Photophysical and morphological properties of dye aggregates

The DSA2+ amphiphile has a distyrylanthracene core bearing two trimethylammonium groups (Fig. 1c). It was synthesized as diiodide salt (DSAI), and counterion exchange reactions afforded DSAX compounds as chloride (DSACl), bromide (DSABr), nitrate (DSANO3) and hexafluorophosphate (DSAPF6) salts (Supplementary Information section 3). The molecules were characterized by 1H and 13C nuclear magnetic resonance spectroscopies, mass spectrometry, ultraviolet (UV)–visible absorption spectroscopy and infrared spectroscopy (Supplementary Figs. 1–32).

We first measured the absolute fluorescence quantum yields (ΦFL) of DSAX salts in water (0–1.5 mM) to monitor the aggregation, as ΦFL is a sensitive reporter of restricted intramolecular rotation in the aggregate state and thus of AIE44,45. DSAX samples with hydrated counteranions, such as Cl−, Br− and NO3−, showed no aggregation, as indicated by low fluorescence (ΦFL ≤ 1.1%), whereas hydrophobic PF6− caused the molecules to be always aggregated, as evidenced by relative high ΦFL ≈ 6.9% in these suspensions (Fig. 2a and Supplementary Figs. 33 and 34). Among the DSAX salts, only the iodide form (DSAI) exhibited concentration-dependent aggregation and emission behaviour, from weakly emissive dilute solutions to progressively fluorescent yellow suspensions at higher concentrations (Fig. 2a and Supplementary Figs. 35–38). This mirrors previous observations that higher concentrations of amphiphilic monomers provide self-screening for repulsive electrostatic interactions46. The specific ion effect is opposite to the conventional Hofmeister series in which charge-dense ions (for example, Cl−, Br− and NO3−) promote aggregation of macromolecules by reducing solubility47,48. Here, instead, charge-diffuse ions (I− and PF6−) promote aggregation, consistent with a reverse Hofmeister effect driven by ion pairing and monomers stacking.

Fig. 2: Supramolecular polymerization of DSA2+ amphiphiles into fluorescent and crystalline nanostructures.The alternative text for this image may have been generated using AI.

a, Assembly landscapes at different concentrations of DSA2+ amphiphile (mM) with selected counteranions (Cl−, Br−, NO3−, I− and PF6−). b, Assembly landscape at different concentrations of DSAI amphiphile (mM) under various ratios between water and organic solvent (DMSO). c, Assembly landscape at different concentrations of DSAI amphiphile (mM) with different amount of salt (NaI, mM). d, Transmission electron microscope (TEM) image of DSAI supramolecular nanostructures in water (DSAI 0.05 mM, NaI 0.5 mM). Scale bar, 500 nm. e, Atomic force microscope image of drop-casted aqueous suspension of DSAI supramolecular nanostructures (DSAI 0.05 mM, NaI 0.5 mM). Scale bar, 400 nm. f, Height profiles by AFM of selected sections (indicated by 1 and 2 from e) and relative step heights of ∼2 nm (grey arrows). g, A schematic representation of the dominant crystallographic faces of a DSAI nanocrystal, with a zoom-in on the layered slipped π–π stacking viewed along the b axis, as determined from 3D-ED data (atoms shown as ball and stick). h, Scanning electron microscope image of DSAI supramolecular nanostructures in water (DSAI 0.05 mM, NaI 0.5 mM). Scale bar, 400 nm. i, WFM image of DSAI supramolecular nanostructures in water (DSAI 0.05 mM, NaI 0.5 mM). Scale bar, 20 μm.

Source data

We further investigated the polymerization behaviour of DSAI, which can exist as either dissolved monomers or aggregates depending on concentration. Organic solvent minimizes interactions between monomers and promotes disaggregation through dissolution, whereas salts and electrolytes enhance charge screening and thus promote supramolecular polymerization. Hence, plotting ΦFL as a function of [DSAI] and either fraction of organic solvent (fDMSO, as vol% with water) or [NaI] provide assembly landscapes that visualize the propensity of DSAI to self-assemble under different conditions (Fig. 2b,c and Supplementary Figs. 40–43). In both cases, aggregation is promoted either by increasing the water content, which strengthens hydrophobic interactions or by increasing [NaI], which screens the positive charges, reduces electrostatic repulsion and promotes hydrophobic collapse. Consistent with this mechanism, aggregation of DSAX (X = Cl−, Br−, NO3− and I−) can be induced by addition of the corresponding sodium salts (NaX), which increase charge screening and promote aggregate formation (Supplementary Figs. 44–46). Aggregation can also be triggered by salts bearing counteranions different from those originally paired with the dye; for example, NaI induces aggregation of DSACl (Supplementary Figs. 47 and 48), showing that both ionic strength and counteranion identity contribute. Supporting this interpretation, counteranions are not merely dissociated but are embedded within the nanostructures (Supplementary Figs. 48–54).

