Edinburgh Research Explorer Size-Selective Photoelectrochemical Reactions in Microporous Environments

Graphitic carbon nitride (g-C 3 N 4 ) with photo-attached platinum (Pt@g-C 3 N 4 ) is known to generate hydrogen under illumination in aqueous environments in the presence of carbohydrate hole quenchers. Here, Pt@g-C 3 N 4 is embedded into a polymer of intrinsic microporosity (PIM-1) host material with a molecularly rigid structure to maintain active unblocked catalyst surfaces and to control transport to/from the photocatalyst. A Clark-type oxygen/hydrogen sensor is employed with Pt@g-C 3 N 4 embedded into PIM-1 applied as a film to be the gas-permeable sensor membrane. Oxygen reduction and hydrogen production are observed in situ as a function of light exposure and quencher concentration. Significant size-selectivity favouring smaller more flexible saccharides or carbohydrate/ hydrocarbon quenchers is observed and attributed to rate limiting PIM-1 micropore transport. Effective hydrogen production through a Teflon membrane is demonstrated. The underlying hydrogen production/ photocurrent enhancing effects of the microporous PIM-1 film on the photochemical process are revealed.


Introduction
Photochemical hydrogen generation in aqueous solution environments is important for renewable energy generation [1] and for biomass conversion. [2]For photocatalyst materials such as graphitic carbon nitride, g-C 3 N 4 , waste polymer degradation/ conversion by photocatalysis has been reported. [3]Many other types of photocatalysts have been developed to generate hydrogen including TiO 2, [4] ZnO, [5] and BiVO 4. [6] Also new types of sustainable co-catalysts for hydrogen generation have been developed based for example on MoS 2 [7] and on metal phosphides. [8]Particularly strong contenders for effective hydro-gen generation is the family of graphitic carbon nitrides or g-C 3 N 4 and their hybrids. [9]Graphitic carbon nitride has been initially discovered being one of the oldest artificial polymers [10] and recently extensively studied as a semiconductor photocatalyst from 2009 onwards. [11]Modified g-C 3 N 4 materials have been developed by doping, [12] by heterojunction formation, [13] by structural tuning, [14] and by single atom immobilisation. [15]uthoritative reviews on g-C 3 N 4 in photocatalysis have appeared. [16]New opportunities arising from the combination of g-C 3 N 4 with polymer materials have been highlighted. [17]olecular engineering has been employed to broaden and tune properties of g-C 3 N 4 materials [18] and new emerging areas of applications for the g-C 3 N 4 photocatalysts have been reviewed. [19]olymer supported nanocomposites for environmental applications have raised attention for combining advantages of polymers and nanoparticles. [20]In recent studies, we have investigated the immobilisation of particulate photocatalysts embedded into films of microporous polymers. [21]Polymers of intrinsic microporosity (or PIMs) [22] are formed with a highly contorted molecularly rigid backbone, and therefore provide uniquely processable materials that are soluble in organic solvents and readily cast into microporous films and deposits. [23]mportantly, the rigid polymer backbone has been proposed to not strongly interact with the surface of nanocatalyst guests to maintain an active unblocked catalyst surface. [24]At the same time photocatalysts would be expected to be less likely to degrade the PIM polymer host due to insufficient molecular interaction.
Here, PIM-1 is employed as a film over a deposit of Pt@g-C 3 N 4 on the Teflon membrane of a commercial Clark-type sensor (used here for detecting both O 2 and H 2 , see Figure 1).PIM-1 films in conjunction with photocatalysts have been reported previously for Pt@TiO 2 [25] and for Pt@g-C 3 N 4 on platinum surfaces [26] or on hydrogen-permeable palladium membranes. [27]A key feature observed for PIM-1 and similar polymers is the ability to affect gas reactivity in liquid (or triphasic) environments. [28]Here, the effects of PIM-1 on photoelectrochemical hydrogen generation from Pt@g-C 3 N 4 in the presence of aqueous quenchers (hexanol, sorbitol, gluconic acid, glucose, sucrose, or raffinose) is reported.

Reagents
Graphitic carbon nitride [29] and PIM-1 [30] were prepared following literature procedures.The obtained graphitic carbon nitride was loaded with platinum nanoparticles cocatalyst to give Pt@g-C 3 N 4 by photo-deposition as described previously. [26]All other chemicals reagents were purchased in analytical reagent quality from Sigma Aldrich and used as received without further purification.Ultrapure (18.2 MΩ cm at 18 °C) water from a Thermo-Fisher water purification system was used for all experiments.

