1 Introduction
Solar-driven water splitting is acknowledged to be a sustainable strategy for producing green H
2 to address issues such as energy shortages and environmental pollution [
1–
12]. The photoelectrochemical (PEC) system has received extensive attention due to its combination of the advantages of both photocatalysis and electrocatalysis [
13–
17]. The key for practical PEC application is to develop efficient and stable photoelectrode, which generally based on light-harvesting semiconductor. To date, a series of semiconductors have been utilized as photoelectrodes in PEC water splitting, including inorganic materials (e.g., TiO
2 [
18–
20], WO
3 [
21–
23], α-Fe
2O
3 [
24–
26], and BiVO
4 [
27–
32]), organic polymers (e.g., N,N′-bis(phosphonomethyl)-3,4,9,10-perylenediimide (PMPDI) [
33], poly(benzimidazobenzophenanthroline) (BBL) [
34], g-C
3N
4 [
35–
37], and poly[(2,6-(4,8-bis(5-(2-ethylhexyl)thiophen-2-yl)-benzo[1,2-b:4,5-bʹ]dithiophene))-alt-(5,5-(1ʹ,3ʹ-di-2-thienyl-5ʹ,7ʹ-bis(2-ethylhexyl)benzo[1ʹ,2ʹ-c:4ʹ,5ʹ-cʹ]dithiophene-4,8-dione))] (PBDB-T) [
38]), and metal-organic frameworks [
39]. In view of the advantages of easily adjustable band structure, abundant resources, high light absorption capacity, low cost and simple preparation process, polymer films for PEC water splitting have received significant attention recently [
40–
42]. However, most of the reported polymer photoelectrodes still suffers from ineffective charge separation and poor PEC performance. Therefore, exploring a breadth of polymer materials is imperative to open up new perspectives for further PEC studies.
As a polymer with excellent properties, such as strong mechanical strength, corrosion resistance, prominent electrical properties, and chemical stability, polyimide (PI) has been traditionally used in aerospace, electronic devices, transportation, and other fields [
43–
47]. In 2012, crystalline PI was synthesized with amine and dianhydride as monomers by thermal condensation, which demonstrated that it is a promising photocatalyst for H
2 generation [
48]. Compared with g-C
3N
4, PI is synthesized under relatively mild conditions (≤ 350 °C). In addition, the donor-acceptor (D-A) structure of PI is beneficial for space charge separation, which also allows its chemical structure and optical properties to be flexibly regulated by tuning amine and dianhydride monomers [
49]. To date, a wide range of measures have been explored for the improvement of the photocatalytic properties of PI, such as monomer regulation [
50–
57], doping [
58–
60], nanostructure engineering [
61–
64], and heterojunction engineering [
65–
71]. Although PI has been extensively studied as a powder photocatalyst for various applications such as H
2 evolution and organic pollutants degradation, its application as a film photoelectrode for PEC application has been less investigated.
Herein, for the first time the preparation of PI films by a simple spin-coating method was reported and their PEC properties investigated. Four dianhydrides with different conjugate sizes of aromatic unit (phenyl, biphenyl, naphthalene, perylene) were adopted to construct the corresponding D-A PI photoelectrodes. The naphthalene-based PI was found to possess the highest PEC activity. The structure–activity relationships were elucidated by both experimental and theoretical studies.
2 Experimental sections
2.1 Materials
Melamine (MA), pyromellitic dianhydride (PMDA), 3,3′,4,4′-biphenyl tetracarboxylic dianhydride (BPDA), 1,4,5,8-naphthalene tetracarboxylic dianhydride (NTDA) and 3,4,9,10-perylene tetracarboxylic dianhydride (PTCDA) were purchased from Shanghai Macklin Biochemical Co., Ltd., China. All chemicals were analytically pure, and used as received without further purification.
