Introduction

Developing environmentally friendly, efficient, and economically viable strategies has long been a priority in both academic and industrial chemistry. Photocatalysis1,2,3,4, particularly visible-light-driven processes, aligns well with these goals by facilitating a broad range of chemical transformations under mild conditions5,6,7. It has significantly enriched the synthetic toolbox with diverse mechanisms such as single electron transfer (SET), energy transfer (EnT), hydrogen atom transfer (HAT), and halogen atom transfer (XAT), among others8,9,10,11,12,13. However, the application of visible light in photocatalysis is constrained by its limited penetration depth and the extensive solvent volumes required for dilution, complicating scalability14,15,16,17,18. Conventional scaling strategies, such as enlarging reactor size, have proven less effective, especially for turbid suspensions where light attenuation reduces reaction efficiency18,19,20. While flow chemistry offers an engineering solution to the problem of light attenuation in large-scale reactions, issues such as the handling of insoluble reagents and high solvent consumption due to highly dilute reactions still persist (Fig. 1a)15,16,17,18. Since solvents constitute about 85% of the total chemical mass in the pharmaceutical industry21, solvent-minimised photocatalysis offers great potential for advancing green and sustainable processes (Fig. 1b)20.

Fig. 1: Photo-mechanochemistry.
figure 1

a Limitations of traditional solution-based photocatalysis. b Waste composition in the pharmaceutical industry. c This work: Photo-mechanochemical platform and its applications.

Full size image

Mechanochemical synthesis, on the other hand, has emerged as a useful tool across materials science, as well as polymer, inorganic, and organic chemistry22,23,24,25,26,27,28. Mechanochemistry can reduce or eliminate solvent use, shorten reaction times, and facilitate the exploration of novel synthetic pathways29,30,31,32,33,34,35,36,37. In 2019, its potential for environmental and methodological improvements led IUPAC to name it among the top ten emerging technologies that will change our world38. The integration of photocatalysis and mechanochemistry builds on these strengths, offering synergistic potential to expand the synthetic landscape. Despite its promise, the emerging field of photo-mechanochemistry remains underexplored, largely due to the challenges of integrating light sources into mechanochemical systems. Early studies by MacGillivray et al.39,40,41,42, König et al.43,44, Hernández45, and Štrukil et al.46 demonstrated the feasibility of photo-mechanochemical reactions, while more recent advances by Rappe and Braunschweig et al.47, Zysman-Colman et al.48, and Borchardt et al. 49 further expanded this field by integrating blue and UV light into ball milling setups. Wu and Wang et al. also demonstrated the use of mechanoluminescent materials as internal photon sources50. Although these findings are encouraging, these methods remain limited by experimental complexity, deficiencies in reaction efficiency and scalability, and reaction versatility51,52. Addressing these barriers requires a versatile and scalable platform to propel photo-mechanochemistry beyond its early stages.

Among various mechanochemical tools, Resonant Acoustic Mixing (RAM) emerges as a promising candidate for photo-mechanochemical applications. Recently introduced into organic synthesis53,54,55,56,57,58,59,60,61, RAM offers two key advantages. First, it eliminates the need for milling media, such as balls or screws, reducing contamination risk and simplifying reactor design. This allows the use of glass reactors, typically impractical or not feasible in ball milling due to the risk of breakage. Second, RAM’s powerful mixing capabilities are particularly suited to solvent-minimised photocatalytic processes, where solid reactants often form paste-like mixtures rather than homogeneous solutions. Such conditions pose challenges to traditional photochemical methods due to non-uniform reactant distribution, uneven irradiation, and light scattering, which reduces light penetration and hinders efficient photon absorption20. We envisioned that by leveraging RAM, photo-mechanochemical processes could overcome these issues by enhancing reactant mixing and increasing reactant exposure frequency.

