Styryldipyrylium monomers (1–4) were synthesized by trimolecular condensation reaction of pyrylium tetrafluoroborate salt with diformylarenes in ethanol (Supplementary Figure S1). The products were precipitated as yellow-to-orange powders (labeled as 1–4 M initial) and washed with ethanol and diethyl ether for purification. The synthesis procedures were performed in ambient conditions on a gram scale (Supplementary Figure S2). Single-crystal X-ray crystallography was performed to gain insight into the orientation of the monomers in the solid state. High-quality crystals of monomers 1 and 2 were grown by the slow diffusion of diethyl ether to acetonitrile solutions (Fig. 2a, Supplementary Figures S3 and S4) and revealed to adopt a 1D columnar packing (Fig. 2b for compound 2) where the distances between the neighboring olefins are 7.77 Å and 8.51 Å. The 1D columnar crystal structure of compound 1 is reported, showing a similar packing to 2.44 The large intermolecular distances exceeding 4.2 Å rationalize the photostability of the crystals,17 i.e., no [2 + 2] photocycloaddition even after 24 hours of irradiation at 470 nm, and such photostable monomers are labeled as 1-sM and 2-sM, respectively (Supplementary Figure S5). A 470 nm LED (21.4 µW/mm2) was employed for irradiation, based on the maximum absorption wavelength of 1 at 485 nm in CH2Cl2 (vide infra). During the crystal growth, monomer 1 also afforded a minor polymorph (less than 4.8%, Supplementary Figure S6) that appears as orange crystals and displays rapid and complete photopolymerization under a broadband tungsten-halogen lamp (visible to infrared; used during the optical microscopy of crystals) or a 470 nm LED, accompanied by a discoloration of the crystals (Fig. 2c, Video S1). The crystal structure of this polymorph shows an offset π-stacked packing and inter-olefin distance of 3.64 Å (Fig. 2d, Supplementary Figures S7–S9), which allows for facile [2 + 2] photocycloaddition (Fig. 2e). This photoactive crystalline form of compound 1 is labeled as 1-cM to distinguish it from the photoactive amorphous form of compound 1 (1-aM, vide infra). The colorless polymer crystals (1-cP) produced by the irradiation were solved to be triclinic P-1, identical to the space group of 1-cM, demonstrating single-crystal-to-single-crystal photochemical conversion with a small volume change of 2% (Supplementary Figures S10-S11, Table S1). Thus, 1-cM is distinguished from a previously-reported photoactive styryldipyrylium (X– = [SnCl4(HCOO)]–) that underwent an incomplete crystal-to-crystal conversion (80%) to yield short oligomers with less than 5 repeating units.44
As mentioned before, most of the single crystals obtained from the recrystallization of monomer 1 were photostable (1-sM), with a minor fraction of photoactive 1-cM. In order to maximize the yield of photoactive 1-cM, various recrystallization methods were explored, and rapid addition of diethyl ether to monomer 1 in acetonitrile was discovered to produce photoactive orange powder (Fig. 3a top). The monomer conversion (%) was calculated as the 1H NMR integral ratio between the residual monomer and an internal standard (1,3,5-trimethoxybenzene). The orange powder undergoes a facile TCP (84% conversion of monomer in 3 h and 95% in 24 h) under 470 nm irradiation (Supplementary Figures S12-S13). The powder X-ray diffraction (PXRD) patterns of the orange powder before and after irradiation closely matched the patterns of 1-cM and 1-cP, respectively, which were simulated based on the crystallographic data of their single crystals (Fig. 3b, Supplementary Figure S14). Thus, the orange powder obtained by the rapid recrystallization is considered to adopt the crystal packing of 1-cM as displayed in Fig. 2d. On the other hand, the slow addition of diethyl ether to monomer 1 in acetonitrile over 30 minutes primarily afforded photostable yellow powder (Fig. 3a bottom) that is expected to adopt molecular packing of 1-sM. Thus, we were able to achieve a scalable and selective synthesis of 1-cM and 1-sM by controlling the kinetics of recrystallization.62
