Biasing mechanistically distinct reaction pathways by mechanochemistry

Abstract

Due to the solvent-free nature of the mechanochemical approach, it offers an opportunity to drive reactions with different selectivities from those under conventional solution-based conditions. Nevertheless, literature examples of pathway switching remain scarce. Herein, we report an unusual example where ball-milling offers three distinct pathways leading to three different products. By simply adjusting the amount of liquid additive, arene C–H functionalization can be biased between bromination and oxidation via either radical (Br˙) or ionic (Br⁺) intermediates, respectively.

---

In recent years, mechanochemistry has made a comeback in chemical synthesis due to the increased interest in developing greener and more sustainable methods to perform chemical transformations.¹ However, most reported examples emphasize the rate-enhancement aspect of mechanochemical approaches, while their unique role in controlling selectivity,² distinct from conventional thermal reactions, has received less attention. For example, Ito reported that mono-arylation is favored over diarylation for unbiased dibromoarenes by palladium catalysis under ball-milling conditions.³ Browne described that liquid assisted grinding (LAG) can kinetically bias C–H mono-fluorination over difluorination.⁴ Likewise, Mack showed that LAG with various polar solvents lead to different outcomes of Pd-catalyzed alkyne–alkyne coupling reactions.⁵ Furthermore, Friščić and co-workers reported that the isolation of elusive aryl N-thiocarbamoylbenzotriazole intermediates, which are unobservable by solvent-phase methods, becomes possible under mechanochemical conditions.⁶

Electrophilic bromination is an important industrial reaction for introducing bromine atoms in arenes,⁷ as the installed C–Br groups enable subsequent selective functionalizations.^8–10 Traditionally, bromination reactions using either elemental bromine or mild organic bromination agents are predominantly performed under solution-based conditions. Recent advances have demonstrated the feasibility of electrophilic bromination under ball-milling conditions.¹¹ For example, Mal^11b and Banerjee^11l independently showed that bromination proceeded with a broad range of arenes with ball-milling. Hernández presented a mass production protocol using extrusion techniques.^11e

Although arene bromination has been realized by both solution-based and mechanochemical approaches, most studies have focused on synthetic applications, leaving questions regarding reaction mechanisms and selectivity control largely unaddressed. Understanding the fundamental mechanistic differences between these approaches is critically important, yet it has been largely neglected. This knowledge gap exists because the divergent reactivity observed in mechanochemistry, in our opinion, is difficult to rationally anticipate or design.

Our group is particularly interested in uncovering the insights that explain how these two approaches can drive divergent reactivity from the same starting materials. Consequently, understanding the chemical origin of this unpredictable behavior is paramount for the rational design of divergent synthesis, representing a significant, uncharted area in chemical synthesis. We recently showed a divergent reaction pattern in the reaction of periodic acid and polycyclic aromatic hydrocarbons (PAHs) between ball-milling and solution-based approaches.¹² Specifically, solvent-free mechanochemistry afforded quinones via C–H oxidation, whereas the solution-based approach favored a C–H iodination pathway.

During a literature search, we encountered a report by Natarajan,¹³ in which the reaction of 9-anthracenecarboxaldehyde (1a) with N-bromosuccinimide (2) in a DMF/H₂O mixture preferentially yielded quinone oxidation products instead of the anticipated C–H bromination product. Motivated to develop chemical processes with reduced carbon footprints, we sought to investigate this unusual transformation under ball-milling conditions. In this work, we discovered a remarkable reaction pattern, where three distinct pathways are operative and tunable through ball-milling techniques (Fig. 1a). Explicitly, solvent-free milling favors C–H bromination, while LAG can bias the outcome toward either a selective deformylative bromination (with a trace LAG additive) or an oxidation pathway (with an excess LAG additive) (Fig. 1b). Mechanistic studies indicate that this pathway divergence stems from the generation of different reactive intermediates, dictated by the amount of liquid additive used during ball-milling.

Fig. 1 (a) Divergent reaction pathways controlled by different reaction approaches (amount of LAG solvent: neat milling, trace LAG and excess LAG). (b) the chemical example described in this work.

