Printable Newtonian fluid photocatalysts for scale-up solar CO₂ conversion

Abstract

Photocatalysts are usually in powder form, which makes it difficult to expand the scale of photocatalysis. Here, we report a Newtonian fluid photocatalyst, which consists of an internal nano-hollow imidazole framework and an external light-excitable liquid chain striking a pose on the stage. Due to the intermolecular interaction at the solid-liquid interface and significant steric hindrance effect, the Newtonian fluid catalyst with higher surface tension can firmly adhere to different kinds of scaffolds via simple printing, such as curved surfaces, inclined walls, and grids. In addition to the easier scale-up, the pore structure of frameworks favors faster CO₂ mass transfer, and the liquid chain with a co-catalytic effect serves as the electron donor for efficient CO₂ photoreduction. In this work, the Newtonian fluid photocatalyst achieves a 57.8-fold increase in CO overflow efficiency (100% selectivity), and this universal synthesis method can convert common organic/inorganic photocatalysts into similar Newtonian fluid photocatalysts.

The Newtonian fluid photocatalyst composed of imidazole framework and light-excitable liquid chain can be firmly attached to different types of carriers through printing for easier photocatalytic scaling up and more stable CO₂ photoreduction.

Introduction

Photocatalysis is a green and sustainable manner for converting solar energy into high-value chemicals, which has the potential to become an important tool for addressing climate change and environmental crises^1–4. In the past few decades, various semiconductor photocatalysts have been developed, including nanotubes⁵, nanosheets⁶, amorphous nanoparticles⁷, alloys⁸, and so on^4,9, dedicated to improving the intrinsic catalytic activity during the reaction process. Although significant progress has been made in the performance of photocatalysts and the exploration of reaction mechanisms, the demand for suitable large-scale reactions has not been well explored. The amplification of photocatalytic systems needs to consider the practicality of catalysts beyond their performance¹⁰. At present, immobilized photocatalyst reactors are the main form for achieving large-scale photocatalysis, where photocatalysts are fixed to a matrix to prevent loss of catalysts in outdoor reaction environments^11,12. For example, Wang et al. prepared a porous hydrogel nanocomposite photocatalysis platform with an area of 1m² by embedding a large number of photocatalysts into the hydrogel macroporous matrix¹³; Zhu et al. prepared a 1600 cm² photocatalytic hydrogen production platform by immobilizing powder photocatalysts on a super-hydrophilic wood fiber matrix using gradient alignment theory¹⁴. Although these scaled-up methods have broad applicability in accommodating various types of nanoscale photocatalytic fillers, the matrix of the encapsulated photocatalyst interrupts the active material conduction chain between the catalysts and the effective charge transfer between the catalyst and reactants (such as small gas molecules) at the catalytic interface. Therefore, it is required to design a photocatalyst that can achieve simple amplification without the need for a large matrix.

Newtonian liquid is an ideal viscous liquid proposed by Newton in 1687, whose viscosity remains constant regardless of the shear rate. This characteristic makes it capable of  becoming a stable coating material that can be easily coated on flat surfaces. Compared with embedding a large amount of catalyst into the matrix, the method of expanding the photocatalytic reaction scale by simply coating and brushing to increase the reaction area can not only improve performance, but also eliminate the cumbersome “matrix-catalyst” binding relationship and additional pre-operation costs. The stable and uniform liquid catalyst formed by grafting liquid ends on the surface of microporous nanoparticles has been proven to have promising prospects in the field of photocatalysis^15–17. However, these liquid catalysts lack viscosity due to weak surface tension, making it difficult to stably coat them on steep slopes, grilles, or in strong wind environments. It is considered that the intramolecular interaction of the solvent or the solvent and solute causes the liquid to generate resistance in the flow, which increases the surface tension of the liquid and thus produces viscosity¹⁸. Inspired by this, the liquid chain with a large steric hindrance is spatially confined around its side chains, and then grafted with microporous nano-semiconductors through weak molecular forces, such as electrostatic interaction, to form a strong viscous Newtonian fluid that responds to non-horizontal or open natural conditions^19,20. Unfortunately, since the liquid chain cannot be photoexcited, the internal catalyst surface radiation is attenuated, giving fluidity while sacrificing photocatalytic activity. Therefore, a rational design of high-performance and strong surface tension liquid catalysts would hold further interest.

In this study, we propose a viscous Newtonian fluid catalyst composed of internal microporous nano-hollow spheres with positive potential and external negative potential assisted catalytic liquid chains for simple scaling up of photocatalysis. Due to the electrostatic interactions between the inner and outer components and the significant steric hindrance effect of the liquid chains, the Newtonian fluid catalyst exhibits robust flow resistance, making it easy to apply and stabilize loads in strong winds and non-horizontal environments such as grids, slopes, and even downward orientations. On the other hand, the liquid chain that can be light-excited is designed as an electron donor to improve the electron enrichment degree of internal nano-hollow spheres, thereby promoting reactant adsorption and rapid light conversion. Theoretical and experimental verification demonstrates the practical advantages of Newtonian fluid catalysts in photocatalysis.

