Structure-property relationships for covalent triazine-based frameworks: The effect of spacer length on photocatalytic hydrogen evolution from water

Covalent triazine-based frameworks (CTFs) are a subclass of conjugated microporous polymers (CMPs) that can be used as organic photocatalysts for photocatalytic hydrogen evolution from water. Seven materials with varied spacer units from phenylene to quarterphenylene were synthesized, either by tri ﬂ uoromethanesulfonic acid (TfOH) catalysis from nitriles or by Suzuki-Miyaura polycondensation. The photocatalytic performance under visible light of all materials was systematically studied in the presence of a hole-scavenger, showing that both synthesis routes produce CTFs with similar hydrogen evolution rates (HER), but different optical properties. The highest hydrogen evolution rate in the cyclotrimerized series was found for CTF-2 with an apparent quantum yield of 1.6% at 420 nm in a mixture of water and triethanolamine with a platinum co-catalyst. Based on (TD-)DFT calculations, the highest performance was expected for CTF-1 and this discrepancy is explained by a trade-off between increased light absorption and decreased thermodynamic driving force.


Introduction
Until recently, direct photocatalytic hydrogen production using particulate photocatalysts has been dominated by inorganic materials [1e3], but there is now a growing interest in organic and polymeric photocatalysts [4]. Graphitic carbon nitride (g-C 3 N 4 ) is the most extensively studied material in this class for hydrogen evolution from water [5e10] and it has been reported to drive overall water splitting via a four-electron pathway [11,12]. One of the drawbacks of graphitic carbon nitride is the limited scope for finetuning the structure and properties as the synthesis is carried out at temperatures above 520 C [9]. Materials are typically modified by varying the nitrogen rich starting materials [13,14] or by doping methods; for example, by adding sulfur [15] or potassium hydroxide [16]. This difficulty in tailoring the photophysical properties of g-C 3 N 4 makes it attractive to study polymeric organic photocatalysts, which can be synthesized by coupling a diverse range of monomers at low temperatures, thus allowing fine synthetic control over the primary structure. Poly(p-phenylene)s [17,18], oligo(p-phenylene)s [19] and poly(pyridine-2,5-diyl) [20] show limited catalytic activity for hydrogen evolution from water in the presence of a sacrificial hole-scavenger, mostly under UV light. More recently, conjugated microporous polymer (CMP) photocatalysts synthesized by Suzuki-Miyaura polycondensation of substituted pyrenes and benzenes [21,22] have shown promising photocatalytic activities, and the absorption spectrum can be shifted into the visible region by varying the co-monomer composition. A further study on CMPs [23] explored the influence of substitution patterns and conjugation lengths on the hydrogen evolution rates. CMPs based on 4,8di(thiophen-2-yl)benzo [1,2- [24,25] have also shown good activity. Besides these CMPs, linear conjugated copolymers of planar units were shown to have even higher activities for photocatalytic hydrogen evolution in the presence of triethylamine as sacrificial hole-scavenger [26]. Subsequently, poly(phenylco-benzothiadiazole)s [27] and poly(fluorene-co-benzothiadiazole) [28] were both shown to promote hydrogen evolution from water in the presence of sacrificial hole-scavengers.
In this study, we prepared a range of structurally related CTFs using low temperature synthesis routes and explored the effect of structure and synthesis conditions on photocatalytic hydrogen evolution under sacrificial conditions.

Hydrogen evolution experiments
A flask was charged with the polymer powder (25 mg), water (20.0 mL), triethanolamine (5.0 mL), H 2 PtCl 6 solution (0.02 mL, 8 wt % in H 2 O) or with the polymer powder (25 mg), water (20.0 mL), triethylamine (2.5 mL), methanol (2.5 mL), H 2 PtCl 6 solution (0.02 mL, 8 wt % in H 2 O) and sealed with a septum. The resulting suspension was ultra-sonicated for 10 min until the photocatalyst was well dispersed before degassing by N 2 bubbling for 30 min. The reaction mixture was side-illuminated with a 300 W Newport Xe light-source (Model: 6258, Ozone free) for the time specified. The lamp was cooled and IR light cut out by water circulating through a metal jacket with quartz windows. Gas samples were analyzed with a Bruker gas chromatograph.