After further screening, we found that the solution reaction between 1a and 2 in EtOH, in addition to DMF/H₂O, also afforded the oxidation product 3a (Table 1, entry 13). In contrast, when this reaction was conducted under ball-milling conditions (40 Hz, 1 h) with 2 equiv. of 2, bromination products were formed selectively (98% combined yield, entry 2). In the product mixture, monobromo- (3a-Br) and dibromoanthracene (3a-Br₂), with the loss of a formyl group, were obtained in 62% and 36% yields, respectively. Only 2% yield of oxidation product 3a-O₂ was detected. While other electrophilic brominating agents were investigated, 2 proved the most effective under ball-milling conditions (Table S1). Increasing the milling time to 6 h had little effect on the reaction yields or product distribution (entry 5).¹⁴ Similarly, varying the amount of 2 (1 or 4 equiv.) completely suppressed the oxidation pathway (entries 1 and 3), albeit with diminished bromination efficiency (combined yield of 3a-Br + 3a-Br₂) (entries 1 and 3). Reaction temperature had a less pronounced impact on the outcome, as the reaction proceeded even at −20 °C (entries 6–7). In contrast, decreasing the milling speed (20 Hz) drastically hampered reactivity (entry 4). This strongly suggests that mechanical impact, rather than the heat generated during the milling process, is a prerequisite for inducing an effective transformation.

Table 1 Condition screening in the mechanochemical reaction between 1a and 2 under neat milling^a

Entry	ν (Hz)	Time (h)	Equiv. of 2	EtOH additive (mL mg^−1)	Yield (%) of 3a-Br/3a-Br2/3a-O2^b
a Ball milling parameters: 2 mL PP tube and ZrO2 balls (3.0 mm × 3) under open air condition.b Yields and product ratios determined by ^1H NMR spectroscopy with hexamethylbenzene as an internal standard.c Milling conducted at jar temperatures of 80 °C and −20 °C.d Solution reactions conducted in the EtOH solution reaction at 80 °C.e or at rt.
1	40	1	4	0	59/30/0
2	40	1	2	0	62/36/2
3	40	1	1	0	52/23/0
4	20	1	2	0	31/12/0
5	40	6	2	0	41/48/5
6	40	1^c	2	0	67/30/0
7	40	1^c	2	0	42/20/0
8	40	1	2	0.1	19/71/trace
9	40	1	2	0.2	20/59/13
10	40	1	2	0.5	3/21/36
11	40	1	2	2.0	3/10/60
12	40	1	2	4.0	0/2/41
13	—	6	2	Neat	—/—/90^d(75)^e

---

These divergent results prompted us to consider whether 3a-Br serves as an intermediate in the formation of either 3a-Br₂ or 3a-O₂. To test this hypothesis, a purified sample of 3a-Br was subjected to neat milling conditions identical to those in entry 2. However, no further reaction was observed. Similarly, control experiments starting from either 3a-Br₂ or 3a-O₂ also showed no conversion under these conditions (Fig. 2a). To further elucidate the reaction pathway, we performed kinetic profiling to monitor the product distribution over time (Fig. 2c(i)).¹⁵ These experiments revealed that the ratios of 3a-Br/3a-Br₂ remained constant throughout the course of reaction, fluctuating only slightly between 2.1 and 2.3 (Fig. 2c).

Fig. 2 (a) Control experiments of (a) solvent-free reactions between 3a-Br, 3a-Br₂ and 3a-O₂ and 2 and (b) a trace LAG reaction between 3a-Br and 2. (c) Reaction progress between 1a and 2 under different milling conditions: (i) solventless; and LAG with EtOH, (ii) 0.2 µL mg⁻¹ and (iii) 1.0 µL mg⁻¹.

Based on the observed divergent reactivity between solvent-free milling (3a-Br and 3a-Br₂) and solution processes (3a-O₂), the solvent effect was inspected in the context of LAG. To this end, each sample was injected with a trace amount (0.2 µL mg⁻¹) of a common organic solvent prior to milling treatment. Regardless of dielectric constant or proticity of the solvents used, these reactions predominantly favored bromination products (Table S2). Among them, only alcoholic solvents such as EtOH bias the formation of 3a-Br₂, up to 19 : 71 ratio for 3a-Br and 3a-Br₂ (Table 1, entries 8–9). With increasing EtOH additive content, the formation of 3a-O₂ increased up to 60% (entries 10–11). Up to 4.0 µL mg⁻¹, the bromination pathway is completely shut down, obeying a thermal reaction pattern in solution (entries 12–13). The kinetic traces (0.2 µL mg⁻¹ and 1.0 µL mg⁻¹) also revealed that the major product under each LAG condition (0.2 µL mg⁻¹: 3a-Br₂; 1.0 µL mg⁻¹: 3a-O₂) evolved rapidly and the product ratio was maintained for the rest of reactions (Fig. 2c(ii) and (iii)). In addition, 3a-Br₂ only minimally converts to 3a-O₂ under LAG conditions (1.0 µL mg⁻¹; 3% yield), while 3a-O₂ was completely recovered in an analogous reaction with 2 (2 equiv.). Meanwhile, an isolated sample of 3a-Br converts to 3a-Br₂ and 3a-O₂ in 57% and 22% yields under LAG conditions (1.0 µL mg⁻¹), respectively (Fig. 2b). This information suggests that three reactions between 1a and 2 proceed under kinetic control driven by the amount of solvent employed (solvent-free mechanochemical vs. LAG vs. solution-phase methods).