However, the rapid recrystallization method was not as effective in producing photoactive crystals (analogous to 1-cM) for compounds 2–4. Table 1 shows the low topochemical photo-conversion of recrystallized monomers 2 and 4 (< 1% and 7%), measured by 1H NMR (Supplementary Figures S15–S17). The monomer conversion only increased by 1–9% even after the rapid recrystallization, indicating that the majority of molecules adopt a photostable packing (Supplementary Figure S18). Monomer 3 exhibits a more substantial increase in photoconversion by 23% upon rapid recrystallization; however, the maximum conversion remains 62%, highlighting the low efficacy of solution-based crystallization in achieving molecular arrangements that are susceptible to [2 + 2] photocycloaddition. The optical images and powder XRD patterns of compounds 1–4 prepared by the slow and rapid recrystallization as well as grinding are shown in Supplementary Figures S19–22. We have also explored other solution-state recrystallization conditions, including the combinations of CHCl3, Et2O, EtOH, and hexanes, to find that the photostable packing is primarily obtained. Although further extensive screening of recrystallization conditions may be able to alter the crystal packing, herein, we highlight that the mechanoactivation of photostable monomers is an effective and alternative tool that yields higher photoconversion of monomers (Table 1, highlighted in green).
Table 1. Conversion (%) of monomers upon irradiation at 470 nm for 24 h in solid state, analyzed by 1H NMR.
Monomer 1 changed from yellow crystals to orange amorphous powder, losing diffraction peaks, after 30 minutes of solvent-free grinding in an agate mortar (Fig. 3b bottom, Fig. 4a, Video S2). The amorphous powder (1-aM) is photoactive, turning off-white upon 470 nm light irradiation through photopolymerization and loss of conjugation. We monitored the PXRD pattern changes during the gradual dry grinding process (5, 15, and 30 minutes), which confirmed that mechanical stimulation results in amorphization, rather than a crystal-to-crystal phase transition (Supplementary Figure S23). The solid-state diffuse reflectance spectra of 1-aM and 1-cM display similar profiles (Fig. 4b), both red-shifted from the spectrum of the initial photostable monomer 1. The red shift indicates the formation of J-aggregates consisting of offset π-stacked monomers, which is supported by the TD-DFT calculated low-energy electronic transition of two neighboring monomers in 1-cM crystal structures (Supplementary Figure S24). In addition, the solid dispersion fluorescence spectrum of 1-aM (λem = 554 nm) is also red-shifted from the spectrum of initial monomer 1 (λem = 517 nm), corroborating the formation of J-aggregates (Supplementary Figure S25, vide infra). Compounds 2–4 also display the loss of diffraction peaks and red-shifted fluorescence upon grinding. Moreover, matching chemical shifts of 1-aM and 1-cM in solid-state 13C NMR suggest similar local molecular environment in both photoactive solids (Supplementary Figures S26a–c and S27).