Results and Disscussion

The structure of Newtonian fluid photocatalysts

To achieve the viscous and flowing effect of Newtonian fluids while maintaining photocatalytic performance, we divide the catalyst into two parts: an inner and an outer part. The microporous nano-hollow spheres, serving as the inner part, act as the primary catalytic contributor. In this regard, the polymerization of imidazolium (PIL), which possess strong positive potential and electron-enriched site, can serve as the adsorption and conversion sites of reactants. The outer parts are long polymer liquid chains (EY[M]) grafted by Eosin Y (EY) and Polyether amine (M2070), which can ionize anions EY[M]^∙- for combining with PIL from electrostatic attraction and act as electron donors. Schematic (Fig. 1a) and optical images show the main preparation process. Firstly, the PIL was synthesized by the polymerization of divinylbenzene and 1-vinylimidazole on the surface of the PS sphere as templates (Fig. S1). Then, by etching templates, PIL forms the hollow states and micropore surface. The optical images show that the PIL is a white powder. The EY[M] is formed by the dehydration condensation reaction (Fig. S2). EY[M] appears as a dark red semi-translucent liquid. Finally, PIL and EY[M] are self-assembled by electrostatic action to form a viscous fluid catalyst (PIL-EY[M]). The molecular structure is shown in Fig. 1a. The fluid rheology experimental results of PIL-EY[M] show that the shear rate of PIL-EY[M] is proportional to the shear stress (Fig. 1b), exhibiting a straight line passing through the origin, which is consistent with the Newtonian fluid equation. At the same time, as the shear rate increases, the viscosity remains constant (Fig. S3). In addition, viscosity-time curve tests were conducted over a wide range, with four constant shear rates (1, 10, 100, 1000 S^-1) selected for testing. The results showed that the viscosity of PIL-EY[M] remained stable under all four conditions (Fig. S3-S4), with no downward trend, which directly ruled out the possibility of it being a thixotropic fluid. Therefore, the viscosity of PIL-EY[M] belongs to a Newtonian fluid¹⁹.

Fig. 1. Characterization of Newtonian fluid photocatalyst structure.

a Schematic and optical images of the synthesis process. The dotted lines and dotted box are used to explain the internal structure of the PIL-EY[M]. b Rheological experiment of PIL-EY[M]. c High-resolution TEM images of PIL. The dotted lines and arrows are used for intuitive measurement and display of length. d Solid-state ¹³C NMR of PIL. e FTIR spectroscopy of PIL, EY and PIL-EY[M]. The dotted line and dotted box are used to indicate the position of the peak.

Subsequently, a series of characterizations were conducted to demonstrate the structure of PIL-EY[M]. The morphology of PIL is spherical with a hollow diameter of around 370 nm inside (Fig. 1c, Fig. S5), and there are positive potential (+32.6 mV) multiple microporous structures on its surface (Fig. S6, Tab. S1). These hollow and microporous structure are used to increase reactant mass transfer inside the Newtonian fluid photocatalyst. The ¹³C Solid State Nuclear Magnetic Resonance (SSNMR) was used to demonstrate the composition of PIL (Fig. 1d). The peaks at 138 ppm and 128 ppm belong to the imidazole carbon²¹, the peaks around 145 ppm and 113 ppm are attributable to the phenyl carbon²², and the other peaks belong to the carbon chain. Two broad diffraction peaks at 13.3° and 27.5° were observed in the X-ray diffraction (XRD) pattern of PIL (Fig. S7), attributed to the planar stacking of imidazole rings and the interlayer stacking of aromatic rings, respectively. The PIL-EY[M] did not detect the above two peaks, and their signals were completely covered by the EY[M], showing a peak located at 21.6°, which belongs to the amorphous diffraction of M2070²³. The X-ray photoelectron spectroscopy (XPS) and Fourier transform infrared (FTIR) spectroscopy further confirmed the detailed information of the PIL (Fig. 1e and Fig. S8). In the TEM image of EY[M], small negatively charged (−32.3 mV) nanoclusters (<5 nm) can be observed (Fig. S9), which are formed due to the monomers and dimer-aggregates of EY²⁴ (Fig. S10). These nanoclusters serve as the “head” for binding with PIL. In the FTIR spectrum of EY[M], there is an added peak at 1670 cm^-1 (Fig. 1e), which represents the -NH- groups formed by the dehydration condensation of EY and M2070. The PIL-EY[M] still maintains the nanosphere morphology. TEM and mapping images indicate that EY[M] is wrapped around the outer surface of PIL (Fig. S11-S12).

Characterization of Newtonian fluid photocatalysts

The stable fluidity and viscosity of the catalyst at room temperature are expected to generate unique features that are advantageous for the scale-up of photocatalysis. To achieve a coating effect, a viscous and slowly flowing Newtonian fluid photocatalyst can be controlled by adjusting the synthesis ratio of the inner and outer parts. In the optical images (Fig. 2a), it can be observed that when flipping the bottle containing PIL-EY[M], the PIL-EY[M] flows slowly downwards under the influence of gravity. By pressing the PIL-EY[M] in the syringe, it can drip (Fig. 2b). By dipping the needle into the PIL-EY[M] and applying an upward force, the PIL-EY[M] adheres to the needle and exhibits strong upward flow resistance (Supplementary Movie 1). These indicate that the PIL-EY[M] takes into account both viscosity and fluidity at room temperature and can be coated onto a carrier using a brush, greatly improving efficiency compared to the matrix preparation and catalyst embedding processes in conventional photocatalytic amplification platforms. There properties make it suitable for brush coating on various carriers (such as plastic, wood, metal, et al.) and even grids (Fig. 2c). In contrast, the powder photocatalysts are limited to planar carriers in sealed conditions, as powder are difficult to fix under non-horizontal plane environments such as inclined plain or facing downwards, and are also prone to mass loss due to being blown away in open air. In a simulated open environment with a wind speed of 18 m/s (equivalent to an 8-level typhoon) for 20 min (Fig. 2d), the PIL-EY[M] has no significant photocatalyst mass loss (Fig. S13), and it can maintain stability in non-horizontal planes or even vertically downward states for 180 days (Fig. 2e). Despite all this, strong viscosity does not mean difficult to recycle, the PIL-EY[M] can be removed from the carrier using organic solvents such as N,N-Dimethylformamide and ethanol, and can be recovered by washing with these solvents (Fig. 2f). The thermal decomposition temperature and melting temperature of PIL-EY[M] are 235 °C and −40 °C (Fig. 2g, Fig. S14), respectively, which indicate that it can remain the fluidity and viscosity for most climate environments on Earth.