We next turned our attention to mechanistic understanding in the bromination reaction with 2 in ball-milling methods. When a spin-trapping agent, such as TEMPO, BHT or allyl-(2,2,6,6-tetramethylpiperidin-1-yl)oxyl (TEMPO)-based radical trap CHANT¹⁶ (2 equiv.), was added to a neat milling reaction between 1a and 2, the reactions were inhibited and 1a was fully recovered (Fig. 3a(i)). In the reaction with CHANT, the ESI-MS spectrum revealed the formation of new adducts 4 (calc. [M + H]⁺ 264.0489; found 264.0487) and 5 (calc. [M + H]⁺ 265.1547; found 265.1547) (Fig. 3a(ii)), derived from Br˙ or succinimide radical olefin addition in CHANT, respectively. The radical nature of the bromination reaction with 2 is further supported by the negative attempt with 5,5-dibromo-2,2-dimethyl-1,3-dioxane-4,6-dione (2-ii), as an alternative brominating agent in the reaction with 1a, where cleaving the C–Br bond is unlikely (Table S1). Moreover, KIE experiments with anthracene (1b-H₁₀/1b-d₁₀) gave k_H/k_D values of 3.09–3.66, supporting C–H bond cleavage as the rate determining step (Fig. 3b). Considering the solid-state packing effect, a milling reaction (40 Hz) between 1a and 2 was intentionally paused after 2 min. Storing or “aging” the milled sample in the dark for 48 h facilitated the conversion of 3a-Br and 3a-Br₂ from 20% and 10% (immediately after ball-milling) to 43% and 25% yield, respectively (Fig. 3c).^17,18 These experiments suggest that bromination with 2 likely follows a radical pathway, and the initiation occurring through homolysis of the N–Br bond of 2 via mechanical impact generated from a ball-milling process. Once mechanoradicals (Br˙ or succinimide radicals) are formed and stabilized within solid-state packing (see PXRD discussion below), as evident from the aging experiments, bromination could follow a chain-like mechanism even in the absence of mechanical impact, suggested by the aging experiments.

Fig. 3 Mechanistic studies with (a) spin-trapping reagents; and (b) KIE experiments determined (i) from an intermolecular competition and (ii) from two parallel reactions. (c) Solid-state aging experiments; and (d) PXRD experiments of 2 under different milling treatments.

To identify the active brominating species in LAG and solution-based processes, we analyzed the reaction mixture. When 2 was milled with a trace amount of EtOH (under LAG conditions), the ¹H NMR spectrum of the milled sample (6) in CDCl₃ revealed a new set of ethyl C–H resonances ∼0.5 ppm more downfield shifted than those in free EtOH, indicating the formation of a more electrophilic species (Fig. S8). This unexpected species 6 was tentatively assigned as [EtOHBr]⁺, structurally analogous to other active species responsible for electrophilic bromination formed in situ from Lewis bases and 2.¹⁹ We propose that highly reactive intermediates, such as bromine radicals (Br˙) in solvent-free milling and complex 6 (“Br⁺”) in LAG, are stabilized by the confinement effect within the crystalline phases. This hypothesis is supported by PXRD studies (Fig. 3d), which showed well-defined diffraction patterns for both neat milling (pure crystal phase of 2 in Fig. 3d(2)) and trace LAG samples (pure crystal phase of succinimide (7) in Fig. 3d(4)), indicating the existence of a crystalline environment. In the absence of the crystalline lattice, species 6 can neither be isolated nor detected under conventional solution-based reaction conditions (Table 1, entry 13), consistent with the absence of such a stabilizing solid-state confinement effect.

