for a while, then filtered and washed with an excess of the corresponding solvents to remove the unreacted alkyl bromides. Acetone was used to wash off the excess solvents, and the reaction products were finally filtered using a Büchner funnel. The reaction products were dried in an oven before further use. The PANI-SSA after different N-alkylation processes are named C₅-IPA (modified in IPA with C₅H₁₁Br), C₁₂-IPA (modified in IPA with C₁₂H₂₅Br), C₅-DMF (modified in DMF with C₅H₁₁Br), C₁₂-DMF (modified in DMF with C₁₂H₂₅Br), respectively.

Fourier infrared spectra (FTIR) were recorded with KBr pellets using a FTIR spectrometer (Nicolet Magna-IR IR560) within the wavenumber range of 2000-500 cm^-1 to obtain the chemical structural characteristics of the fillers (PANI-SSA, C₅-IPA, C₁₂-IPA, C₅-DMF and C₁₂-DMF).

X-ray photoelectron spectroscopy (XPS) performed on Escalab 250 (ThermoFisher VG) was used to characterize the composition and structure of all the fillers. The analysis was conducted at 6.1 × 10^-8 Pa using a monochromatized Al Kα (1486.6 eV) X-ray source. All spectra were energetically calibrated by placing the adventitious C 1s peak at 284.6 eV on the binding energy scale. XPSPEAK software was further used to resolved the N 1s spectra of all the fillers into component peaks.

2.2. Coating preparation

5083 Al alloy (with chemical composition shown in Table 1) was cut into 10 mm × 10 mm × 5 mm for electrochemistry tests and the corresponding adhesion measurements. All the substrate samples were ground with SiC paper to 800 grit size, and then put into acetone to degrease, washed with distilled water and subsequently dehydrated with ethanol. Finally, all samples were put into desiccators for further use.

Three different organic coatings were prepared for present study. Xylene was used as diluent and the epoxy E-44 (WRS6101, Nantong Xingchen Synthetic Material Co. Ltd) was dissolved in it. Following, PANI-SSA, C₅-DMF, C₁₂-DMF, which were used as fillers, were added into the epoxy/xylene solution. The mixed paint was then ultra-sonicated for 1 h and stirred for 30 min, respectively. The curing agent is polyamide (Shenyang Zhentraid Anti-corrosion Materials Co. Ltd). The mass ratio of the epoxy resin, the curing agent and the solvent is 1:0.8:0.5. The three fillers (PANI-SSA, C₅-DMF and C₁₂-DMF) contents are 6% (mass percent) of the dry coating mass. As xylene solvent is considered to be completely volatilized, the dry film mass, recorded as m₀, is the mass of epoxy resin plus the mass of curing agent. Thus, the mass of the fillers is 6% × m₀. Finally, the prepared coatings are PANI-SSA/epoxy coating, C₅-DMF/epoxy coating and C₁₂-DMF/epoxy coating, respectively. To get the coating/Al samples, the prepared coating paints were applied on the Al alloy surface by a hand brush. The free film samples were used for SEM cross-sectional and gravimetric measurements, and the method to prepare free film is as follows: firstly, the paints were brushed on a silica gel plate, after being cured in an oven at 40 ℃ for 4 h and 60 ℃ for 20 h, finally the free film was peeled off from the plate. For gravimetric test, the cured free films were cut into the specific dimension, while for SEM cross-sectional tests the cured free films were brittle-fractured at low temperature. The coating was cured in an oven at 40 ℃ for 4 h and 60 ℃ for 20 h, and then at ambient conditions (25 ℃, 30% RH, 168 h) to volatilize the remaining solvent (xylene) in the cured coatings. A hand-held electronic gauge (PosiTector 6000 of Defelsko) was used to test the average dry coating thickness according to ISO 2808-1997. The coating thickness was found to be 100 ± 10 μm. The samples were stored in a desiccator prior to all tests to keep them dry and avoid property changes.

2.3. Characterization

The surface and cross-sectional morphologies of the coatings were characterized by scanning electron microscopy (SEM, Inspect F50).