General procedure for the acid catalyzed trimerization polymerization
Trifluoromethanesulfonic acid and CHCl 3 were added to a flask under nitrogen and the monomer dissolved in chloroform was added with a syringe pump at a rate of 2 ml min À1 . The reaction was stirred for 48 h before it was poured onto an aqueous ammonia solution (28.0e30.0% NH 3 basis) to neutralize the trifluoromethanesulfonic acid. The suspension was left stirring for 1 h after the addition of the ammonia solution. The resulting solid was filtered, washed successively with dichloromethane, ethanol, water and methanol. The samples were dried under vacuum at 75 C overnight to obtain the polymer.

General procedure for the Pd(0)-catalyzed Suzuki-Miyaura polycondensation
The flask was charged with the monomers, N,N-dimethylformamide, an aqueous solution of K 2 CO 3 (2.0 M) and degassed by bubbling with N 2 for 30 min [Pd(PPh 3 ) 4 ] (1.2 mol %) was added and the solution was degassed by bubbling with N 2 for 10 min before heated to 150 C for 48 h. The mixture was cooled to room temperature and poured into water. The precipitate was collected by filtration and washed with H 2 O and methanol. Further purification of the polymer was carried out by Soxhlet extraction with cyclopentyl methyl ether for two days and the product was dried under reduced pressure at 125 C.

Results and discussion
We present here a series of CTFs and compare the influence of the phenylene spacer length and the synthesis methods on the photocatalytic hydrogen evolution rate. We further compare the measured properties of these materials with theoretical (TD-)DFT predictions.
All CTFs were insoluble in common organic solvents tested (i.e., acetone, methanol, hexane, dichloromethane, toluene, tetrahydrofuran, and N,N-dimethylformamide). Fourier-transform infrared spectroscopy (FT-IR) shows a significant decrease in intensity for the aryl nitrile band around 2230e2220 cm À1 compared to the monomer for the TfOH catalyzed series (CTF-1 to CTF-4), qualitatively indicating a high degree of polymerization ( Fig. S-1, ESI). The FT-IR spectra of the Suzuki-coupled polymers show similar features compared to those prepared by TfOH catalysis, except for the expected absence of a nitrile band. Even though no distinct features in the FT-IR can be assigned to hydrolysis products, partial hydrolysis of unreacted nitrile groups at the end of the polymerization or during work-up in polymers CTF1, CTF-2, and CTF-4 is conceivable. All polymers are porous to nitrogen at 77 K after ball milling, except CTF-1 and CTF-4. CTF-2 shows the highest apparent Brunauer-Emmett-Teller surface area (SA BET ) of 560 m 2 g À1 , while all others have SA BET in the range from 200 to 400 m 2 g À1 . Polymers CTF-1 and CTF-2 were reported previously to have similar FT-IR spectra and surface areas [43]. The polymers also show moderate gas uptakes for H 2 , CO 2 , Kr, and Xe (Table S-1, ESI), with CTF-2 showing the highest uptakes for H 2 (5.35 mmol g À1 at 77 K, 1 bar), CO 2 (1.25 mmol g À1 at 298 K, 1 bar), Kr (0.45 mmol g À1 at 298 K, 0.9 bar), and Xe 1.73 mmol g À1 (298 K, 1 bar). Powder X-ray diffraction patterns were featureless, indicating that both synthesis routes result in amorphous materials