The origin of oxygen atoms in 3a-O₂ represents the final piece of the mechanistic puzzle. To identify this source, we conducted the following experiments. First, the reaction of 1a with succinimide (7) under standard conditions did not produce 3a-O₂, ruling out 2 as the oxygen source. Second, LAG reactions of 1a and 2 with EtOH under a N₂ atmosphere still gave 3a-O₂. Third, and conclusively, an LAG reaction employing H₂¹⁸O (η = 1.0) furnished 3a-¹⁸O₂, as confirmed by HRMS analysis (calc. for [M + H]⁺ 213.0682; found 213.0682) (Fig. S5). In addition, reactions under anhydrous ethanol still produced 3a-O₂, and these results confirm that protic solvents, including alcohols and water, can both serve as the source of oxygen atoms in 3a-O₂.

To evaluate the synthetic utility of solvent-free mechanochemical bromination, various naphthalene and anthracene derivatives were treated with 2 (2 equiv.) at 40 Hz for 1 h (Fig. 4). Substrates including anthracene (1b), 9-chloroanthracene (1c), 9-bromoanthracene (1d), 9-(chloromethyl)anthracene (1e), and 9,9′- bianthracene (1f) were efficiently converted to their corresponding dibrominated products (3b-Br₂ to 3f-Br₂). In contrast, anthracene derivatives bearing -Bpin (1h), -(Bpin)₂ (1i), -B(OH)₂ (1j), and -I (1k) underwent a tandem defunctionalization and dibromination, yielding 3-Br₂ as the sole product. The reaction of 9-methylanthracene (1g) with 2 resulted in sp³-CH bromination (3g-Br₂) in 44% yield, a result that supports a radical-mediated mechanism. Naphthalene derivatives containing either electron-withdrawing (1m) or electron-donating groups (1n–1p) were well tolerated, affording 3m-Br to 3p-Br₂ in good yields. Notably, the solution-phase reaction of 1m with 2 led to the formation of 1,4-naphthoquinone, mirroring the divergent reactivity observed for 1a and highlighting a key distinction from the mechanochemical process.¹³ Finally, 7H-dibenzo[c,g]fluoren-7-one (1l), a twisted aryl ketone and key precursor for organic semiconductors,²⁰ persistent organic radicals,²¹ and Feringa-type molecular rotors,²² was successfully dibrominated under the standard conditions, yielding 3l-Br₂ in 50% yield.

Fig. 4 Substrate scope in the mechanochemical reaction between 1 and 2 under solvent-free conditions. Yields determined by ¹H NMR spectroscopy with an internal standard. The experimental details for the isolation procedure and isolated yields for some of the products are described in the SI.

Conclusions

In summary, we have developed a mechanochemical strategy that selectively biases reaction pathways toward either C–H bromination or oxidation. Mechanistic investigations indicate that the mechanochemical environment favors a radical-chain pathway for C–H bromination, whereas solution conditions promote oxidation via brominium-like intermediates. Crucially, this work demonstrates that mechanochemistry can stabilize reactive species and thereby unlock distinct chemical space with reactivity divergent from that observed in solution. This ability to access hidden reagent reactivity underscores the potential of mechanochemistry to complement and expand the toolbox of conventional synthetic methods.

Author contributions

The work was conceptualized by H. L. and K. Y. All experiments were conducted by H. L, Z. H, Y. L, Y. J. and T. W. The manuscript was written by K. Y. with the input from all the authors. All authors have given approval to the final version of the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data supporting this article have been included as part of the supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d5mr00104h.

Acknowledgements

Financial support for this work was generously provided by the ShanghaiTech University start-up funding. We also thank other staff members at the Analytical Instrumentation Center of SPST, and ShanghaiTech University (contract no. SPST-AIC10112914) for characterization support.