We confirmed that mechanoactivated monomer 1-aM was effectively polymerized by the irradiation at 470 nm in the solid state. The polymerization of amorphous monomer was monitored by 1H NMR, UV-Vis absorption spectroscopy, and size exclusion chromatography (SEC) of the produced polymer. 1H NMR confirms the complete loss of monomer peaks after 15 hours of irradiation, which is accompanied by the appearance of broad polymer signals (Supplementary Figures S28). Similar 1H NMR spectral changes were observed during the photopolymerization of ground monomers 2–4 (Supplementary Figures S29-S31), and time-dependent monomer conversion (%) under irradiation was monitored (Supplementary FigureS32). Due to the broad signals, it was challenging to estimate the molecular weight of polymer by end group analysis. UV-Vis absorption spectra (Fig. 4c, Supplementary Figure S33), obtained by dissolving aliquots of solid-state reaction mixture, also show the progress of reaction. The prominent peak at 485 nm sharply decreased while the peak at 293 nm emerged over reaction time, which resembles the spectral change of styrylpyrylium during solid-state photo-dimerization, as recently reported.63
Figure 4d shows the continual growth of the peak at 293 nm beyond 15-hour irradiation, which marks the complete consumption of monomers, suggesting the further conversion of oligomers to longer polymers. SEC analysis confirmed the continuous growth of the polymer during the extended irradiation, and the weight average molecular weight (Mw) of the produced 1-aP after 72 h of irradiation was 1.8 × 104 g mol–1 (Fig. 4e, Supplementary Table S2, Figure S34). A lower Mw of 4.6 × 103 g mol–1 was obtained when the monomer 1-aM was irradiated at a longer wavelength (530 nm), and a higher Mw of 2.5 × 105 g mol–1 when the same amount of monomer was spread thin over a larger irradiation area. These values are comparable to the Mw of other recyclable polymers64 and topochemical polymers.9,10,11,16 Another monomer (3-aM) polymerized to similar Mw (6.7 × 104 g mol–1), while 2-aM and 4-aM afforded oligomers (Mw < 2.0 × 103 g mol–1). This observation indicates that monomers prone to form photoactive crystals (cM) upon recrystallization are also more likely to be mechanoactivated and photopolymerized. Therefore, we confirm that the success of solid-state photochemical reaction is dependent on both molecular designs and activation methods (Supplementary Figure S35, Table S3). We note that the molecular weights of ionic polymers are reported to be often underestimated because of the strong interactions between polymers and the stationary phase of the SEC, which also results in overestimated polydispersity indices (PDI) (Table S2).11,65 In our experiments, even the ionic monomers 1–4 exhibited broad SEC signals and large PDI values ranging from 1.9 to 3.2, because of such strong interactions (Supplementary Figure S36). To further evaluate the molecular weights of polymers, we also performed diffusion ordered NMR spectroscopy (DOSY),66 which revealed a wide range of diffusion coefficients of 1-aP (5.9–2.0 × 10–10 m2 s–1) and a number average molecular weight (Mn) of 7.0 × 103 g mol–1 (Supplementary Figure S37), corroborating the SEC analysis of polydisperse 1-aP. We hypothesize that such polydispersity also results from the non-uniform grinding and irradiation, intrinsic to solid-state activation and reaction conditions.
Solid-state 13C NMR was performed to clearly monitor the photo-induced conversion of olefin to cyclobutane. The matching chemical shifts of 1-aP and 1-cP confirm the formation of cyclobutane moieties upon the irradiation of both 1-aM and 1-cM (Fig. 5a, Supplementary Figures S26 d-e), and 1-aP and 1-cP also exhibit similar infrared (IR) spectra (Fig. 5b). The broader 13C NMR and IR peaks of 1-aP, compared to those of 1-cP, are attributed to the polydispersity and low crystallinity of 1-aP, which is supported by the comparative PXRD patterns of 1-aP and 1-cP (Fig. 5c). From these spectroscopic results, we conclude that the chemical structures of 1-cP and 1-aP are analogous, despite the drastic difference in their crystallinity.
A large solubility difference between amorphous 1-aP and crystalline 1-cP was observed, highlighting the potential of mechanoactivated amorphous-state polymerization for creating soluble and processable polymers. 1-aP shows a high solubility in acetonitrile (> 67 g L–1) and moderate solubility (2–6 g L–1) in other organic solvents including dichloromethane, chloroform, and acetic acid (Table S4), in sharp contrast to completely insoluble 1-cP (Fig. 5d left). Due to the insolubility of 1-cP in all common organic solvents tested, solution NMR and SEC analysis of 1-cP is not viable. In addition, MALDI-TOF mass analysis performed to compare the molecular weights of 1-cP and 1-aP did not yield any signals above the trimer, presumably due to fragmentation (Supplementary Figure S38). The larger solubility of 1-aP could be attributed to multiple potential factors including the larger disorder in polymer backbones, lower inter-chain interactions, shorter polymer chains, and larger polydispersity, compared to 1-cP. The solutions of 1-aP could be spin-coated to form highly uniform and transparent films, whereas dispersions of 1-cP formed inhomogeneous and opaque films (Fig. 5d right, Supplementary Figure S39).