Herein, we present a new photo-mechanochemical platform that combines visible-light photocatalysis with RAM to enable an efficient, general, and scalable method for metallaphotoredox catalysis under solvent-minimised conditions. As a proof of concept, we select the photoredox dual-catalysed C–X (X = N, O, P, S) bond cross-coupling reactions, which are known for their industrial relevance (Fig. 1c)62,63,64,65,66,67,68,69,70,71,72. We also explore five additional types of photoinduced reactions, further showcasing the versatility of this platform. This strategy features operational simplicity, broad substrate tolerance, reduced reaction times, and scalability with exceptionally low catalyst loading. By addressing key limitations in both photocatalysis and mechanochemistry, this approach broadens the accessible chemical space and opens new opportunities in synthetic chemistry.

Results and discussion

Evaluation of the reaction conditions

Photo-mechanochemical experiments were conducted utilising a commercial Resodyn LabRAM II instrument with a capacity of 1 kg. The Resonant Acoustic Mixer (RAM) employs a non-contact, forced-vibration method that uses a robust acoustic field to facilitate effective particle collisions at low frequencies, commonly around 60 Hz54. The energy delivered during the RAM procedure is controllable through variations in the oscillation amplitude of the sample container, which adjusts its acceleration from 0 to 100 g (gravity, g = 9.81 m·s−2). To facilitate light exposure, a modular lamp holder was designed to accommodate four lamps and two cooling fans. The holder is engineered for easy installation and removal. Additionally, a custom-designed 3D-printed sample holder, made from ABS polymer, was tailored to fit 4 mL glass vials. This design enhances the throughput of photo-mechanochemical reactions, enabling up to 12 reactions to be conducted simultaneously while maintaining excellent reproducibility (Fig. 2 and Supplementary Fig. 1).

Fig. 2: Photo-mechanochemical platform set-up.
figure 2

Left: Twelve 4 mL glass vials and a 3D-printed sample holder positioned at the core of the labRAM II instrument. Middle: A modular lamp holder equipped with four Kessel lamps and two cooling fans for thermal management. Right: External view of the assembled set-up.

Full size image

Our optimisation studies began with conditions previously established by König and co-workers for nickel photoredox catalysis67. The results are summarised in Table 1. In the photo-mechanochemical setup, 4-bromobenzonitrile (1a) and aniline (2a) were used as representative starting materials in a 4 mL glass vial. The reaction mixture comprised 4CzIPN (0.1 mol%), NiBr₂·glyme (5 mol%), DABCO (2.0 equiv.), and Et₃N (0.25 equiv.) in DMA (0.4 mL) and was irradiated with 90 W Aqua Blue LEDs at 90 g for 90 min. Under these conditions, a 90% yield of compound 3 was achieved, as determined by GC-FID using dodecane as the internal standard (entry 1). This positive outcome led us to further explore the reaction’s parameters. Remarkably, omitting the solvent led to only a marginal decrease in yield, though it resulted in the formation of a sticky, heterogeneous, and cloudy mixture (entry 2). The application of grinding agents commonly used in ball milling, such as sea sand and silica gel, resulted in a decrease in yield for 3 to 82% and 43%, respectively (entry 3 and 4). Control experiments showed that light, a photocatalyst, and nickel salt are all essential to the reaction (entries 5–7). In the absence of DABCO or Et3N, or both, the reaction proved to be ineffective (entries 8–10). To our delight, the concurrent use of 1 equiv. each of DABCO and Et3N achieved a quantitative yield and reduced the reaction time to 30 min (entry 11). Replacing NiBr2·glyme with alternative nickel catalysts did not alter the outcome (entries 12–15). Moreover, the optimal conditions can be applied to 4-chlorobenzonitrile (1b) with an extended reaction time of 90 min (entry 16). Crucially, the approach demonstrated scalability, smoothly transitioning from a 0.2 mmol scale directly to 2 mmol and 5 mmol scales within a 4 mL glass vial without necessitating any modifications (entries 17 and 18). Such scalability highlights the protocol’s potential for wider application in synthetic chemistry. When the argon atmosphere was replaced by air, no reaction was observed (entry 19).