Fig. 2. Stable coating properties of Newtonian fluid photocatalyst.

a Fluidity of Newtonian fluid photocatalysts. b Viscosity of Newtonian fluid photocatalysts. c Printing coating on various carriers. d Newtonian fluids facing strong wind conditions. The arrow indicates the direction of the wind. e Newtonian fluid photocatalysts applied face down. f Easily erasable Newtonian fluid photocatalysts. g Melting point of Newtonian fluid photocatalysts. h The RDG analysis of PIL and EY[M]. The circle represents weak interaction forces. i The IGM analysis of PIL and EY[M]. The circle represents weak interaction forces. j Simulated size of EY[M].

Weak intermolecular forces, such as electrostatic interactions, can cause the flow resistance, which is one of the reasons for the PIL-EY[M] viscosity. Reduced Density Gradient (RDG) and Independent Gradient Model (IGM) analysis were used to evaluate the weak intermolecular forces between PIL and EY[M] (Fig. S15, Supplementary Data 1). The RDG analysis results confirmed the existence of a prominent weak interaction force (blue iso surface) between PIL and EY[M], which can be reasonably inferred as an electrostatic force between them (Fig. 2h). As shown in Fig. S16, the black and red dots are defined as intramolecular and intermolecular forces. The IGM analysis results clearly indicate the intermolecular forces between PIL and EY[M]. After eliminating the interference of atomic density gradients, IGM analysis visualizes the real binding environment. The results reveal an ideal intermolecular interaction between PIL and EY[M] (Fig. 2i), where the blue surface is attributed to electrostatic binding force, while the green surface reflects the spatial potential resistance and dispersion effect with low electron density²⁵. It is worth noting that the length and width of EY[M] reach 22.97 Å and 21.22 Å, respectively (Fig. 2j, Supplementary Data 2). Such a large-sized EY[M] cannot pass through the micropores on the PIL surface or enter the hollow structure. The micropores of PIL are well preserved for enhancing the adsorption and mass transfer of reactants in photocatalytic reactions.

The photocatalytic advantages and mechanism of Newtonian fluid photocatalysis

The photocatalytic CO₂ conversion reaction is one of the most common reactions in the field of photocatalysis, and the unique advantages of Newtonian fluid photocatalysts in practical applications are analyzed through CO₂ reduction reactions. The detailed description and light source information of the photocatalytic CO₂ reduction measurement experiment are provided in Fig. S17. The performance testing was conducted under gas-solid conditions, where the PIL-EY[M] was coated onto a carrier film, and the PIL is dispersed on the carrier with the same area for detection. Optical images indicate that compared to powder catalysts, Newtonian fluid photocatalysts can be coated more uniformly on the surface of the carrier with less mass (Fig. S18), which greatly saves catalyst usage and maximizes the reaction contact area between the catalytic surface and CO₂. After five hours of simulated light reaction, CO was the only product in all samples. The results showed that the CO yield of PIL, EY, and EY[M] were 21.24 μmol h^-1 m^-2, 87.82 μmol h^-1 m^-2 and 80.83 μmol h^-1 m^-2, respectively (Fig. 3a, Fig. S19). In contrast, the CO evolution rates of PIL-EY[M] reached 1228.68 μmol h^-1 m^-2, which was 57.84, 13.99 and 15.20 times of PIL, EY and EY[M], respectively. The long-term reaction test of the photocatalysts showed that the yield of PIL and EY gradually decreased after stable for 10 hours during the continuous reaction of 25 hours (Fig. 3b). The CO yield of PIL-EY[M] remained stable, with an average yield of 1206.28 μmol h^-1 m^-2 over 25 hours, which was 58.61 times and 58.99 times higher than PIL and EY, respectively. The AQY efficiency of catalyst PIL-EY[M] reaches 0.43%. In addition, the turnover frequency (TOF) of PIL-EY[M] for CO was 2.28 × 10^-2h^-1, which was significantly higher than that of PIL (6.79 × 10^-6 h^-1). Compared with the same type of imidazole photocatalysts, ion composite photocatalysts, and different types of photocatalysts (Tab. S2-S3), PIL-EY[M] has competitive performance increased multiplier factors, emphasizing the advantages of Newtonian fluid photocatalysts in photocatalytic systems (Fig. S20). The FT-IR, XPS and NMR characterization confirmed that there were no functional or structural changes observed after the PIL-EY[M] reaction (Fig. S21-S24). Subsequently, the isotopic ¹³CO₂ experiment and a series of comparative experiments were conducted to demonstrate the photocatalytic activity of the product derived from Newtonian fluids (Fig. S25, Fig. S26). Subsequently, we tested the CO performance of PIL-EY[M] coated on various carriers (Fig. S27).