2.2 Preparation of PI films
Using fluorine-doped tin oxide (FTO) glass (1 cm × 3 cm) as the substrate, the PI films were obtained via a spin-coating method (Scheme R1). Briefly, 0.1 g of PI powder was ultrasonicated into 2 mL of ethanol for 4 h to obtain the upper suspension. The FTO glass underwent a 30 min ultrasonic cleaning with ethanol before being used. The side part of FTO glass was then protected with a high-temperature resistant tape (1 cm × 0.5 cm). 50 μL of the resulting suspension was uniformly applied to the conductive surface of FTO glass by spin coating at a speed of 3000 r/min. Finally, to improve surface adhesion, the FTO glass with the PI film was heat-treated at 200 °C for 1 h in the air atmosphere.
The PI powder mentioned above was synthesized in advance based on the method described in Ref. [
48]. Typically, MA (1 mol) and anhydride monomers (1.5 mol of PMDA, BPDA, NTDA, and PTCDA, respectively) were homogeneously blended and then heated at 350 °C for 4 h with a heating rate of 7 °C/min in the air atmosphere. Subsequently, the product was cleaned three times by using deionized water and ethanol respectively, and dried at 90 °C for 1 h. Finally, the PI powder was obtained with a mass yield of 85% ± 5%. The resulting products were labeled PI-PM, PI-BP, PI-NT, and PI-PT, corresponding to PMDA, BPDA, NTDA, and PTCDA, respectively.
2.3 Characterization
Thermogravimetric (TG, Netzsch TG 209 F3 Tarsus) analysis was performed to reveal the thermal stability of the samples with a heating rate of 10 °C/min in the air atmosphere. Scanning electron microscopy (SEM, Hitachi Regulus 8100) images were obtained to analyze the surface morphology. X-ray diffraction (XRD, Rigaku Ultimate IV) measurements were conducted to show the crystallinity, using Cu Kα radiation. Fourier transform infrared (FTIR, Nicolet iS20) spectra, solid-state 13C nuclear magnetic resonance (NMR, Bruker 400M) spectra, and X-ray photoelectron spectra (XPS, Thermo Scientific K-Alpha) were recorded to verify the molecular structure and chemical composition of the obtained samples. The diffuse reflectance ultraviolet-visible (UV-Vis, Shimadzu UV-3600i) spectra were measured to explore light absorption properties. N2 adsorption–desorption isotherms were obtained on Micromeritics ASAP 2460 after the samples were degassed at 160 °C for 4 h. The specific surface areas were calculated using the Brunauer-Emmett-Teller (BET) method.
2.4 Density functional theory (DFT) calculations
The Gaussian 16W program was utilized to perform the DFT calculations. To optimize the molecular geometries, the B3LYP/6-311G method was chosen. The calculated coordinates of all the optimized models were displayed in Texts S(1)–S(5) in Electronic Supplementary Material. The electrostatic potential (ESP), the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) were analyzed on the GaussView program.
2.5 PEC analysis
The electrochemical workstation (CHI760e) was used to perform the PEC analysis with a three-electrode configuration, the prepared PI photoanode (working electrode), the Pt photocathode (counter electrode), and the Ag/AgCl electrode (reference electrode). The 0.5 mol/L Na2SO4 solution was utilized as an electrolyte. To minimize the effect of film thickness on measurement results, the working electrode was illuminated on the backside under simulated sunlight (100 mW/cm2). The linear sweep voltammetry (LSV) was measured in the 0.2–1.6 V (vs. reversible hydrogen electrode (RHE)) range with a rate of 5 mV/s and the current density–time (I–T) curve was recorded at a stable voltage of 1.23 V (vs. RHE). The measured potentials (vs. RHE) were obtained from the Ag/AgCl electrode scale according to the Nernst equation ERHE = EAg/AgCl + E0Ag/AgCl + 0.0591pH, where E0Ag/AgCl = 0.1976 V at 25 °C, and pH = 7. Electrochemical impedance spectroscopy (EIS) experiments were performed with a perturbation signal of 10 mV and a frequency range of 100 kHz–10 MHz.