without any long-range order (Fig. S-7, ESI). This is probably due to the irreversible nature of the low temperature synthesis routes employed, since dynamic bond formation is believed to be necessary to give crystalline materials [44,45]. By contrast, limited crystallinity in CTFs has been reported for materials synthesized using ionothermal [29,30] and microwave routes [43]. Residual palladium was determined via ICP-OES and found to be 0.27 wt % for CTF-2 Suzuki, 0.07 wt % for CTF-3 Suzuki, and 0.65 wt % for CTF-4 Suzuki (Fig. S-28 and Table S-3,  ESI). Energy-dispersive X-ray spectroscopy also confirms that residual palladium is present in the materials synthesized via Suzuki-Miyaura coupling (Table S-4, ESI). Thermogravimetric analysis (TGA) shows that the materials are stable up to 300 C ( Fig. S-2, ESI), except for CTF-1 which has a decomposition onset of 265 C, possibly indicating a lower degree of polycondensation. Comparing both series of materials, the materials synthesized via TfOH trimerization show an earlier decomposition on-set compared to the polymers made by Suzuki coupling. This could be due to lower molecular weight of the polymers, which might also affect the photo-physical properties (infra vide). In contrast to previous reports [36,43], we found that when CTF-1 was synthesized at room temperature, the onset of decomposition was even lower (245 C; Fig. S-3, ESI) and a photocatalytically inactive material was obtained. The samples were isolated after work-up as macroscopic particles (CTF-1, CTF-2 Suzuki, CTF-3 Suzuki, CTF-4 and CTF-4 Suzuki) or as macroscopic flakes (CTF-2 and CTF-3). After ball milling, SEM images of all materials (Fig. S-30eS-33, ESI) were obtained to estimate the decrease in particle size. The particle size of CTF-1 could be reduced from around 100 mme10 mm. The average particle size of CTF-2, CTF-3, CTF-2 Suzuki and CTF-3 Suzuki were decreased from several millimeters to smaller particles in the range of a few hundred micrometers. By contrast, the particle size of CTF-4 and CTF-4 Suzuki was not decreased after ball milling, probably due to the softness and lower density of the networks.
UV-visible spectra (Fig. 2) show that the optical gap of the polymer depends on the length of the 1,4-phenylene linker between the triazine cores. Within the series CTF-1eCTF-4, the optical gap decreases from 2.95 eV to 2.48 eV (Table 1). A similar trend is observed for series CTF-2 Suzuki to CTF-4 Suzuki, but these materials fall within a much narrower range (2.93 eVe2.85 eV). As such, the synthesis route seems to have an important impact on the photophysical properties of the materials, possibly due to differences in defects within the material and end-groups. In both series, the material with four-phenylene spacers (CTF-4 and CTF-4 Suzuki) is the most red-shifted and has the smallest optical gap (see Fig. 3).
All materials are weakly fluorescent upon excitation at 280 nm and show two characteristic maxima (Fig. S-5 and Fig. S-6, ESI). The maxima fall between 367 and 374 nm for the first peak and 448 and 476 nm for the second peak. No coherent trends can be found within either series of polymers, and the observed Stokes' shifts are small, possibly an indication of rigidity.