Gravimetric tests were conducted to obtain the water absorption of the three free films. According to previous research [13,23,24], the water absorption was calculated using the equation below:

$Q_{t}=(m_{t}-m_{0})\times 100\%/m_{0}$ (1)

where m_t is the free film mass at time t, and m₀ is the mass before immersion. Q_t is the water absorption at time t. The Q_t-t curves can be obtained by plotting Q_t at different time intervals. The water absorption rate was calculated from the Q_t-t curve by differentiating the water absorption.

EIS is carried out in a three-electrode cell with 3.5% NaCl solution using an Autolab electrochemical station (Metrohm). The coated-metal samples were used as working electrode, a platinum plate of size 20 mm × 20 mm and a saturated calomel electrode (SCE) were selected as the counter electrode and the reference electrode, respectively. A sinusoidal ac perturbation of 20 mV (rms) amplitude was used at initial immersion time due to the highly resistive system of coatings/metal. After a period of immersion time (about 24 h), a 10 mV (rms) amplitude was used because the system had stabilized. Subsequently, the EIS measurements were carried out at the open circuit potential (OCP) with the two different amplitudes over a frequency range of 100 kHz to 10 mHz. All experiments were conducted in a Faraday cage to minimize the influence of external noise on the results. ZSimpWin software (Princeton Applied Research) was finally used to analyze and fit to the electrical equivalent circuit models of the collected EIS data.

The adhesion is the bonding force between organic coating and metal, which is also the force required to peel the organic coating from the metal substrate. The adhesion tests of coatings/metal samples were carried out by PosiTest Pull-off Adhesion Tester after the corresponding EIS measurement in the light of ASTM D4541-02, a standard evaluation method to estimate the force needed to separate the coatings from the metal surface. A digital camera (Nikon) was used to depict the corrosion macro-morphologies of the underlying 5083 Al alloy after immersion.

3. Results and discussion

3.1. Structural characteristics

The FTIR spectra of PANI-SSA before and after N-alkylation are compared in Fig. 1. The spectrum of all the fillers exhibited two distinct peaks, the characteristic absorption peak of quinoid (Q) ring unit appeared at 1560 cm^-1 and benzene (B) ring unit at 1482 cm^-1, which both are associated with the aromatic ring stretching. The increased intensity and significant broadening of the characteristic absorption peak at 1112 cm^-1 of N-alkylated PANI-SSA implies a higher degree of electron delocalization in the molecule and less intermolecular interactions. Thus, with weaker interactions, a better dispersion of N-alkylated PANI-SSA in solutions would be possible to realize. Additionally, a new peak observed at 1233 cm^-1 was due to the aliphatic C-N stretching of N-alkylated PANI [18,19,25]. It suggests that the alkyl substituents are linked to the nitrogen of the PANI-SSA chain, the N-alkylation reaction of PANI-SSA equation can be seen in Fig. 2.

To clarify the N-alkylation degree of PANI-SSA, the N 1s XPS core-level spectra of all the fillers (PANI-SSA, C₅-IPA, C₁₂-IPA, C₅-DMF and C₁₂-DMF) were compared, the results are shown in Fig. 3. The N 1s spectra of N-alkylated PANI-SSA contains three peaks, including =N- at 398.2 eV (± 0.2 eV), -NH- at 399.4 eV (± 0.2 eV) and N⁺ at 401.3 eV (±0.2 eV). Obviously, N⁺ was formed when the alkanes attached to the chain of PANI-SSA. It can be concluded that the higher the N⁺ fraction, the higher the N-alkylation degree of PANI-SSA [18,19,26]. The proportion of N⁺ increased in the N 1s spectra of N-alkylated PANI-SSA (Table 2), while the = N- and -NH- proportion decreased (Table 2). As there is no N⁺ peak (398.2 eV) in pristine PANI-SSA, the high percentage of N⁺ in N-alkylated PANI-SSA is due to the formation of the positively charged nitrogen during N-alkylation procedure. The results also demonstrate that the N⁺ fraction increased after N-alkylation, while the fraction of the -N- decreased (Table 2), further confirmed that N-alkylation of PANI-SSA occurs on the nitrogen of the -N- units rather than the = N- units [18,19]. The N-alkylation of PANI-SSA by the two bromoalkanes in DMF provided higher degree of substitution, for C₁₂-DMF and C₅-DMF the N-alkylation degrees were 39.9% and 32.3%, respectively. N-alkylation of PANI-SSA is a nucleophilic substitution reaction, which easily occurs in a polar solvent. The greater polarity of the solvents, the higher the degree of N-alkylation. The polar solvent can reduce the energy of the ion transition state in N-alkylation step, so it facilitates the reaction [[15], [16], [17], [18], [19]]. The results showed N-alkylation degree is higher in DMF solution because of its larger polarity. The linked alkane chain on PANI-SSA facilitates the compatibility between the fillers and epoxy matrix, and thereby may decrease the coating defects (Table 3).