Notes and references

1. (a) S. L. James, C. J. Adams, C. Bolm, D. Braga, P. Collier, T. Friščić, F. Grepioni, K. D. Harris, G. Hyett, W. Jones, A. Krebs, J. Mack, L. Maini, A. G. Orpen, I. P. Parkin, W. C. Shearouse, J. W. Steed and D. C. Waddell, Chem. Soc. Rev., 2012, 41, 413–447; (b) K. J. Ardila-Fierro and J. G. Hernandez, ChemSusChem, 2021, 14, 2145–2162; (c) E. Colacino, V. Isoni, D. Crawford and F. García, Trends Chem., 2021, 3, 335–339; (d) J. L. Howard, Q. Cao and D. L. Browne, Chem. Sci., 2018, 9, 3080–3094; (e) D. Virieux, F. Delogu, A. Porcheddu, F. García and E. Colacino, J. Org. Chem., 2021, 86, 13885–13894; (f) F. León and F. García, in Comprehensive Coordination Chemistry III, ed. E. C. Constable, G. Parkin and L. Que Jr, 3rd edn, 2021; (g) R. Liu, X. He, T. Liu, X. Wang, Q. Wang, X. Chen and Z. Lian, Chem.–Eur. J., 2024, 30, e202401376.
2. (a) C. Bolm and J. G. Hernandez, Angew. Chem., Int. Ed., 2019, 58, 3285–3299; (b) K. Y. Jia, J. B. Yu, Z. J. Jiang and W. K. Su, J. Org. Chem., 2016, 81, 6049–6055; (c) Y. Liu, F. Z. Liu and K. Yan, Angew. Chem., Int. Ed., 2022, 61, e202116980; (d) N. Haruta, T. Sato and K. Tanaka, Tetrahedron Lett., 2013, 54, 5290–5293; (e) J.-L. Do and T. Friščić, ACS Cent. Sci., 2017, 3, 13–19; (f) J. G. Hernández and C. Bolm, J. Org. Chem., 2017, 82, 4007–4019; (g) F. Cuccu, L. De Luca, F. Delogu, E. Colacino, N. Solin, R. Mocci and A. Porcheddu, ChemSusChem, 2022, 15, e202200362.
3. T. Seo, K. Kubota and H. Ito, Angew. Chem., Int. Ed., 2020, 58, 3285–3299.
4. J. L. Howard, M. C. Brand and D. L. Browne, Angew. Chem., Int. Ed., 2018, 57, 16104–16108.
5. S. Kaabel, R. S. Stein, M. Fomitsenko, I. Jarving, T. Friščić and R. Aav, Angew. Chem., Int. Ed., 2019, 58, 6230–6234.
6. L. Chen, M. Regan and J. Mack, ACS Catal., 2016, 6, 868–872.
7. (a) I. Saikia, A. J. Borah and P. Phukan, Chem. Rev., 2016, 116, 6837–7042; (b) D. A. Petrone, J. Ye and M. Lautens, Chem. Rev., 2016, 116, 8003–8104.
8. (a) A. Krasovskiy, A. Tishkov, V. Del Amo, H. Mayr and P. Knochel, Angew. Chem., Int. Ed., 2006, 45, 5010–5014; (b) T. Nagano and T. Hayashi, Org. Lett., 2005, 7, 491–493; (c) For their mechanochemical analogues, see ref. 9d–f; (d) R. Takahashi, A. Hu, P. Gao, Y. Gao, Y. Pang, T. Seo, J. Jiang, S. Maeda, H. Takaya, K. Kubota and H. Ito, Nat. Commun., 2021, 12, 6691; (e) P. Gao, J. Jiang, S. Maeda, K. Kubota and H. Ito, Angew. Chem., Int. Ed., 2022, 61, e202207118.
9. (a) Metal-Catalyzed Cross-Coupling Reactions, ed. A. de Meijere and F. Diederich, Wiley-VCH, Germany, 2nd edn, 2024; (b) For mechanochemical analogues, see ref. 9c and d; (c) K. Kubota and H. Ito, Trends Chem., 2020, 2, 1066–1081; (d) C. Wu, J. Lv, H. Fan, W. Su, X. Cai and J. Yu, Chem.–Eur. J., 2024, 30, e202304231.
10. (a) C. K. Prier, D. A. Rankic and D. W. MacMillan, Chem. Rev., 2013, 113, 5322–5363; (b) K. Kubota, Y. Pang, A. Miura and H. Ito, Science, 2019, 366, 1500–1504; (c) C. Schumacher, J. G. Hernandez and C. Bolm, Angew. Chem., Int. Ed., 2020, 59, 16357–16360.