We tested all materials for photocatalytic hydrogen evolution from water in the presence of triethanolamine (TEOA, 20 vol %) acting as a sacrificial hole-scavenger [32,34,36]. For improved performance, the materials were loaded with 3 wt % platinum as the co-catalyst by in situ photodeposition of H 2 PtCl 6 [24,36,46]. For CTF-2, Pt particles with diameters in the range of 17e33 mm were observed ( Fig. S-34, ESI). All materials, with the exception of CTF-4 and CTF-4 Suzuki, were found to produce hydrogen under visiblelight irradiation (>420 nm). The hydrogen evolution rates of all as-synthesized CTF samples were low, as the polymer powders were macroscopic flakes that were only poorly dispersible in the aqueous mixture (Fig. S-35, ESI). These issues were solved by ball milling the samples and we obtained hydrogen evolution rates of 35 mmol g À1 h À1 for CTF-1, 296 mmol g À1 h À1 for CTF-2, and 45 mmol g À1 h À1 for CTF-3 under >420 nm irradiation. When the materials were synthesized via Pd(0)-cross coupling reaction, slightly lower photocatalytic activities for the analogous polymers CTF-2 Suzuki (265 mmol g À1 h À1 ) and CTF-3 Suzuki (44 mmol g À1 h À1 ) were found when tested in the presence of the platinum co-catalyst, although these differences are probably smaller than the reproducibility of these measurements. These CTF materials show lower, but still significant, activity even without added platinum co-catalyst; e.g. for CTF-2 Suzuki, the hydrogen evolution rate drops from 265 mmol g À1 h À1 with platinum co-catalyst to 121 mmol g À1 h À1 without added platinum (Fig. 19,   ESI). As for organic solar cells [47e49], residual palladium is found as particles in these polymers, which have been suggested to act as co-catalysts in a recent report on photocatalytically active CMPs [24]. Palladium was also reported to be active in conjunction with g-C 3 N 4 [50], albeit with a lower activity than platinum, in line with the trends observed here. In contrast, CTF-2 does not produce hydrogen in absence of the Pt co-catalyst from water/ triethanolamine solution ( Fig. S-19, ESI). No hydrogen could be detected from suspensions of CTF-2 in water ( Fig. S-20, ESI), CTF-2 water platinum particle suspensions (Fig. S-20, ESI), and from water/triethanolamine solutions in the absence of photocatalysts ( Fig. S-21, ESI). The apparent quantum yield (AQY) [51] of CTF-2 was calculated to be 1.6% at 420 nm (±10 nm, FWHM, Fig. 4), which is higher than poly(p-phenylene) (AQY 420 nm ¼ 0.4%) but lower than poly(dibenzo[b,d]thiophene sulfone-co-phenylene) (AQY 420 nm ¼ 7.2%) [26]. No photocatalytic activity was found under 600 nm, 550 nm or 500 nm illumination using band-pass filters, indicating that the hydrogen evolution is indeed a photocatalytic process. Furthermore, photocurrents were measured for CTF-1, CTF-2, and CTF-4 under >420 nm illumination, these show that the process is indeed photocatalytic (Fig. S-29, ESI). The photocatalytic stability of CTF-2 was also tested under 1 Sun illumination (using an ABA-certified solar-simulator as the light source) for a total of 38 h. The performance slightly decreased after three hours, but stabilized to a constant evolution rate of 526 mmol g À1 h À1 . After 38 h, the polymer evolved more hydrogen than present in the material, ruling out decomposition of the polymer as the source of hydrogen.
Calculations were performed using our established computational approach [52,53] based around density functional theory calculations on cluster models of the CTF materials ( Fig. S-44, ESI). The results of these calculations are presented in Fig. 5      that the thermodynamic driving force for proton reduction is similar for all materials (difference between EA; red line, and the proton reduction potential; blue line). However, the predicted driving force for the oxidation of the sacrificial hole-scavenger decreases within the series from CTF-1 to CTF-4. In Fig. 5 the oxidation of triethylamine to diethylamine and acetaldehyde is shown (difference between IP; blue line, and the triethylamine oxidation potential; grey line), but similar conclusions can be drawn for any sacrificial hole-scavenger as the oxidation potential of a sacrificial hole-scavenger is independent of the material studied. Results for the excited state EA* and IP* potentials are discussed in more detail in the supporting information.

and suggest
In comparison to our measurements, this at first appears to be inconsistent with the fact that CTF-2 is the most active material, especially under illumination with a solar simulator. The photocatalytic activity of these materials is, however, the product of several properties, such as the driving force for both half-reactions and the number of photons absorbed, as well as the ease of charge transport [54]. We therefore investigated in more detail the spectral effects on the observed activity for CTF-1 and CTF-2.