To test the dispersion of the N-alkylated PANI-SSA, a sedimentation measurement was carried out to monitor whether the filler stratification phenomenon exists in epoxy/xylene solution [27,28]. The fillers, PANI-SSA and N-alkylated PANI-SSA, were grinded and then dispersed in the prepared epoxy/xylene solution completely. Epoxy/xylene solution was selected as a solvent because the fillers were dispersed in epoxy/xylene solution during coatings preparation process. In general, if there is a superior compatibility between fillers and the epoxy/xylene solution, the fillers would not show apparent sedimentation, nor aggregations or pores in the resulting coatings. Fig. 4 shows the digital photos of the dispersion performance of the pristine PANI-SSA and the N-alkylated PANI-SSA in epoxy/xylene solution after different time intervals. Obviously, all the mixed solutions initially exhibited a black color. The black spots sticking on the upper part of the sample bottles are the filler aggregations when the liquid (epoxy/xylene) flows down. However, the pristine PANI-SSA obviously precipitated just within 5 d of standing, and showed a larger color difference (the upper solution is light-green, while the bottom solution appears dark-green) after 12 d of sedimentation test. This indicates that the sedimentation of PANI-SSA is severe, the compatibility between pristine PANI-SSA and epoxy/xylene solution is poor. The C₅-IPA and C₁₂-IPA also showed slight settlement and a dark green coloration in the upper liquid after settling only for 5 d, especially for C₅-IPA, which was found with an obvious sedimentation, the upper liquid became light-green. However, C₅-DMF and C₁₂-DMF remained highly dispersed in epoxy/xylene solution even after 12 d of standing and showed a uniform and near-black colored mixed solution. These results show that N-alkylation of PANI-SSA in DMF solution significantly prevented the fillers from sedimentation in epoxy/xylene solution, which may correspond to the higher degree of N-alkylation. Consequently, a higher dispersion of filler in the epoxy/xylene solution enables high compatibility between N-alkylated PANI-SSA and epoxy resin, in addition to a better combination between the two components which indeed help to enhance the barrier property of the coatings. Degree of N-alkylation in DMF solution is higher, and the N-alkylated PANI-SSA also show a better settlement performance and better compatibility with epoxy resin.

3.2. SEM measurements

The SEM surface morphologies of PANI-SSA coating and N-alkylated PANI-SSA coatings are shown in Fig. 5, and their SEM cross-sectional morphologies after brittle-fracture are presented in Fig. 6. The PANI-SSA coating revealed a scraggly surface with many holes, which definitely allow the permeation of corrosive particles (H₂O, Cl^- and O₂) into the coating, leading to Al alloy corrosion. Besides, large aggregations were found in the cross-sectional morphology, which is indicative of a poor dispersion performance of PANI-SSA fillers in epoxy resin. However, the two N-alkylated coatings were homogeneous, smooth and dense in the surface SEM morphology (Fig. 5(b) and (c)) with no obvious holes and rough areas. Small aggregations along with little fracture stripes were found in C₅-DMF coating in its cross-sectional morphology. The cross-sectional microstructural images of C₁₂-DMF coating are homogeneous with plenty of cracked fringes (shown in Fig. 6(f)). These results demonstrate a superior compatibility of N-alkylated PANI-SSA with epoxy resin when compared with pristine PANI-SSA. The uniformity and compactness of C₁₂-DMF coating has been promoted by the better compatibility between epoxy resin and the long chains of C₁₂-DMF, which enables uniform and compact coatings and facilitate superior anticorrosion properties of N-alkylated PANI-SSA/epoxy composite coatings.