Instrumentation
Electrochemical measurements were conducted with an Autolab PGSTAT (Metrohm, UK) controlling a commercial polarographic dissolved oxygen Clark probe (HI-76407/2, Hanna Instruments, UK).A LED light (λ = 385 nm, Thorlabs, UK, 1200 mA, 100 % intensity) was used in all photoelectrochemical procedures.Scanning electron microscopy (SEM) images were captured with a Hitachi SU-3900 SEM with attached Oxford Instruments Ultim Max 170 mm 2 EDX detector.Transmission electron microscopy (TEM) was performed on a JEOL JEM-2100Plus instrument with a 200 kV maximum accelerating voltage.PXRD pattern were recorded in transmission mode on a STOE STADI P equipped with a multi Mythen detector using monochromated Cu-Kα radiation (1.54060 Å).

Procedures
For the photoelectrochemical measurements, Pt@g-C 3 N 4 was embedded into PIM-1.Both were suspended/dissolved in chloro-form solution with a weight ratio of 5 : 1 and concentration 1 mg cm À 3 .A loading of 60 μg Pt@g-C 3 N 4 with 12 μg PIM-1 was immobilised centrally on the Clark probe PTFE (Teflon) membrane cap by drop-casting.After drying in air, the Clark probe is immersed in the testing solution and aligned to the LED light source (LED, λ = 385 nm, Thorlabs, UK, type M385LP1, operated at 100 % nominal power, distance approx. 1 cm).The experimental setup is illustrated in Figure 1. Figure 2 shows typical SEM images of the deposited catalyst composite on the gas permeable PTFE membrane.These images reveal the distribution of PIM-1 with photocatalyst on the PTFE surface.The approximate location of the Pt disk electrode in the Clark probe is indicated.Due to the high content of solid g-C 3 N 4 in the composite, the deposit appears course/porous on a 1 to 10 μm scale and this is likely to affect the transport of gaseous products and the Clark sensor responses.Effects from this uneven photocatalyst film morphology could be important and will require further study in future.
EDX analysis (see Figure S1) proved the successful deposition of platinum nanoparticle cocatalyst on the g-C 3 N 4 sheets.In addition, TEM images (Figure S2) show the morphology of the Pt@g-C 3 N 4 photocatalyst.Platinum nanoparticles are present with a diameter of typically 2-3 nm.The PXRD pattern of a prepared Pt@g-C 3 N 4 sample (Figure S3) show distinct platinum peaks compared to the powder diffraction peaks for g-C 3 N 4 .Chronoamperometry data was recorded with a fixed applied potential of 0.6 V vs. Ag/AgCl for hydrogen sensing and À 0.7 V vs. Ag/AgCl for oxygen sensing.