Table 1 Optimisation of reaction conditions
Full size table

Scope of substrates

Encouraged by the initial findings, the scope of the substrates was examined (Fig. 3). We first investigated different aryl halides 1 in reactions with aniline (2a). A wide range of aryl bromides bearing electron-withdrawing and electron-donating groups were well tolerated, giving the corresponding products in moderate to good yields. Several functional groups at the para position such as cyano, ester, ketone, sulfone, trifluoromethyl, alkyl, and phenyl were all compatible. Boronic esters and chloro functionalities, providing the possibility for further late-stage modifications, also reacted well, affording 9 and 10 in 75% and 74% yields, respectively. The presence of ortho-cyano groups introduced steric hindrances, mildly affecting yield (16, 50%). Furthermore, di-substituted variants and bicyclic aryl bromides containing dioxole and ketone functionalities, as well as polyaromatic bromides, also demonstrated good reactivity in this transformation (18–24). Notably, the photo-mechanochemical method accommodated a variety of heterocyclic substrates with pharmaceutical relevance, including structures bearing indole, pyridine, quinoline, dibenzofuran, dibenzothiophene, and carbazole motifs (25–34). A set of biologically active moiety and drug molecules, including adamantanecarboxylic acid (35), pregnenolone (36), fenofibrate (37), and indomethacin (38), were also successfully applied, further demonstrating the generality of this methodology.

Fig. 3: Substrate scope.
figure 3

Reactions conditions: aryl halides (0.5 mmol, 1.0 equiv.), nucleophiles (2.0 − 6.0 equiv.), 4CzIPN (0.1 – 0.5 mol%), and nickel catalyst (5 mol%) in a 4 mL glass vial under irradiation of 90 W Aqua Blue LEDs. Yields after column chromatography. aUsing 40 W 440 nm Blue LEDs. bWith 4-chlorobenzonitrile (1b). cThe compound was isolated as its respective oxide. dAdding 200 μL of DMA.

Full size image

Next, the substrate scope with respect to nucleophiles was explored. 4-Bromobenzonitrile (1a) reacted smoothly with a wide array of N-, O-, P-, and S-containing nucleophiles. Various anilines regardless of their electronic nature were well tolerated, giving the corresponding products in good to excellent yields (39–43). Solid state 2-naphthylamine also reacted well, affording 44 in 90% yield. Furthermore, both primary and secondary alkyl-amine nucleophiles displayed high tolerance, affording the corresponding products in moderate to good yields (45–55). At this stage, the project had made significant progress by developing an effective solvent-minimised photo-mechanochemical protocol for aryl amination. However, its performance declined with low-reactivity substrates such as weak nucleophiles (e.g., S,S-dimethyl sulfoximine, benzamide, p-toluenesulfonamide) and an electron-rich aryl bromide (4-methoxyphenyl bromide). Under high-intensity, broad-spectrum irradiation (90 W Aqua Blue LEDs), the organic photocatalyst 4CzIPN might decompose, further reducing efficiency. In contrast, using 40 W 440 nm blue LEDs as the light source overcame these challenges, achieving excellent yields in all aforementioned cases (15, 56–58, for further information, see Supplementary Table 2). With respect to other heteroatom coupling partners, we adopted the conditions reported by König and co-workers67, simply omitting the solvent without additional adjustments. Under the irradiation of 40 W 440 nm Blue LEDs, effective coupling was achieved for primary alcohols (59–62), secondary alcohols (63, 64), and diphenylphosphine (65). Additionally, aliphatic thiol and thiophenol also reacted effectively, although the addition of 200 μL of DMA (η = 0.87-1.10 μL/mg) was required to inhibit their dimerisation (66–68). It is worth noting that all the reaction mixtures were viscous, with some forming heterogeneous phases due to solid content, yet the glass vials remained intact in all cases.