3.3. Water absorption

To better understand the barrier properties and corrosive ion permeation behavior of the coatings, water absorption measurements were performed. Fig. 7 shows the water absorption curves (Q_t-t) of the PANI-SSA/epoxy coating and the two N-alkylated PANI-SSA/epoxy coatings in 3.5% NaCl solution after different time intervals. Initially, the Q_t of the three coatings increased significantly, then approached maximum Q_t at about 144 h and remained steady (from 144 h to the end of the experiment). The value of the stable water absorption in Q_t-t curves is called saturation water absorption. The saturation time is almost the same for all the three coatings at 144 h, and the corresponding saturation Q_t is 5.38%, 4.58% and 3.89%. Obviously, there is a noticeable difference between the saturation water absorption values of the three coatings. The main reason for organic coating’s degradation is the penetration of corrosive solution, so lower water absorption of the N-alkylated coating enables better anticorrosion performance [[29], [30], [31]]. Thus, the lower saturation water absorption of N-alkylated coating demonstrates that the addition of C₅-DMF and C₁₂-DMF in epoxy resin prevented the coatings from adsorbing corrosive solution. Lower water absorption results in less accumulation of corrosive particles (H₂O, O₂, Cl^- etc.) at the coating/metal interface, thereby the coatings can effectively protect the Al alloy for long-term immersion. This better barrier performance of N-alkylated PANI-SSA/epoxy coating is mainly due to the smaller number of aggregations and pores (shown in Fig. 5, Fig. 6), which facilitate the formation of a denser coating structure.

The instantaneous diffusion velocity (V_i) was calculated by dividing and differentiating the Q_t-t curves, the result is shown in the inset of Fig. 7. V_i decreased rapidly (from 350 to 11 mg h^-1 cm^-2 for PANI-SSA/epoxy coating, and 150 to 9 mg h^-1 cm^-2 for C₁₂-DMF/epoxy coating and 240 to 8 mg h^-1 for C₅-DMF/epoxy coating) within 96 h immersion time. At the beginning, the Vi of PANI-SSA/epoxy coating was higher mainly because there were many holes in its surface. The holes provided fast access for the corrosive electrolyte diffusion. After N-alkylation, alkyl chains were linked to the PANI-SSA chain, which facilitates a better compatibility with epoxy resin. As the results show, less holes and aggregations were formed. Less defects also resulted in lower water diffusion rate and smaller saturation Q_t. Especially for C₁₂-DMF/epoxy coating, C₁₂-DMF filler was uniformly dispersed in the epoxy matrix because of the superior compatibility of the long-chain alkane (-C₁₂H₂₅) with epoxy resin. Consequently, the corrosion of Al substrate was slowed down, the N-alkylated coatings exhibited a better protection for underlying Al alloy.

3.4. EIS measurements

The barrier property and anticorrosion performance of the epoxy coatings after compounding PANI-SSA, C₅-DMF and C₁₂-DMF additives can be evaluated and compared by in situ EIS measurements. The EIS were carried out at open-circuit potential in aqueous 3.5% NaCl solution after different immersion times.