Fig. 3. CO₂ reduction performance and mechanism of Newtonian fluid photocatalyst.

a Photocatalytic CO₂ reduction production of CO. Inset: PIL-EY[M] coated on the reaction membrane. The arrow represents the trend of performance improvement. b 25 h CO₂ reduction production of CO. c UV-vis spectra before and after reaction of PIL-EY[M]. Inset: UV-vis spectra before and after reaction of EY. The arrow represents a downward trend. d The TDDFT calculations at the lowest energy excited state and the differential charge density of PIL-EY[M]. The arrow represents the distance between the centers of mass of electron and hole. e The mono-exponential fit of PIL-EY[M] dynamics at 440 nm. Data in panels a and b are presented as the mean ± standard deviation (s.d.) from three independent experiments.

Explain the advantages of Newtonian fluids in photocatalytic reactions from a mechanistic perspective. The UV absorption spectrum of PIL-EY[M] showed stronger absorption than EY[M], which may be due to the charge transfer (CT) state between PIL and EY[M] (Fig. 3c). As the reaction time increases, the peak at 516 nm and 480 nm strengthens, indicating an enhanced π/π* transition in PIL-EY[M]. In contrast, the absorption spectrum of EY decreases significantly with reaction time. By performing time-dependent density functional theory (TDDFT) calculations on the optimized structure of PIL-EY[M], it was found that the distance between the centers of mass of holes and electrons was as high as 3.69 Å at S0-S1 (Fig. 3d, Fig. S28, Supplementary Data 3), corresponding to the contribution of charge-transfer excitation as the lowest energy excited state of PIL-EY[M], indicating a high degree of separation between electrons and holes²⁶. It can be reasonably inferred that it belongs to π/π* charge transfer excitation in the EY to PIL direction. The charge difference calculation of PIL-EY[M] proves that the excited electrons gather on EY and are consumed on PIL, indicating that EY[M] can serve as an electron donor and PIL as an acceptor. This result is consistent with the HOMO, LUMO orbital positions and band structure of PIL and EY[M] (Fig. S29-S30). A larger dipole moment of the excited state is more conducive to the charge transfer from the lowest excited state of EY[M] to PIL. In addition, PIL-EY[M] has a larger ground state dipole moment (19.34 D) compared to PIL (1.63 D) and EY[M] (8.32 D), indicating that PIL-EY[M] is more conducive to photogenerated charge separation²⁷, which is consistent with the results of weaker emission intensity, better electrochemical properties, and charge transfer density (Fig. S31-S32, Tab. S4). However, DFT has certain limitations in simulating real catalytic conditions. To gain a deeper understanding of the excited-state dynamics between EY[M] and PIL. Immediately, femtosecond transient absorption spectroscopy (TAS) was conducted on PIL-EY[M] after excitation by a 420 nm laser pulse, the spectrum displayed a strong broad bleach centered at 538 nm, attributed to the singlet excited state absorption of EY[M] (Fig. S33). The 420−490 nm range exhibited the characteristic absorption of the EY radical anion in ET[M], reaching a maximum at 440 nm. The mono-exponential fit of PIL-EY[M] dynamics at 440 nm resulted in a lifetime of only 857.3 ps (Fig. 3e), indicating that exciton energy has been transferred from EY[M] radical anions to PIL, directly demonstrating effective electron transfer between EY[M] and PIL^28,29.

The availability of reactants and stable adsorption at catalyst sites are key steps in the photocatalytic reduction process. The weak adsorption capacity of CO₂ without porosity limits the development of liquid catalysts. The presence of permanent hollow pores in porous Newtonian fluid photocatalysts is expected to facilitate reactants diffusion and transfer without resistance. PIL-EY[M] exhibits a 3 times higher adsorption rate of CO₂ compared to EY[M] (Fig. S34), improving the mass transfer of the CO₂ reduction reaction. In addition, we monitored the color change of the PIL-EY[M] in a solution containing alkaline phenolphthalein exposed to the CO₂ environment. The color of phenolphthalein fades due to the consumption of hydroxide ions (OH^-) dissolved in CO₂ in the presence of carbonic acid. Therefore, the length of fading time can be used to reflect the diffusion rate of CO₂ in the catalyst solutions³⁰. It is worth noting that the micropores of PIL are extremely hydrophobic (Fig. S35), which allows their pores to be well preserved in aqueous solutions. To prevent color overlap of the orange PIL-EY[M] and EY[M] aqueous solution, the thymolphthalein, which appears blue under alkaline conditions, is replaced by the red phenolphthalein (Fig. S36). At the same CO₂ flow rate, the color change of PIL-EY[M] is more pronounced than that of EY[M] and PIL (Fig. 4a). The complete faded solution in PIL, EY[M] and PIL-EY[M] takes about 126 s, 93 s and 57 s, respectively. This indicates that the permanent porosity in the Newtonian fluid photocatalyst makes gas mass transfer no longer the limiting step of liquid catalysts. However, due to the easy aggregation of powdered catalysts, the contact area between CO₂ and PIL is limited, leading to a hard diffusion in the aqueous phase with the slowest mass transfer time. To decouple the benefits of the “liquid state” properties of Newtonian fluids from the chemical function of amino groups in M2070, the 3,4’-Oxydianiline (3,4’-ODA) and Polyacrylamide (PAM) were chosen as substitutes for M2070 (Fig. S37). They have rich amino groups similar to M2070. The difference is that both have rigid frameworks of phenyl/alkane, making them non-liquid materials. We mixed these two materials with PIL and EY by deposition method, and after mixing, formed the final materials PIL-EY[ODA] and PIL-EY[PAM].The fading time of PIL-EY[ODA] and PIL-EY[PAM] are both above 120 s, similar to the fading time of PIL, and lower than that of PIL-EY[M] (Fig. S38).