Reducing catalyst loading

Motivated by the high efficiency of the current protocol and aiming to further evaluate the performance of the photo-mechanochemical platform, we hypothesised that low catalyst loadings could still achieve comparable reactivity. To test this hypothesis, a series of experiments were conducted to explore the impact of reducing catalyst loading, as illustrated in Fig. 4a. Aniline (2a) and pyrrolidine (2c) were selected as nucleophiles for these studies, with reactions performed on a 0.5 mmol scale and a reaction time of 30 min. Initial experiments, conducted with a fixed nickel catalyst loading of 5 mol%, revealed that photocatalyst loadings below 0.01 mol% led to reduced reaction efficiency for both nucleophiles. Subsequent investigations with the photocatalyst fixed at 0.01 mol% highlighted a critical dependency on nickel catalyst loading, particularly for aniline (2a). Reducing the nickel loading to 1 mol% significantly compromised the yield. Considering that the RAM generates additional heat with larger amounts of reactants, this thermal characteristic was factored into further experiments conducted on a 2 mmol scale with aniline (2a). The results showed that nickel loadings of 0.1 mol% and 0.01 mol% produced yields of 78% and 45%, respectively, suggesting a potential benefit from the slight temperature increase. For pyrrolidine (2c), the system tolerated nickel catalyst loading as low as 1 mol%, but further reduction or scaling up to a 2 mmol scale did not significantly improve performance.

Fig. 4: Control and scale-up experiments.
figure 4

a Investigations of reducing catalysts loading. Reactions were performed on 0.5 mmol scale in a 4 mL glass vial for 30 min. The yields were determined by GC-FID with dodecane as the internal standard. a2 mmol scale. b Scale-up experiments of product 4 with images showing the reaction mixture before and after completion. c Scale-up experiments of product 70 with images showing the reaction mixture before and after completion. d Other scale-up experiments with 0.01% 4CzIPN as a photo catalyst.

Full size image

Scale-up experiments

Scalable synthesis is a fundamental requirement for applications in industrial catalysis. Building on these promising results, we expanded our experiments to larger scales to demonstrate the platform’s synthetic utility. Remarkably, for aniline (2a), product 4 was obtained in a 93% yield within 20 min using only 0.01 mol% of both photocatalyst and nickel catalyst on a 10 mmol scale, representing a 50-fold increase in reaction volume (10 mmol vs 0.2 mmol). This success was readily reproduced at 100 mmol and 300 mmol scales (500- and 1500-fold scale-ups), achieving 95% and 98% yields, respectively, with extended reaction times, as illustrated in Fig. 4b. Notably, the corresponding catalyst turnover number (TON) reached 9800. To the best of our knowledge, such high efficiency for nickel-catalysed cross-coupling reactions has not been previously reported. Similarly, for pyrrolidine (2c), a 92% yield of product 70 was achieved at a 10 mmol scale within 20 min using 0.01 mol% photocatalyst and 1 mol% nickel catalyst. Reducing the nickel loading to 0.5 mol% and prolonging the reaction time to 40 min maintained the yield. However, further reduction in NiBr₂·glyme loading led to incomplete conversion. Scaling up to 100 mmol (200-fold increase compared to the 0.5 mmol reaction scale), product 70 was obtained in a 91% yield after 2 h, with a corresponding turnover number (TON) of 91 for the nickel catalyst, as shown in Fig. 4c. In experiments at 100 mmol and 300 mmol scales, a normal 400 mL glass vial was used, and operations were paused for 5 min after every 20–30 min to prevent overheating. During these reactions, solid products and salts gradually precipitated, turning the mixture into a semi-solid, viscous state. The desired products were easily isolated through a brief water wash and slurry purification. This photo-mechanochemical approach avoids the need for solvents and eliminates more wasteful work-up procedures. In contrast, conventional solution-based photoredox catalysis typically requires large amounts of solvents (DMA, DMF, or DMSO) to ensure complete dissolution and homogeneous reaction conditions. The equipment used in this study is a small research instrument, with the combined mass of the glass vial and reactants at the 300 mmol scale approximately 500 g, well below its capacity limit of 1 kg. Even with the current setup, theoretical throughput rates of 569 g day−1 for product 4 and 200 g day−1 for product 70 could be achieved. Future designs aim to incorporate more effective heat management strategies and larger vials to enhance reaction efficiency further. As shown in Fig. 4d, the approach also proved effective for aryl bromides with electron-donating or neutral groups, and drug molecules containing aryl chloride, all reacting successfully with just 0.01 mol% photocatalyst. When butan-1-amine was used as a nucleophile, it led to a 43% yield of product 71 using 0.01 mol% photocatalyst and 1 mol% nickel catalyst. These results are particularly noteworthy because scaling in photochemical synthesis usually requires substantial re-optimisation of reaction parameters54,73. Additionally, the degradation of organic dyes and nickel64,68, which is commonly encountered in metallaphotoredox catalysis, often necessitates higher catalyst loadings to maintain reaction efficiency. Overall, the photo-mechanochemical platform holds promise as a powerful tool for rapid, large-scale synthesis of high-value chemical products in the pharmaceutical and chemical industries.