We also investigated the thermally-induced depolymerization of 1-cP and 1-aP, building upon the successful [2 + 2] cycloreversion of styrylpyrylium dimers in our previous report.63 Differential scanning calorimetry (DSC) and thermogravimetric analysis (TGA) reveal the complete depolymerization of 1-cP at 262 ℃ (Supplementary Figures S40–43) and high thermal stability of the produced 1-cM below 297 ℃. The crystal structure of 1-cP shows the cleavable C–C bond length of 1.589(3) Å in the cyclobutane units, which rationalizes the ease of cycloreversion.16 In contrast, 1-aP undergoes thermal decomposition at temperatures above 120°C rather than depolymerization. The chemical stability of 1-aP below 120°C is confirmed by 1H NMR, solid-state 13C NMR, UV-vis absorption, SEC, and optical microscopy (Supplementary Figures S44–S48). To explain the different thermal response of 1-cP and 1-aP, we hypothesize that mechanically disordered monomers (1-aM) can undergo [2 + 2] photocycloaddition to generate both all-trans (1r,2r,3r,4r) and cis-trans (1R,2R,3S,4S) cyclobutane isomers within the 1-aP backbone, in contrast to 1-cP solely composed of cis-trans isomer.67,68 Our DFT calculation of model dimers suggests that an all-trans cyclobutane ring possesses a shorter C–C bond (1.571 Å) than that of a cis-trans cyclobutane ring (1.594 Å) (Supplementary Figure S49, Tables S5-S6). The presence of uncleavable all-trans cyclobutane rings in the 1-aP backbone may be a reason for its different thermal properties from those of stereoregular 1-cP. The marked differences in crystallinity, solubility, and thermal stability between 1-aP and 1-cP demonstrate how one can switch activation method (mechanical vs. solution-based) to produce distinct polymers from the identical monomer.
Lastly, the mechanoactivation of monomer and its solid-state photopolymerization enabled a unique encryption application.69–72 The solid-state emission of initial monomer 1 changes from green to orange by grinding, and the initial state is restored upon the exposure of 1-aM to CH2Cl2 vapor (Fig. 6a). Also, 470 nm irradiation on 1-aM generates 1-aP that lacks fluorescence due to the loss of conjugation upon [2 + 2] cycloaddition. This force-, solvent-, and light-responsive fluorescence change of the compound was utilized to generate a security code that is unreadable under ambient light and only detected in the dark under UV illumination (Fig. 6b). A ground mixture of 1-aM (5 wt%) and calcium sulfate was pressed into a chalk pellet and uniformly deposited on paper by drawing (Fig. 6b left, Supplementary Figure S50), which appeared pale orange under room light and showed red emission under UV. The color of the chromophore is diluted in the mixture with calcium sulfate. A QR code was printed on paper by 470 nm irradiation through an optical mask (Fig. 6b center), discoloring the orange paper through photopolymerization and leaving the covered area intact. The initially printed QR code can be read both under ambient light and under UV (395 nm) in the dark, because of the colorless and non-fluorescent polymer 1-aP. The QR code was then exposed to CH2Cl2 vapor for 5 minutes to make the code unreadable under the ambient light, turning the orange code to pale yellow that has insufficient contrast to the background (Fig. 6b right). The background consisting of 1-aP does not show any vapochromism. Thus, the information embedded in the QR code is "locked" under the ambient light and accessed only with a “key”: UV illumination in the dark. The green fluorescence of the QR code has sufficient contrast to the background, rendering it easily readable. Since the final QR code is encrypted by the combination of photostable monomer (1-M initial) and irreversible polymer (1-aP), the information is preserved even upon the repeated exposure to UV and visible light. This security feature can be printed on various substrates including fabric, frosted glass, plastics, and paper by abrasion, followed by the irradiation through a mask and vapor annealing, which offers a simple and versatile encryption method.