11. (a) D. Margetič and V. Štrukil, Mechanochemical Organic Synthesis, Elsevier: Boston, 2016, pp. 235–282; (b) N. N. Karade, G. B. Tiwari, D. B. Huple and T. A. J. Siddiqui, J. Chem. Res., 2006, 366–368; (c) A. Bose and P. Mal, Tetrahedron Lett., 2014, 55, 2154–2156; (d) S. K. Bera and P. Mal, J. Org. Chem., 2021, 86, 14144–14159; (e) A. Bal, T. Kumar Dinda and P. Mal, Asian J. Org. Chem., 2022, 11, e202200046; (f) K. J. Ardila-Fierro, L. Vugrin, I. Halasz, A. Palčić and J. G. Hernández, Chem.:Methods, 2022, 2, e202200035; (g) G.-W. Wang and J. Gao, Green Chem., 2012, 14, 1125–1131; (h) Z. Liu, H. Xu and G. W. Wang, Beilstein J. Org. Chem., 2018, 14, 430–435; (i) D. Barišić, I. Halasz, A. Bjelopetrović, D. Babić and M. Ćurić, Organometallics, 2022, 41, 1284–1294; (j) J. Wong and Y. Y. Yeung, RSC Adv., 2021, 11, 13564–13570; (k) S. Patra, V. Valsamidou, B. N. Nandasana and D. Katayev, ACS Catal., 2024, 14, 13747–13758; (l) S. Patra, B. N. Nandasana, V. Valsamidou and D. Katayev, Adv. Sci., 2024, 11, e2402970; (m) D. Das, A. A. Bhosle, A. Chatterjee and M. Banerjee, Beilstein J. Org. Chem., 2022, 18, 999–1008; (n) E. Bartalucci, C. Schumacher, L. Hendrickx, F. Puccetti, I. d'Anciães Almeida Silva, R. Dervişoğlu, R. Puttreddy, C. Bolm and T. Wiegand, Chem.–Eur. J., 2023, 29, e202203466; (o) D. Barišić, M. Pajić, I. Halasz, D. Babić and M. Ćurić, Beilstein J. Org. Chem., 2022, 18, 680–687.
12. H. Luo, F. Z. Liu, Y. Liu, Z. Chu and K. Yan, J. Am. Chem. Soc., 2023, 145, 15118–15127.
13. P. Natarajan, V. D. Vagicherla and M. T. Vijayan, Tetrahedron Lett., 2014, 55, 3511–3515.
14. The conflicting data in Table 1 (entries 1 vs. 5), where longer milling times between 1a and 2 lead to the conversion of 3a-Br to 3a-Br₂ and 3a-O₂, but not in the reaction shown in Fig. 2a, were further investigated. It was revealed that the conversion of 3a-Br to 3a-Br₂ under neat milling becomes evident upon extended grinding (up to 3 h), as shown in Fig. 2c and Table S7. This supports a stepwise bromination mechanism: (i) initial phase (0–1 h): competitive formation of 3a-Br and 3a-Br₂ via parallel radical pathways (competitive C–H abstraction by Br˙). (ii) Later phase (1–6 h): gradual and slow secondary bromination of 3a-Br by residual 2, driven by persistent mechanical activation. Control experiments with isolated 3a-Br confirmed its conversion to 3a-Br2 (6% after 3 h) only in the presence of fresh 2.
15. As each time point requires pausing the ball-milling reaction and removing samples from the reaction vessel for periodic NMR analysis, each data point was obtained from an individual experiment.
16. P. J. H. Williams, G. A. Boustead, D. E. Heard, P. W. Seakins, A. R. Rickard and V. Chechik, J. Am. Chem. Soc., 2022, 144, 15969–15976.
17. Similar packing-induced reactivity in the solid state was previously observed and reported in ref. 17.
18. W. Ma, Y. Liu, N. Yu and K. Yan, ACS Sustainable Chem. Eng., 2021, 9, 16092–16102.
19. P. H. Huy, Eur. J. Org Chem., 2020, 10–27.
20. T. Nematiaram, D. Padula, A. Landi and A. Troisi, Adv. Funct. Mater., 2020, 30, 2001906.
21. Y. Tian, K. Uchida, H. Kurata, Y. Hirao, T. Nishiuchi and T. Kubo, J. Am. Chem. Soc., 2014, 136, 12784–12793.
22. T. van Leeuwen, J. Pol, D. Roke, S. J. Wezenberg and B. L. Feringa, Org. Lett., 2017, 19, 1402–1405.

---

Footnote

† H. L., Z.-Y. H. and Y.-N. L. contributed equally to this work.

---