Fig. 4. Newtonian fluid photocatalyst promotes CO₂ adsorption and mass transfer.

a CO₂ mass transfer experiments of PIL-EY[M], EY[M] and PIL. b Kinetic simulation snapshot at 0 ps. c Kinetic simulation snapshot at 200 ps. d The radial distribution function of PIL-EY[M]. e In-situ infrared spectroscopy of PIL and PIL-EY[M]. Arrow and circle represent the peak positions of line adsorption and bridge adsorption, respectively.

Molecular dynamics simulation can deeply simulate the dynamic environment of Newtonian fluid. The PIL-EY[M] monomer structure (for the polymer chain, we took ten repeating units to simulate the overall situation) and 80 CO₂ molecules were placed in a box to simulate the adsorption and diffusion of CO₂ on the surface of Newtonian fluids (Supplementary Data 4). Figure 4b is a snapshot of the initial structure at 0 ps. As the simulation progresses, the polymer chain of PIL-EY[M] quickly aggregates into clusters and continuously changes its configuration afterwards, reflecting the flow state of Newtonian fluids. At the same time, PIL-EY[M] continuously attracts surrounding CO₂ molecules. The results showed that 96% of CO₂ in the system was adsorbed on various sites of PIL-EY[M] at 200 ps (Fig. 4c). Meanwhile, a two-dimensional density analysis was conducted on the diffusion location of CO₂. As shown in Fig. S39 (0 ns), CO₂ (red dots) is evenly distributed in the box, corresponding to “free diffusion, no obvious preferred position”. After 200 ps, CO₂ appeared with a density distribution in a very obvious high-density area (red cluster). This indicates that the residence time of CO₂ in a certain spatial region is significantly longer, so it is not an instantaneous aggregation caused by random walks of CO₂, but the existence of stable adsorption sites. The radial distribution function (RDF) is used to determine which part of PIL-EY[M] exerts the strongest force on CO₂. The PIL-EY[M] has four atoms: C, N, O, and Br. Among them, Br-C and Br-O peaks on PIL-EY[M] have the highest intensity (Fig. 4d), indicating its strongest adsorption capacity for CO₂. The peaks of Br and O rise the fastest, indicating that it comes into contact with CO₂ more frequently. The order of adsorption strength of CO₂ by each atom is Br > O > C > N (Fig. S40). CO₂ has a clear quadrupole moment, with a positive C-end and a negative O-end; Therefore, Br (a highly polarized/negatively charged site) attracts the C of CO₂ the most significantly, corresponding to the ultra-high first peak of Br-C. The O site provides secondary polar interactions/directional adsorption. This also indicates that the adsorption and mass transfer of CO₂ by Newtonian fluid catalysts work as a whole at the microscopic level.

Further in-situ infrared spectroscopy was used to detect the adsorption vibration changes of CO₂ intermediates (Fig. 4e). The signal intensity of PIL-EY[M] is significantly stronger than that of PIL. The reaction intermediates in PIL showed a linear adsorption state dominated by *C-O, with a peak at 2263 cm^-1, which may be attributed to the combination of imidazole sites and *CO₂  intermediate. The wide peaks were observed at 2137 cm^-1, 2069 cm^-1, and 1997 cm^-1, which overlapped and belonged to the bridged adsorption of *C-O in PIL imidazole sites (Fig. S41). The bridged adsorption mode is generally considered to be a more stable chemical adsorption than linear adsorption³¹. Bridged adsorption peaks of PIL-EY[M] gradually become dominant compared to PIL during the 0-20 minutes reaction process, which is attributed to the formation of CO₂ intermediates adsorption at the interface between PIL and EY[M]. The DFT theoretical calculation results indicate that CO₂ intermediates adsorbed at the interface between PIL and EY[M] in PIL-EY[M] have more advantageous adsorption energy (Fig. S42). Their advantages are manifested in more electron-enriched sites and lower thermodynamic reaction energy barriers (Fig. S3-S44). Therefore, the interaction between the solid-liquid interface of Newtonian fluid photocatalysts is also crucial for the improved photocatalytic performance.

Large-scale application characteristics of Newtonian fluid photocatalyst

As Newtonian fluid photocatalysts with fluidity and viscosity having application advantages in amplification reactions, the PIL-EY[M] can be uniformly coated on the inner surface of the bottom of a 36 L square glass reactor, and a 0.6 m*0.6 m quartz cover plate was sealed on the top to form a scale-up solar CO₂ reduction device (Fig. 5a). The CO₂ was continuously introduced for two hours to fill the reduction device for outdoor testing, and 10 ml of water was added as the proton donor. The reaction was conducted from 11:00 to 16:00 (Fig. 5b). The CO production rate in five hours was 613.51 μmol/m² on sunny day (Fig. 5c). Comparison of CO evolution performance on day 1, day 4 and day 7 confirms its long-term stability (Fig. S45). Compared to simulated light sources, the light intensity under real sunlight is reduced by a maximum of 39.5 times, but the performance is only reduced by a maximum of 10.1 times. Therefore, the performance of the photocatalyst did not decrease, and the decrease in yield may only be related to the light intensity. Currently, attempts to scale up reaction systems in the field of photocatalysis are still limited by the difficulty of immobilizing powder state, and can only be conducted by laying the devices horizontally^32,33. The variation in solar elevation angle alters the solar radiation intensity, impacting the surface photon absorption rate of the catalysts and consequently leading to the performance decline in horizontal conditions (Fig. S46). The Newtonian fluid photocatalyst can adjust the carrier tilt angle by changing the incident solar elevation angle to increase the production rate without catalyst consumption. In addition, the preparation strategy of Newtonian fluid catalysts can be extended to various photocatalysts. Taking typical powder organic photocatalyst carbon nitride (C₃N₄) and inorganic photocatalyst titanium dioxide (TiO₂) as examples. By coating a Polyimidazolium layer on the surface of TiO₂ and C₃N₄, and then grafting liquid end EY[M] onto the surfaces, Newtonian fluid photocatalysts TiO₂PIL-EY[M] and C₃N₄PIL-EY[M] can be formed in a state similar to PIL-EY[M]. They also exhibit fluidity, uniformity, viscosity, and paintability (Fig. 5d, Fig. S47, Supplementary Movie 2). At the same time, Newtonian fluids TiO₂PIL-EY[M] and C₃N₄PIL-EY[M] also showed improved CO₂ reduction ability (Fig. S48). However, when removing the Polyimidazolium layer or EY from TiO₂PIL-EY[M] and C₃N₄PIL-EY[M], the TiO₂ and C₃N₄ will separate (Fig. 4f, Fig. S47), ultimately preventing the formation of Newtonian fluids. Therefore, the electrostatic interaction between the solid and liquid ends is the key to the formation of this type of Newtonian fluid photocatalyst. The construction strategy of solid-liquid interaction can achieve universal synthesis of various Newtonian fluid photocatalysts.