3 Results and discussion
3.1 Morphology and structure analysis
The PI-PM, PI-BP, PI-NT, and PI-PT were obtained via thermopolymerization of MA and different anhydride monomers (PMDA, BPDA, NTDA, and PTCDA), containing benzene, biphenyl, naphthalene and perylene units, respectively (). During the calcination process, the anhydride monomer was melted first and the anhydride parts were then dehydrated with the amino parts of MA to obtain imide oligomers, which continued to grow into a copolymer with a high crystallinity. To investigate the thermal stability of the samples, the TG analysis was performed. As shown in Fig. S1, all the samples displayed an excellent thermal stability of up to 300 °C before decomposing. This indicated that the structures of the samples were not damaged in the heat treatment process at 200 °C. The top and cross-sectional view of the PI films were examined by SEM analysis (Fig.1). The surface morphologies of the PI films are almost unchanged with the tuning of anhydride monomers. All PI films were formed by numerous tiny PI particles with a size ranging from 100 to 200 nm on the conducting surface of FTO and measured approximately to be 400 nm in thickness.
The crystallinity of the PI samples was characterized using XRD analysis (Fig.2(a)). Distinct anhydride units significantly affect the XRD spectra of the PI samples, indicating the anhydride moiety figures prominently on the chain orientation of the PI. It is not unexpected that all the samples exhibit an excellent crystallinity due to the melt polymerization process, which is reflected in the several prominent diffraction peaks between 10° and 30° [
51]. Typically, the dominant peaks of PI-PM, PI-BP, PI-NT, and PI-PT are located at 29.6°, 26.9°, 28.2°, and 27.5°, respectively, which is ascribed to the (002) plane of inter-layer stacking, indicating that PI crystals grow preferentially layer by layer in the [001] direction [
50,
51,
72–
74]. Additionally, the grain size of all the samples was calculated to be in the range of 26–35 nm according to Scherrer’s formula.
To demonstrate the successful synthesis of the PI samples, FTIR and XPS analysis were performed. The absorption peaks located at 1777, 1722, and 725 cm
−1 in Fig.2(b) are attributed to the asymmetric stretching, symmetric stretching, and deformation vibrations of the imide carbonyl groups in five-membered imide rings, respectively. The stretching vibration of the C−N−C in five-membered imide rings is observed at 1376 cm
−1. The peak at 810 cm
−1 is ascribed to the triazine structure [
72,
75–
77]. The molecular structures of the PI samples were further confirmed via XPS analysis, where the C, N, and O bonds of the PI-NT structure were analyzed (Fig.2(c)–2(e)). The three main peaks at 284.7, 285.3, and 288.1 eV in Fig.2(c) (C 1s) correspond to the C−C/C=C bonds of benzene rings of anhydride moiety, the C=O bonds of imide groups, and the C−(N)
3 bonds of triazine rings, respectively. In the N 1s spectrum (Fig.2(d)), the signals at 398.6, 399.9, and 400.8 eV are assigned to the C=N−C bonds from the amine moiety, the O=C−N−C=O bonds in five-membered imide rings, and the N−H bonds from the unreacted amine moiety, respectively. The two peaks at 531.3 and 533.2 eV in Fig.2(e) (O 1s) can be ascribed to the C−O bonds in five-membered imide rings and the C−O−C bonds from the unreacted anhydride moiety, respectively [
72,
76]. Compared to the spectra before and after use in photoelectrochemistry, no position change was found for all the peaks, indicating that no surface restructuration occurred during the PEC test. Similarly, the XPS spectra of PI-PM, PI-BP, and PI-PT were displayed in Figs. S2–S4.
In addition, the molecular structure of PI-NT reflected by the
13C NMR spectrum (Fig.2(f)) was consistent with the above verification. The
13C NMR spectrum displays several different signals representing the different carbon environments in the PI-NT structure. A peak at 167 ppm can be observed, resulting from the carbons connected to anhydride moieties in triazine rings. The carbon resonance at 163 ppm originates from the carbons of the C=O parts and carbons attached to the amino groups [
75,
78]. The carbons in the aromatic units of the anhydride moieties are responsible for the signals at 131 and 127 ppm [
79].