Monitoring Photoelectrochemical Oxygen Consumption and Hydrogen Generation in the Presence of Glucose
In previous reports electrochemical methods based on photocatalyst-modified platinum electrodes were employed to detect hydrogen generation from Pt@g-C 3 N 4 embedded into PIM-1. [26]his direct electrochemical approach is versatile, but also sensitive to effects from impurities.PIM-1 is known to undergo some degradation/restructuring with UV light and in the presence/absence of oxygen. [31]Exploratory experiments (not shown) suggested that the detection of oxygen with immersed platinum electrodes was unreliable probably due to fragments from PIM-1 photodegradation impacting on the catalytic activity of the platinum electrode surface.Therefore, a different (indirect) approach based on the Clark probe was studied.Clark probes have been developed for the detection of dissolved oxygen, which permeates through a Teflon membrane to a  platinum electrode operating as an amperometric sensor. [32]oth experimental methods based on Clark probes [33] and theory for the time dependent sensor responses [34] have been studied.In order to reach a stationary state, the amperometric Clark probe usually requires 30 s to 60 s.Surface modified Clark probes may require more time to reach a stationary state due to slower diffusion through the surface layers (e. g. photocatalyst films) and time constant effects from chemical processes.Here, we measure the Clark probe current at close to stationary state after switching on a light source.In recent work also the detection of hydrogen with a Clark probe was suggested simply by changing the applied potential. [35]xperiments were conducted with Pt@g-C 3 N 4 embedded in PIM-1 (5 : 1 weight ratio) immobilised onto the Teflon film of a Clark probe.The Clark probe was employed to measure both oxygen in the solution (at À 0.7 V vs. Ag/AgCl) and hydrogen in the solution (at + 0.6 V vs. Ag/AgCl).For reference, for airsaturated water (with approx.20 % oxygen) a cathodic current of typically À 0.4 μA was detected.For hydrogen saturated water (100 % hydrogen) an anodic current of typically + 0.4 μA was detected (see Figure S4).For these current responses, the permeability of H 2 and O 2 in Teflon-like materials can be compared [36] and a higher response for the more permeable hydrogen would be expected.However, the conditions at the Clark probe sensor electrode are not fully transport/diffusion controlled and current responses are therefore sensitive to the applied bias potential.
With glucose added into the solution and the light source turned on (at t = 50 s in Figure 3A), a transient response is observed and then a new steady-state current reading is detected after about 4 minutes (typical Clark probe time constants are somewhat shorter, but here a more complex steady state involving the photocatalyst film needs to be established).Figure 3B shows average current data plotted (error bars based on triplicate measurements) as a function of the logarithm of glucose concentration.
The transient signals in Figure 3A for 0.1 M glucose suggest that the oxygen concentration immediately plummets when the light is switched on.This is consistent with oxygen being readily reduced/consumed at illuminated Pt@g-C 3 N 4 .The oxygen removal is followed by the production of hydrogen after a short delay.When performing this experiment under an atmosphere of argon, the competing oxygen reduction can be eliminated.The process becomes more effective and, perhaps surprisingly, the Clark probe current becomes significantly higher compared to that expected for hydrogen saturated aqueous solution (0.7 μA instead of 0.4 μA).This observation can be assigned here (i) to the formation of hydrogen directly at the surface of the Teflon membrane and (ii) to the presence of the PIM-1 causing hydrogen to be directed into the Teflon membrane of the Clark probe rather than into the electrolyte solution (compare results obtained with PIM-1 photocatalyst films on platinum electrodes [26] ).Sensor current fluctuations (see error bars) are likely to be associated with uneven photocatalyst deposition and with hydrogen bubble formation affecting the process.Plots in Figure 3B are therefore based on the average from three repeat measurements.
Figure 3B shows that in the presence of 10 mM glucose in water already a significant effect on the oxygen signal is observed.The lower reading of the oxygen signal can be attributed to the consumption of oxygen during the photochemical reaction.The sequence of reaction steps in simplified form can be expressed as in equations 1 to 5.
After excitation and charge separation, the quencher can react with the hole on the g-C 3 N 4 surface leaving the electron on the Pt nanoparticles.The metal and the quencher function as electron acceptor and electron donor respectively to promote the charge carrier separation during the photochemical process.Due to the presence of Pt nanoparticles, oxygen reduction directly to water (rather than H 2 O 2 ) seems most likely thermodynamically (equation 4).Only when sufficiently low levels of oxygen are reached, the production of hydrogen can commence (equation 5).After hydrogen is generated locally at the photocatalytic sites, possible hydrogen diffusion paths are (i) towards the Teflon membrane to result in a sensor response or (ii) towards the electrolyte solution.
In the data in Figure 3B, it can be seen that the production of hydrogen increases with higher concentrations of glucose.Here, 10 mM, 20 mM, 50 mM, 80 mM, 100 mM, 200 mM, and 500 mM glucose in 0.1 M phosphate buffer solution at pH 7 were employed.This concentration effect has been observed previously in related experiments with Pt@g-C 3 N 4 /PIM-1 photocatalyst films on platinum electrodes [26] and could be at least in part related to the adsorption of the quencher onto the photocatalyst surface or also related in part to the interaction of glucose with the microporous PIM-1 host.
When using argon deaerated solutions of glucose, the observed Clark probe currents increase significantly.In fact, these currents are higher than those expected for 1 bar hydrogen in solution (I = + 0.4 μA).This can be explained with hydrogen production close to the Teflon membrane causing preferential permeation through the Teflon and not through the PIM-1 film.The flux of hydrogen is driven photochemically and therefore higher than that expected for the anticipated diffusion limited flux.This is consistent with previous observations of PIM-1 films increasing photocurrents due to localised hydrogen production. [26]In aerated solutions, the Clark probe current remains lower probably due to a constant flow of oxygen from the solution towards the photocatalyst.