Fig. 5. The outdoor applications and universal synthesis of Newtonian fluid photocatalysts.

a Schematic of the overview of a 0.6*0.6*0.1 m²-scale CO production facility. The lines and arrows are used for intuitive measurement and display of length. b Optical image of scale-up reaction under sunlight. c Photocatalytic CO₂ reduction production of CO in scale-up reaction under sunlight. d Optical image of Newtonian fluid photocatalysts TiO₂PIL-EY[M]. The dotted box represents the TiO₂ precipitate.

In summary, we have synthesized a flowing viscous Newtonian fluid material, which is grafted from internal hollow spheres and external liquid chains, for more convenient material applications and more efficient photocatalytic conversion. Experimental results show that, compared to powder catalytic systems, Newtonian fluid photocatalysts can be more uniformly and stably painted on non-planar surfaces, which enables the catalyst to cope with the effects of wind speed and solar altitude angle under natural conditions. At the same time, the permanent porosity and the solid-liquid interface of a Newtonian fluid enable better adsorption, mass transfer, and conversion. In the application of photocatalytic reduction of CO₂, the PIL-EY[M] achieved a 100% selective CO overflow efficiency, which is 57.8 times higher than that of the PIL, with a stable performance under long-time and large-scale solar radiation. Viscous Newtonian fluid photocatalysts have shown potential for commercial feasibility by expanding the application scope of photocatalysis. In addition, by taking the most commonly used inorganic and organic photocatalysts TiO₂ and C₃N₄ as examples, Newtonian fluid photocatalysts TiO₂PIL-EY[M] and C₃N₄PIL-EY[M] with similar viscosity and fluidity were synthesized, demonstrating the universal synthesis of this type of Newtonian fluid photocatalysts.

Methods

Materials

Deionized water (DI water) was employed in all experiments. Sodium dodecyl benzene sulfonate (SDBS, A.R. ≥ 99.0%), Potassium persulfate (KPS, A.R. ≥ 99.0%), Styrene (St, A.R. ≥ 99.5%), Divinylbenzene (DVB, A.R. ≥ 55.0%), 1-Vinylimidazole (VIM, A.R. ≥ 99.0%), tetrahydrofuran (THF, A.R. ≥ 99.5%), Polyether amine (M2070), 1,6-Diaminohexane (A.R. ≥ 99.0%), acetic acid (A.R. ≥ 99.0%), formaldehyde (A.R. 36-38 wt% in water), methylglyoxal (A.R. 40% solution in water) were obtained from Sinopharm Chemical Reagent Co., Ltd. INHONG GAS Company supplied carbon dioxide (CO₂, ≥ 99.0%) and nitrogen (N₂, ≥ 99.0%).

Synthesis of polystyrene (PS) microspheres

The PS microspheres were synthesized by emulsion polymerization, using SDBS as an emulsifier and KPS as an anionic initiator. The reaction was conducted under a nitrogen atmosphere, and 200 mL of DI water and 50 mL of ethanol were added into a 500 mL three-necked flask containing 0.5 g of SDBS. After adequately dispersing the SDBS, 12.5 mL of St, 3 mL DVB, and 0.25 g KPS were added into the reactor. Then, the polymerization was conducted in an oil bath at 80 °C, with shaking at 300 rpm, for 12 h. Eventually, the PS microspheres were collected by high-speed centrifugation at 5180 xg, and finally vacuum dried at 50 °C for further use.

Synthesis of PIL

The PIL hollow spheres were obtained by combining the template strategy. Specifically, 0.05 g of the as-prepared PS microspheres was added to 37.5 mL DI water in a flask under stirring. In this state, 1 mL DVB and 2 mL VIM were added to the flask with continuous agitation for 1 h. Then, 6 mL KPS solution (0.5 wt%) was added to this system at 75 °C. After keeping stirring for 24 h at this temperature, the precursor of PIL spheres was obtained. Then, the PIL hollow spheres were obtained by etching the PS core with THF and separating by high-speed centrifugation at 5180 x g.

Synthesis of PIL-EY[M]

PIL-EY[M] were prepared by electrostatic self-assembly of PIL hollow spheres and EY[M]. The PIL hollow spheres (0.1 g) were dispersed in 15 mL DI water, a specified amount of EY was added to the PIL hollow spheres suspension. After stirring for 1 h at 60 °C, M2070 in the same amount as EY was slowly added to the mixture solution, stirred at 70 °C until it turned into a red viscous liquid.