3.2 Optical properties and band structure
To determine the differences in the light absorption characteristics, the UV-Vis spectra were analyzed for all the resulting PI materials (Fig.3(a)). Apparently, all the samples displayed excellent optical responses. The absorption edge of PI-NT was considerably redshifted compared to that of PI-PM and PI-BP, resulting from the expansion of electron delocalization caused by the enhanced conjugated degree of fused rings in anhydride moiety [
80]. In addition, the redshift was more evident when increasing the number of fused rings, which was reflected from the light absorption spectrum of PI-PT. The result implied that the light absorption capacity of the PI was improved by introducing anhydride unites with fused rings, which is beneficial for enhancing the PEC activity by generating more charge carriers. Based on the formula (
Ahν)
1/n =
B (
hν −
Eg), where
A represents absorption coefficient,
h = 4.1356676969 × 10
−15 eV·s,
ν represents incident photon frequency,
n = 1/2, and
B is a constant, the band gap energy
Eg was estimated by making the maximum slope tangent line of the Tauc plots intersected with the axes (Fig.3(b)). The evaluated band gap values of PI-PM, PI-BP, PI-NT, and PI-PT correspond to 2.71, 2.76, 2.60, and 1.90 eV, respectively.
In addition to light absorption, the photooxidation capacity determined by the valence band (VB) level of photocatalyst is also an important index in PEC water splitting. Therefore, VB-XPS measurements were performed for the estimation of the VB level (Fig.3(c)). The
EVB, XPS values of PI-PM, PI-BP, PI-NT, and PI-PT were estimated at 1.86, 2.13, 2.13, and 1.48 eV, respectively. According to formula
EVB, RHE =
φ +
EVB, XPS – 4.44 + 0.0591pH [
81], where
φ is the work function of the instrument (4.2 eV) and pH = 7, the
EVB, RHE values of PI-PM, PI-BP, PI-NT, and PI-PT were obtained to be 2.03, 2.30, 2.30, and 1.65 eV (vs. RHE), respectively. According to the equation of
Eg =
EVB −
ECB, the conduction band (CB) values of PI-PM, PI-BP, PI-NT, and PI-PT were determined to be −0.68, −0.46, −0.30, and −0.25 eV (vs. RHE), respectively. Based on the above results, the energy band structures of all the samples were then summarized in Fig.3(d). It can be observed that the VB values of all the samples were larger than the H
2O/O
2 potential (1.23 eV (vs. RHE)), indicating they had appropriate energy levels as photoanode materials. However, PI-PT possessed a small VB value corresponding to a poor photooxidation capacity, which was considered as a barrier for efficient PEC water splitting. In contrast, PI-BP and PI-NT exhibited a higher photooxidation capacity. In addition, the reaction efficiency is also affected by the specific surface area. The BET specific surface areas of PI-PM, PI-BP, PI-NT, and PI-PT were measured to be 4.06, 4.17, 3.29, and 1.86 m
2/g, respectively. In particular, the small specific surface area of PI-PT provided less reaction sites, which would reduce the efficiency of PEC water splitting to a certain extent.
3.3 DFT calculations
With the intention of further understanding of the effect of aromatic units in the acceptor on the electron structure of PI, DFT calculations were conducted, including ESP, HOMO, and LUMO distributions. The ESP distributions can be used to analyze the redox reactive sites. As shown in Fig.4(a) and S5, the ESP distributions of all the PI models were investigated. The negative potential (oxidation sites) of all the PI models is mainly located at the connection region between the amine and anhydride moiety and at the terminal of anhydride moiety (red parts), while the positive potential (reduction sites) occupies the remaining region (blue parts). Notably, the negative potential is mainly concentrated in the O atom, which is attributed to its high electronegativity. Hence, the different aromatic units in the acceptor have no obvious influence on the redox sites of PI. HOMO and LUMO distributions were studied to further analyze the electronic structure. As shown in Fig.4(b), Fig.4(c), and S6, the HOMO and LUMO of PI-PM, PI-BP, and PI-NT models are located in the amine and dianhydride part respectively, indicating the spatial charge separation, which will contribute to a higher PEC performance. In particular, the overlap of HOMO and LUMO can be observed for the PI-PT model from Fig. S6, which will result in a rapid recombination of photogenerated carriers and decrease PEC activity.