Monitoring Photoelectrochemical Hydrogen Generation in the Presence of Small and Large Molecular Quenchers: Pore-Size Effects
It has been demonstrated recently that Pt@g-C 3 N 4 is reactive towards a wide range of quenchers including non-reducing (e. g. sucrose) and reducing saccharides (e. g. fructose). [27]In order to better understand the role of the molecular quencher, here a comparison of five potential quenchers is investigated: sorbitol, gluconic acid, glucose, sucrose, and raffinose.Experiments are performed with Pt@g-C 3 N 4 embedded into PIM-1 films on the Teflon film of a Clark probe.Aqueous quencher solutions are aerated in all experiments (see Figure 4).
Data in Figure 4A for transient responses suggest that the photocatalyst nano-composite is reactive towards all five different types of quenchers.They all produce a steady state photocurrent signal after approximately 120 seconds.Among these quenchers, gluconic acid provided the highest photocurrent signal followed by sorbitol, while the other three show lower photocurrents (or lower rates of hydrogen production).Considering the structure and size of the five quencher molecules, gluconic acid and sorbitol are "flexible" chain molecules rather than more "stiff" ring structures.Smaller and more flexible quencher molecules are more likely to permeate into the microporous PIM-1 structure and participate in the photocatalytic process, thereby contributing to the enhanced photocurrents.As a secondary predictor for hydrogen generation one could consider the "hydrogen content" of each quencher molecule (see data in Table 1).However, the trends in hydrogen production here are more clearly linked to size-effects on mobility/transport in the microporous host of the photocatalyst.The nature of products generated in these photochemical reactions is currently not known.
In order to further investigate the influence of quenchers on the photochemical hydrogen production, experiments with five saccharides in a wider concentration range were carried out.Figure 4B shows that for all types of quenchers an increase in hydrogen production occurs with higher quencher concentration.Especially for the larger quencher molecules like raffinose it seems plausible that transport through micropores becomes rate limiting.PIM-1 is a highly permeable membrane able to provide good selectivity due to its well-defined microporous channels. [37]Molecules diffuse into the PIM-1 micropores with size-sieving apparently leading to different mobilities of quencher molecules.The pore size distribution for PIM-1 has been critically assessed by Kupgan and coworkers [38] suggesting the presence of micropores typically in the range from 1 Å to 10 Å.
Figure 5 summarises the photocurrent responses from different quenchers (at 0.5 M concentration) comparing with their size information (an average diameter estimated from cell volume; see Table 1) with the known PIM-1 pore size distribution. [38]The micropores in PIMs arise from their unique rigid and contorted structure, producing interconnected free volumes elements with channels that are typically 1 nm in diameter.Literature data suggest that a large proportion of PIM-1 micropore sizes are in the range of 0.6-0.8nm [38] to allow facile transport of smaller molecular species. [39]

Monitoring Photoelectrochemical Hydrogen Generation in the Presence of Hydrophobic Molecular Quenchers: Partitioning Effects
Polymers of intrinsic microporosity have been reported to show host-guest partitioning effects favoring hydrophobic molecules

Figure 5.
(A) Pore size distribution (smooth shift) for PIM-1. [38](B) Effect of quencher molecule size on the corresponding photocurrent responses in comparison with PIM-1 pore size distribution.The quencher concentration for each solution was 0.5 M. in heterogeneous electrocatalytic processes, thus controlling catalyst reactivity and selectivity. [46]To investigate the partitioning (accumulation) effect in the presence of hydrophobic quenchers, additional experiments are conducted with 1hexanol as the quencher during the photocatalytic process.Figure 6A shows chronoamperometry data for 1-hexanol in the electrolyte solution with concentrations ranging from 10 μM to 50 mM.In the presence of 1 mM and less hexanol, data suggest only very minor quenching effects and very similar equilibrium current Clark probe readings essentially close to zero (no hydrogen production).When the concentration of hexanol is increased to 10 mM, a current of 0.25 μA was achieved.This is higher when compared to all previously tested quencher systems at the same concentration (see Figure 4B).This result implies that a relatively low hydrophobic quencher concentration in solution can give a correspondingly higher photoresponse or a higher rate of hydrogen production.Hexanol therefore appears to preferentially accumulate into the PIM-1 pores to result in a higher activity at the photocatalytic sites.Unfortunately, in this case the photocurrent response is not further increasing at higher hexanol concentrations possibly due to flooding of micropores.The data are preliminary and the nature of photo-products from this hexanol photooxidation is currently unknown, although the application of alcohols in photocatalytic hydrogen production has been reviewed. [47]