Synthesis of EY[M]

A specified amount of EY was added to the PIL hollow spheres suspension. After stirring for 1 h at 60 °C, M2070 in the same amount as EY was slowly added to the mixture solution, stirred at 70 °C until it turned into a red viscous liquid.

Synthesis of TiO₂ and C₃N₄ covered with the imidazole cation layer

0.5g TiO₂ or C₃N₄, DI water (10 mL),1,6-Diaminohexane (0.5 mL), acetic acid (2 mL) were sequentially added to a round bottom flask and sonicated for 0.5 hours. After that, formaldehyde (0.71 mL) and methylglyoxal (1.47 mL) were added and stirred at room temperature for 24 hours. The final product is washed with water and ethanol.

Synthesis of TiO₂PIL-EY[M] and C₃N₄PIL-EY[M]

Similar to the PIL-EY[M] synthesis process, the difference is that TiO₂ and C₃N₄ covered with the imidazole cation layer were used instead of PIL.

Photocatalysis experiments

The entire photocatalytic CO₂ reduction reaction was carried out in a gas-solid mode in a 250 mL quartz glass device with good translucency. The PIL-EY[M] were painted on a poly (vinylidene fluoride) membrane with a radius of 25 mm while weighing the weight of the PIL-EY[M] load, and the membrane coated with the photocatalyst was put on the platform in the reaction device. The water vapor environment was constructed by the introduction of 1 mL of water. A low-energy 360 nm LED visible light source was 5 cm away from the membrane. The detailed schematic diagram of the device is shown in Fig. S17. Before turning on the light, breathe with high-purity CO₂ gas for 15 min to eliminate other gases, and maintain the reaction temperature at room temperature by flowing water. During light irradiation, the gas products were analyzed by a Tetchrom gas chromatograph (GC2030) with a flame ionization detector (FID) and thermal conductivity detector (TCD). The experimental light source and outdoor sunlight intensity are both measured by the light intensity meter (model: PLD MOPM-I; PerfectLight). Before testing, the probe of the light intensity meter should be directly exposed to light, and the distance between the probe and the lamp source should be consistent with the distance between the lamp source and the catalysts. Wait for the light intensity meter index to stabilize before reading for each detection. After testing, the intensity of the experimental light source used is 220 mW/cm^-2, and the intensity in outdoor experiments are 9.32-58.6 mW/cm^-2 (sunny) and 5.57-38.2 mW/cm^-2 (cloudy). The outdoor light intensity detection time is from AM 11:00 to PM 4:00, with hourly detection.

TA spectroscopy

Time-resolved experiments were carried out on laser-based spectroscopy, with laser powers equating to less than one photon absorption per particle. Samples for transient absorption experiments were kept in dark between each measurement. A Coherent Legend Ti: Sapphire amplifier (800 nm, 100 fs pulse length, 3 kHz repetition rate) was used.

The output is split to pump and probe beams. Excitation pulses at the wavelength of 450 nm were acquired using an optical parametric amplifier (Topas C, Light Conversion). The probe pulses (a broad supercontinuum spectrum) were generated from the 800-nm pulses in a CaF₂ crystal and split by a beam splitter into a probe pulse and a reference pulse. The probe pulse and the reference pulse were dispersed in a spectrograph and detected by a diode array. Instrumental response time is ∼100 fs.

Characterization

Power X-ray diffraction (XRD) was performed by a 9 kW Rigaku Smartlab X-ray diffractometer with Cu Kα radiation. The morphology and element distribution were characterized by high-resolution transmission electron microscopy (FEI TALOS F200X) and scanning electron microscope (Regulus 8100) with energy dispersive spectroscopy. XPS was implemented on an X-ray photoelectron spectrum (Thermo SCIENTIFIC ESCALAB Xi + ), and C 1 s 284.8 eV as reference. The element-selective X-ray absorption fine structure measurements were performed on the Shanghai Synchrotron Radiation Facility, and the data were analyzed by Athena software. The optical absorption properties were measured over a UV-2600 UV-vis DRS spectrophotometer with BaSO₄ powder as a reference. 13C-isotopic tracer was investigated through gas chromatography-mass spectrometry (Shimadzu GC-MSQP2020). Brunauer-Emmett-Teller (BET) specific surface areas were recorded on a Micromeritics TriStar Ⅱ 3020 M instrument. CO₂ adsorption was analyzed by the QUADRASORB evo instrument. The gas products were detected by a gas chromatography-mass spectrometer (GC-MS, Agilent Technologies) in the isotopic ¹³CO₂ experiment.

In-situ FT-IR experiments were recorded on a Thermo Scientific Nicolet iS50 FTIR with drift accessory. First, the material is loaded into the reaction chamber and paved, and the sample position is located below the center of three windows, including a quartz window and two infrared-transparent windows. Then, high-purity nitrogen containing water vapor is run through the reaction tank to keep the signal stable under dark conditions for 60 min. Subsequently, infrared data of CO₂ adsorption on the catalyst began to be recorded under a continuous flow of high-purity CO₂ gas. After the signal is stabilized, turn on the LED light and record the infrared data during the CO₂ photoreduction process.