Comparison between photo-mechanochemical and solution-based approaches

To further demonstrate the efficiency of the newly devised photo-mechanochemical platform, a direct comparison was made with a traditional solution-based photocatalytic approach (Fig. 5a). The study focused on four different amines: aniline, an aromatic amine; 4-chloroaniline, a solid aromatic amine; pyrrolidine, a secondary amine; and butan-1-amine, a primary amine. Each amine was reacted with methyl 4-bromobenzoate (1 d) at a 10 mmol scale with low catalyst loading. In the photo-mechanochemical system, the reaction temperature stabilised at approximately 50–55 °C upon completion, while solution-based reactions were conducted at 55 °C in 0.5 M DMA for 48 h to ensure comparable conditions. The photo-mechanochemical method consistently delivered higher results across all tested amines, with reaction times significantly shorter (20 or 90 min vs. 48 h). Moreover, the superiority of this approach was highlighted by assessing green metrics such as the E factor, mass intensity, and molar efficiency (Fig. 5b, for details, see section 7 of the Supplementary Information). These results illustrate the platform’s potential to outperform traditional solution-based methods in both efficiency and sustainability. The platform’s versatility was validated through five additional photoinduced reactions, including: (i) Dual Ni/photoredox decarboxylative C(sp2)–C(sp3) cross-coupling reaction74, (ii) Ullmann-type C−N cross coupling75, (iii) intermolecular [2 + 2] dimerisation76, (iv) Arbuzov-type α-C(sp3)–H phosphonylation of aliphatic amine77, and (v) decarboxylative alkylation via electron donor–acceptor (EDA) complex78. All reactions demonstrated shortened reaction times and comparable yields under solvent-minimised conditions (Fig. 5c–g).

Fig. 5: Comparison between photo-mechanochemical and solution-based photocatalysis.
figure 5

a Methyl 4-bromobenzoate (1d, 10 mmol) was reacted separately with four different amines. The yields were determined by GC-FID with dodecane as the internal standard. PM (purple): photo-mechanochemical approach. SOL (blue): solution-based photocatalytic approach. PC: photo catalyst. b Calculation of green metrics. c–g Other reaction types, results comparing the two methods shown below the equations.

Full size image

In conclusion, we have developed a novel photo-mechanochemical platform that enables efficient dual-catalysed cross-coupling reactions under solvent-minimised conditions. This approach demonstrates multiple advantages, including operational simplicity, broad substrate tolerance, shortened reaction times, low catalyst loading, and compatibility with paste-like mixtures. The platform’s scalability and synthetic utility have been validated across reaction scales ranging from 0.2 to 300 mmol. Compared to traditional solution-based methods, this photo-mechanochemical strategy offers superior efficiency and environmental benefits. We anticipate that this platform will inspire further advancements in industrially viable, solvent-minimised photocatalytic processes, particularly in applications where conventional photochemistry is limited. Additionally, the high customizability of RAM opens possibilities for its integration into continuous production workflows, potentially expanding its industrial relevance.

Methods

Representative procedures

All reactants, catalysts and additives were loaded into a glass vial. The vial was then sealed and subjected to the photo-mechanochemical platform. After irradiation and mixing, the crude product was analysed by GC-MS and then purified by flash column chromatography using hexane/EtOAc as eluent. For scale-up experiments, if the starting materials achieved full conversion, the reaction mixture was washed with water and the collected solid was mixed with a small amount of ethyl acetate (EtOAc) to form a slurry. Hexane was then added to precipitate the product which was filtered and dried under reduced pressure.