Fig. 8 shows the Bode and Nyquist plots of the PANI-SSA/epoxy, C₅-DMF/epoxy, and C₁₂-DMF/epoxy coated electrodes in aqueous 3.5% NaCl solution after different immersion times (1 h, 1 d, 5 d 15 d and 30 d). The electrical equivalent circuits shown in inset were used to fit the corresponding EIS data. The fitting error results were also presented in the figure. For PANI-SSA coating, there are two obvious stages (distinguished by the number of capacitive loops), the two equivalent circuits shown in inset corresponds to the two stages, one containing the solution resistance (R_s), the coating capacitance (C_c) and the coating pore resistance (R_c) was used to fit the impedance date after 1 h. Considering the corrosive particles (H₂O, O₂, Cl^- et al.) reach the Al alloy surface through micro-pores of the coating after immersion (1-30 d), the other equivalent circuit which adds the charge transfer resistance (R_ct) and the double layer capacitance (Q_dl) corresponding to interfacial reaction on the underlying Al alloy is applied to fit the experimental data. The Nyquist plots of N-alkylated coatings consist only of one capacitive loop, indicating that the N-alkylated coatings act as barrier layer. The equivalent circuits (Fig. 8(d) and the inset in Fig. 8(f)) were used to fit the EIS results. The plot of impedance modulus at low frequency (|Z|_0.01Hz) against immersion time for the three coatings is shown in Fig. 9. |Z|_0.01Hz includes coating resistance and interfacial charge transfer resistance, it can be used to characterize the overall performance of an organic coating. The anticorrosion performance of the coating can be easily obtained by monitoring |Z|_0.01Hz changes. |Z|_0.01Hz is an appropriate parameter for the characterization of the anticorrosion property of the coatings, as reported in previous studies [11,[27], [28], [29], [30], [31], [32], [33], [34]]. After the initial 1 h immersion time, the Nyquist plot reveals one capacitive characteristic, with one capacitive loop on the equivalent circuits, the value of |Z|_0.01Hz exceeded 10¹¹ Ω cm² for all three coatings. Subsequently, |Z|_0.01Hz of the PANI-SSA coatings decreased to 10⁸ - 10⁹ Ω cm², which was due to the rapid entry of corrosive electrolyte and its reaction with the underlying metal. The two capacitive loops in the equivalent circuit indicate the occurrence of metal corrosion. In contrast, the Nyquist plots of C₅-DMF/epoxy and C₁₂-DMF/epoxy coatings presented only one capacitive loop after 30 d immersion. That meant the barrier properties of the two modified coatings is superior, with no corrosion occurring at the coating/metal interface. |Z|_0.01Hz of C₅-DMF/epoxy and C₁₂-DMF/epoxy coating stayed > 10¹⁰ - 10¹² Ω cm², especially for C₁₂-DMF/epoxy coating, |Z|_0.01Hz stabilized at around 1.11 × 10¹¹ Ω cm² after 30 d immersion, confirming its better barrier performance. Possibly the longer molecular chain of C₁₂-DMF/epoxy led to improved compatibility between the fillers and the epoxy matrix when the curing reactions occur. Thus, C₁₂-DMF/epoxy coating yielded the best corrosion protection property for the underlying 5083 Al alloy.

The adhesion results also demonstrate the superior protection property of C₁₂-DMF/epoxy coating. Fig. 10 shows the wet adhesion and the macro-morphologies of Al alloy surface beneath three coatings measured after the in situ EIS tests (after 30 d immersion in 3.5% NaCl solution). The adhesion of PANI-SSA/epoxy coating was 2.13 MPa after 30 d immersion, many corrosion pits along with a few blisters appeared on the underlying Al alloy. In comparison, the adhesion of C₅-DMF/epoxy and C₁₂-DMF/epoxy coatings stayed stronger at 3.26 and 3.44 MPa, respectively. Especially the Al alloy surface beneath C₁₂-DMF/epoxy coating kept metallic luster and no obvious corrosion was observed. These results further demonstrate the most superior anticorrosion performance of the C₁₂-DMF N-alkylated PANI-SSA/epoxy coating.

4. Conclusions

(1) PANI-SSA was N-alkylated with C₅H₁₁Br and C₁₂H₂₅Br in polar solvent IPA and DMF, and characterized by FTIR, XPS and sedimentation experiments. Results showed that alkanes were successfully linked onto the PANI-SSA chain. The C₅-DMF and C₁₂-DMF yielded better dispersion in epoxy/xylene solution after modification because more alkanes were linked to PANI-SSA chain.

(2) The anticorrosion performance of the PANI-SSA/epoxy, C₅-DMF/epoxy and C₁₂-DMF/epoxy coatings were examined by EIS and adhesion measurement. The protection property of PANI-SSA coating was enhanced by N-alkylation of PANI-SSA. C₁₂-DMF/epoxy coating yielded the highest values of impedance modulus and best anticorrosion performance. After being linked with an alkane, PANI-SSA was well dispersed in epoxy resin, thus reducing the number of holes and aggregations in the coatings and providing superior anticorrosion properties.

Acknowledgments

This work was financially supported by the National Natural Science Foundation of China (Nos. 51622106 and 51871049).

---