Photoelectrochemical measurements

The photoelectrochemical properties were measured by using CHI660e electrochemical workstation equipped with a calomel electrode as the reference electrode and the platinum wire as the counter electrode. 5 mmol/L K₃Fe(CN)₆, 5 mmol/L K₄Fe(CN)₆ and 0.1 mol/L KCl were used as impedance electrolyte solution, and 0.5 mol/L Na2SO4 aqueous solution was used as Mott-Schottky and photocurrent electrolyte solution. Pt and Ag/AgCl electrodes were used as counter and reference electrodes, respectively. The saturated calomel electrode, Pt wire, and FTO were used as the reference electrode, counter electrode, and working electrode. The catalyst (1 mg) was added into 1 mL of ethanol, ultrasonic for 30 min, and uniformly added to an area of 1 cm × 1 cm on FTO. The calomel electrode potential is converted into a standard hydrogen electrode potential according to the Nernst equation (1):

$$\mathrm{ENHE}=ESCE+0.059\times \mathbf{pH}+{\mathbf{E}}^{\mathbf{\theta }}\mathbf{SCE}$$

Among them, ENHE represents the potential relative to the standard hydrogen electrode potential, ESCE represents the potential relative to the saturated calomel electrode potential, E^θSCE represents the standard saturated calomel electrode potential (0.245 eV) at 25 °C, and the electrolyte pH = 7.02.

Apparent quantum efficiency calculation

The apparent quantum yield (AQY) was determined under monochromatic LED light irradiation at a certain wavelength (λ = 365 nm). The reported AQY value here is maximum attainable results after varying the amounts of photocatalysts, the light intensities, and the light absorption areas. In this experiment, 100 mg catalyst was dispersed in 50 mL aqueous solution with continuous CO₂ bubbling during the photocatalytic test.

AQY is calculated as follows equation (2):

$$\mathrm{AQY}=\frac{2\times C\times {\mathbf{N}}_{\mathbf{A}}}{S\times P\times t\times \mathbf{\lambda }/\left(\mathbf{h}\times \mathbf{C}\right)}$$

where C is the number of evolved CO molecules, N_A is Avogadro constant (6.022 × 10²³mol^-1), S is the irradiated area (2 cm²); P is light intensity (26.9 mW cm^-2); t is illumination time (s); λ is the wavelength of monochromatic light (nm); h is Planck constant (6.626 × 10^-34J/s); and c is the speed of light in free space (3 × 10⁸m s^-1).

The AQY of PIL-EY[M] for CO was calculated accordingly: [(2 × 1.25 × 10^-6 × 6.02 × 10²³× 6.626 × 10^-34× 3 × 10⁸) / (2 × 26.9 × 10^-3 × 3600 × 360 × 10^-9)] × 100% ≈ 0.429%.

Turnover frequency calculation

The turnover frequency (TOF) calculations are used to further standardize and validate the economic feasibility claims. The TOF was calculated according to the following equation (3):

$${\mathrm{\eta }}_{\mathrm{TOF}}=N_{CO}/\left(t\times N_{catalyst}\right)$$

Where N_CO is the amount of generated CO, t is the reaction time, and N_catalyst is the amount of catalytic sites (this refers to the imidazole active sites of PIL and PIL-EY [M]).

The TOF of PIL for CO was calculated accordingly: (21.24 × 6.25 × 10^-10)/(1 × 0.016 × 0.022/0.18) = 6.79 × 10^-6h^-1.

The TOF of PIL-EY[M] for CO was calculated accordingly: (1200 × 6.25 × 10^-10)/(1 × 0.007 × 3.846% × 0.022/0.18) = 2.28 × 10^-2h^-1.

DFT calculations

The geometry optimization and excited-state calculations were performed by B3LYP-D3BJ exchange-correlation functional with the 6-311 G(d) basis set based on the Gaussian 16 C.01 code. Frequency calculations were performed to ensure that the stability configuration has no imaginary frequency. Independent gradient model (IGM)^34,35, Reduced density gradient (RGD), Frontier molecular orbital, the distribution of electrons and holes in the electron excitation process³⁵, and interfragmentary charge transfer were calculated with Multiwfn 3.8 (dev)³⁶. Molecular dynamics simulation, conducted using CP2K software. All structures and isosurfaces images were visualized by VMD 1.9.3. and VESTA. The single-point energy was calculated by M06-2X/def2-TZVPP level. The adsorption energy (E_ads) of CO₂ (equation (4)).

$${\mathrm{E}}_{\mathrm{ads}}\left({\mathrm{CO}}_2\right)=E\left(*{CO}_2\right)-E\left(*\right)-E\left({CO}_2\right)$$

Where E(*CO₂), E(*) and E(CO₂) were the total energy of samples with CO₂ adsorbed on the surface, the energy of the pristine sample surface and CO₂, respectively.

At room temperature, the Gibbs free energy for each model was calculated by the following equation (equation (5)):

$$\mathrm{G}\left(\mathrm{T}\right)=E_{ele}+G_{corr}\left(T\right)=E_{ele}+ZPE+\mathrm{\Delta }{G}_{0\to T}$$

Where E_ele, G_corr(T), ZPE and $\mathrm{\Delta }{G}_{0\to \mathrm{T}}$ were electronic energy, thermal corrections to Gibbs free energy, zero-point energy and contribution by heating the system from 0 K to 298.15 K, respectively.

Author contributions

Z.L., Y.C. and Y.X. performed the experiments, data analysis and wrote the paper. L.T. and H.L. performed the computer simulations. T.Z., K.Z. and W.S. conceived the ideas and designed the experiments.

Peer review

Peer review information

Nature Communications thanks Bien Tan and the other anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The authors declare that the data that supports the findings of this manuscript can be found in the Supplementary Information and are available free of charge or available from the corresponding author upon request. The source data underlying Figs. 1b,1d-e, 2g, 3a-e, 3g, 4d-e, and 5c are provided in the Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70819-z.