Going from >420 nm irradiation to broad spectrum illumination with a Xe lamp and a >295 nm cut-off filter (Fig. S-24, ESI) leads to an increase in the hydrogen evolution rate from 35 mmol g À1 h À1 to 94 mmol g À1 h À1 for CTF-1 and from 296 mmol g À1 h À1 to 374 mmol g À1 h À1 for CTF-2 ( Fig. S-11, ESI). This increased activity is caused by absorption of UV-photons, showing that CTF-1 is more active under >295 nm UV-light and has limited absorption of visible light photons. We also compared the activity of CTF-1 and CTF-2 when illuminated with exclusively UV light using a U-340 band-pass filter (260e390 nm, see Fig. S-24, ESI for transmission characteristics). Under these conditions, CTF-1 slightly outperforms CTF-2, evolving 118 mmol g À1 h À1 vs. 111 mmol g À1 h À1 , respectively.
Considering all these measurements and our calculations, it appears that the experimentally observed maximum in visible light activity in the series for CTF-2 is the result of a trade-off between the total amount of light absorbed and the thermodynamic driving force for the oxidation reaction. Along the series from CTF-1 to CTF-4, the amount of light absorbed increases as the optical gap decreases, while the driving force for the oxidation of a sacrificial hole-scavenger, i.e. triethanolamine or triethylamine, decreases within the series.
Calculations also predict that CTF-4 should evolve hydrogen in the presence of triethylamine, which is in contrast to the observed lack of activity in experiment when using triethanolamine as sacrificial hole-scavenger. As in previous studies [21,23,26], we therefore tested the materials in water/triethylamine/methanol mixtures under >420 nm illumination. The addition of methanol has been found previously to act as a co-solvent, thus avoiding phase separation between water and triethylamine [23,26], and also to improve surface wettability for hydrophobic polymers. For ball milled CTF-2, the hydrogen evolution rates under visible-light are slightly higher when using water/methanol/triethylamine mixtures compared to water/triethanolamine mixtures (358 vs. 296 mmol g À1 h À1 ). For CTF-1 and CTF-3, the hydrogen evolution rates nearly double from 35 to 66 mmol g À1 h À1 and from 45 to 79 mmol g À1 h À1 , respectively, with the water/triethylamine/ methanol system. More importantly, this mixture leads to a low hydrogen evolution rate of 24 mmol g À1 h À1 for CTF-4, in line with our prediction that this polymer should be able to evolve hydrogen. The fact that no visual improvement could be observed in polymer dispersibility in water/methanol/triethylamine mixtures in comparison to water/triethanolamine mixtures suggests that the increased hydrogen evolution activity for CTFs might result from the easier oxidation of triethylamine compared to triethanolamine.

Conclusions
The photocatalytic hydrogen evolution rates of seven CTFs using sacrificial hole-scavengers were studied and the results were compared with the predictions of (TD)-DFT calculations. The photocatalytic activity is not significantly different for materials synthesized using catalyzed trimerization or Suzuki-Miyaura type polycondensation, despite some differences in the optical properties. In both series of polymers, UV-vis spectra show a trend of decreasing optical gaps as the phenylene spacer length increases. CTF-2 showed the highest hydrogen evolution rate under visible light using triethanolamine as sacrificial hole-scavenger and this shows that microporosity does not preclude efficient photocatalytic water reduction. This is noteworthy, since in principle microporosity might reduce charge transport rates in the solid state, which could offset gains in mass transport. Under UV light, the activity of CTF-2 increased slightly, whereas CTF-1 evolved nearly twice the amount of hydrogen in comparison to the irradiation >420 nm. These results suggest that the absorption of visible photons is limited in CTF-1 and therefore decreases its hydrogen evolution rate under these conditions. This observation is supported by calculations predicting CTF-1 to have the largest thermodynamic driving force in the series from CTF-1 to CTF-4. In general, the trend in the activity of the CTF series and the origin of the maximum in hydrogen evolution rate for CTF-2 appears to be a combination of the thermodynamic driving force and the optical gap, highlighting how the activity of a material can be controlled by structural property tuning. Residual palladium in the Suzuki-Miyaura derived materials seems to act as a co-catalyst, however, the photocatalytic activity can be further improved by the addition of platinum. Finally, CTF-2 shows a stable hydrogen evolution rate under simulated solar light for at least 38 h, evolving more hydrogen than present in the material and showcasing the stability of the material.