3.4 PEC performance
PEC measurements were conducted under standard light conditions (AM 1.5G, 100 mW/cm
2) to reveal the PEC performance. First, the PI films were analyzed by using the LSV method with a voltage span from 0.2 to 1.6 V (vs. RHE) under light and dark conditions (Fig.5(a)). The currents of all photoanodes were negligible under dark conditions. Compared with PI-PM, the photocurrent response was significantly improved within the test voltage range for PI-NT and PI-PT. In particular, the photocurrent was arranged in descending order of magnitude in terms of PI-NT, PI-PT, PI-PM, and PI-BP at the bias voltage of 1.23 V (vs. RHE). Furthermore, PI-NT and PI-BP had a significantly lower onset potential (approximately 0.4 V (vs. RHE)) than PI-PM, resulting from the higher photooxidation capacity. Therefore, the PI-NT photoanode exhibited the best performance for PEC water splitting in terms of photocurrent response and onset potential. The transient photocurrent of the PI films was measured by chronoamperometry technique with chopping condition at a bias voltage of 1.23 V (vs. RHE) (Fig.5(b)). The PI-NT film exhibited the highest photocurrent response (3.3 μA/cm
2), which was 4.1, 5.5, and 2.2 times higher than that of PI-PM, PI-BP, and PI-PT film, respectively. Meanwhile, it was noticed that the transient photocurrent was generated to a higher value when brightening, then decreased, and eventually stabilized for PI-NT, which was ascribed to the imbalance between the generation and transfer of photogenerated charges. Some of the holes failed to reach the electrolyte due to the low transfer rate, resulting in the recombination with excess electrons. Efficient charge transfer and separation capabilities are another requirement for the excellent PEC activity. Therefore, EIS was utilized for the evaluation of the photogenerated carrier separation and transfer kinetics at the interface between the PI film and the electrolyte, as shown in Fig.5(c). Notably, PI-NT had the smallest arc diameter, suggesting the most efficient photogenerated carrier separation and transfer for PI-NT film [
72]. Moreover, the variation trend of the EIS was in agreement with that of transient photocurrent for all the samples.
Based on the above discussion about light absorption properties, photooxidation capacity, specific surface area, and charge transport and separation, the resulting PEC properties of all the samples can be interpreted accordingly. The low PEC activity of PI-PM and PI-BP largely resulted from the poor light absorption and high interfacial impedance. In addition, the obstruction of charge transfer by distorted backbone chains may be another reason for the low performance of PI-BP [
51]. The small specific surface area and weak photooxidation capacity were identified as key factors for the moderate PEC activity of PI-PT. Therefore, PI-NT was demonstrated to be a promising photoanode material with a wide-range visible light absorption, an efficient charge separation and transport, and a strong photooxidation capacity.
Despite the enormous potential advantages of PI as a photoanode material, the PEC activity of the PI films reported here still need to be improved for efficient water splitting. Nowadays, various PIs with a high photocatalytic performance have been obtained through heterojunction engineering, defect engineering, and other effective strategies [
49], which are expected to be further optimized to contribute to efficient PEC water splitting. In addition, it is important for the PI film to achieve a strong adhesion on the conducting surface of substrate. Apart from simple spin coating, the
in situ growth is considered worthy of investigation in subsequent work.
4 Conclusions
In summary, a range of PI photoelectrode films were obtained via polycondensation of MA with dianhydrides containing different conjugate sizes of aromatic unit. It was found that the fused ring was prominent in improving the light absorption capacity of PI, but excessive fused rings were unfavorable for the photogenerated charge separation. Of all the samples, the PI-NT film exhibits the highest photocurrent response, which is ascribed to its wide-range light absorption, efficient charge separation and transport, and strong photooxidation capacity. However, the photocurrent response of the PI film presented here needs to be improved for efficient PEC water splitting, which can be enhanced by catalyst modification (e.g., elemental doping, and composite engineering) or optimizing the preparation method of films in subsequent work. This work is not only a starting point of PI films for PEC water splitting, but also an enlightenment for the rational design of polymer photocatalysts for efficient PEC applications.