Conclusion and Outlook
It has been shown that PIM-1 as a microporous host material for Pt@g-C 3 N 4 photocatalyst offers both (i) the facile mechanically robust immobilisation of photocatalyst into a molecularly rigid environment without loss of activity and (ii) an element of control over the photochemical reaction due to pore size effects and hydrophobicity effects.
The novelty and importance of this work is in the heterogenization of the photocatalyst into a microporous host environment.In this way photocatalyst can potentially be employed and recovered/reused for several processes.The microporous host environment can be employed to tune and control the photocatalytic process.Hydrogen is generated by photocatalysis and guided through a Teflon film.
A variety of quencher molecules were employed in the photocatalytic generation of hydrogen.Both, oxygen reduction and hydrogen production were monitored based on the photocurrent response from a Clark probe.Factors that affect the photocatalytic conversion efficiency were identified such as (i) size of the quencher molecule to fit the PIM-1 micropores, (ii) flexibility/rigidity of the quencher molecule impacting on transport in PIM-1 micropores, (iii) adsorption of the quencher molecules onto the photocatalyst, and (iv) pore hydrophobicity and partitioning of quencher molecules into the PIM-1 film.In this context, it is interesting to note that transport of highly flexible but long chain lipids and oils have been reported to permeate through microporous PIMs also in catalytic processes. [48] potential drawback of the in situ hydrogen monitoring technique is the fact that the Clark sensor is designed for oxygen sensing.Hydrogen detection is feasible in theory, but in experiments a considerable error bar is inevitable.High hydrogen flux can lead to bubble formation in the sensor and may lead to sensor inaccuracy.This implies that a thin Teflon membrane could in future be effective in separating hydrogen for further applications in energy conversion.More work will be necessary to make further improvements and to explore other types of membranes.Being able to measure both oxygen consumption and hydrogen production with a single probe is very useful in the assessment of photoelectrochemical processes and for comparing different types of quenchers and reaction conditions.Based on this methodology, new processes will be developed for hydrophobic biomolecules and biowastes and for utilisation in artificial photocatalytic energy conversion reactions.

Figure 1 .
Figure1.Schematic drawing of a Clark probe with Pt@g-C 3 N 4 deposit embedded in PIM-1 on a gas permeable poly-tetrafluoroethylene (PTFE) membrane cap.The applied potential can be adjusted to either detect oxygen (at À 0.7 V vs. Ag/AgCl) or hydrogen (at + 0.6 V vs. Ag/AgCl).The solution in the photocatalytic compartment is aqueous 0.1 M phosphate buffer solution pH 7 with added quencher.

Figure 2 .
Figure 2. SEM images (A lower and B higher magnification) of Pt@g-C 3 N 4 embedded in PIM-1 film coated on a polytetrafluoroethylene (PTFE) membrane.The approximate position of the Pt disk electrode under the PTFE membrane is indicated.

Figure 3 .
Figure 3. (A) Chronoamperometry data detecting hydrogen (at 0.6 V vs. Ag/AgCl) and oxygen (at À 0.7 V vs. Ag/AgCl) with 60 μg Pt@g-C 3 N 4 embedded in 12 μg PIM-1 on PTFE membrane in 0.1 M glucose pH 7 phosphate buffer solution.At the time t = 50 s the LED light (λ = 385 nm) is switched on.(B) Photocurrent response data showing the Clark probe hydrogen currents plotted versus logarithmic glucose concentration; error bars based on one standard deviation for triplicate measurements.

Figure 4 .
Figure 4. (A) Chronoamperometry data for 60 μg Pt@g-C 3 N 4 embedded in 12 μg PIM-1 on PTFE membrane in five different types of quencher solution (0.1 M quencher in 0.1 M phosphate buffer pH 7).At the time t = 50 s the LED light (λ = 385 nm) is switched on.(B) Average Clark probe photocurrent response data for hydrogen plotted versus logarithmic concentration of five different quenchers (error bars based on one standard deviation with triplicate measurements).

Figure 6 .
Figure 6.(A) Chronoamperometry data with 60 μg Pt@g-C 3 N 4 embedded into 12 μg PIM-1 on PTFE membrane in 0.1 M phosphate buffer pH 7 solution with different concentrations of 1-hexanol quencher.At the time t = 50 s LED light (λ = 385 nm) is switched on.(B) Photocurrent response data showing the currents plotted against the logarithmic hexanol concentration (error bars based on one standard deviation with triplicate measurements).

Table 1 .
Summary of Clark probe photocurrent data and predictor parameters based on molecule volume (obtained from the crystallographic unit cell volume per molecule), molecular diameter (assuming a sphere), and stored H 2 equivalents (calculated for stoichiometric conversion in H 2 O to give CO 2 and H 2 ).