We next characterized the nanostructures formed upon DSAI self-assembly using complementary techniques from the nanoscale to the bulk. Transmission electron microscopy revealed flat, elongated nanoribbons with high aspect ratio (Fig. 2d and Supplementary Fig. 55), indicating a preferential growth direction, probably via hydrophobic collapse of the aromatic cores. Atomic force microscopy showed that these flat nanoribbons are stacked lamellar structures with discrete step heights of ~2 nm (Fig. 2e,f and Supplementary Fig. 56). X-ray diffraction on the bulk powder revealed the high degree of crystallinity of the nanoribbons (Supplementary Fig. 57). Although bulk single crystals of pure DSAI suitable for X-ray analysis could not be obtained in water, continuous three-dimensional electron diffraction (3D-ED)49,50 provided atomically precise structural characterization of the nanocrystalline aggregates. 3D-ED analysis revealed face-on molecular packing in the aggregates (Fig. 2g and Supplementary Figs. 58–60), in which the DSAI chromophores adopt a typical slipped π–π stacked arrangement (θslip ≈ 44.6°, centre-to-centre distance of ~5.7 Å) with an interlayer spacing of ~1.9 nm typical for crystalline aromatic systems such as acenes and related polycyclic π-conjugated molecules. Scanning electron microscopy (SEM) showed that these nanoribbons assemble both laterally and through inter-ribbon stacking into micron-sized crystalline platelets (Fig. 2h and Supplementary Fig. 61), whereas widefield fluorescence microscopy (WFM) revealed bright emissive micron-sized domains (Fig. 2i and Supplementary Fig. 62).

Supramolecular aggregation can be systematically controlled by counterion, solvent environment and charge screening to yield structurally ordered, emissive nanoribbons, establishing the structural basis for the photocatalytic studies that follow.

Aggregation-induced light-driven ROS and H2O2 generation

We next investigated whether supramolecular assembly of DSA2+ improves light-driven generation of reactive oxygen species (ROS), modulating aggregation through counterion exchange, chromophore concentration and ionic strength (Fig. 3a). In each case, we assessed both fluorescence quantum yield (ΦFL) and light-driven ROS generation, the latter using the 3,3′,5,5′-tetramethylbenzidine (TMB)51,52 probe, which is oxidized by ROS to form a blue-coloured charge-transfer complex (Fig. 3a).

Fig. 3: Light-driven ROS generation by DSA2+ nanostructures.The alternative text for this image may have been generated using AI.

a, A schematic representation of aggregation modes (counteranion used, concentration and ionic strength) to obtain fluorescent nanostructures that mediate oxidation of the TMB probe into a blue charge-transfer complex (white-light LED, 180 mW cm−2 under air). b, Fluorescence emission spectra and fluorescence quantum yields (ΦFL) of DSAX (0.3 mM) with different counteranions (X = Cl−, Br−, NO3−, I− and PF6−). The inset shows the DSAX samples under UV light (from left to right: Cl−, Br−, NO3−, I− and PF6−). c, Light-driven TMB (134 μM) oxidation by DSAX over time. Statistics are from two independent groups. d, TMB oxidation rate (0.0 ± 0, 0.0 ± 0, 0.01 ± 0.02, 1.5 ± 0 and 3.7 ± 0.6 μM min−1) extracted from traces in c. Statistics are from two independent groups. e, Fluorescence emission spectra and fluorescence quantum yields (ΦFL) of DSAI at 0.1 mM concentration (yellow trace) and 0.4 mM (purple trace). The inset shows DSAI 0.1 mM (left) and 0.4 mM (right) samples under UV light. f, Light-driven TMB (134 μM) oxidation by DSAI over time. Statistics are from two independent groups. g, TMB oxidation rate (0.14 ± 0.04 and 3.3 ± 0 μM min−1) extracted from traces in f. Statistics are from two independent groups. h, Fluorescence emission spectra and fluorescence quantum yield (ΦFL) of DSAI (0.1 mM) without (yellow trace, 0.0 mM NaI) and with presence of salt (purple trace, 0.6 mM NaI). The inset shows DSAI 0.1 mM samples without (left) and with NaI present (right) under UV light. i, Light-driven TMB (134 μM) oxidation by DSAI without and with presence of NaI over time. Statistics are from two independent groups. j, TMB oxidation rate (0.08 ± 0.11 and 3.9 ± 0 μM min−1) extracted from traces in i. Statistics are from two independent groups.

Source data

Under the buffer conditions used for the light-driven TMB oxidation (acetate buffer, 0.1 M, pH 5), hydrophobic and poorly hydrated anions, particularly I− and PF6−, increased ΦFL from ≤1.1% (for Cl−, Br− and NO3− salts) to 3.5% and 8.7% for DSAI and DSAPF6, respectively (Fig. 3b). This fluorescence enhancement strongly correlates with photocatalytic activity under white-light irradiation: DSAPF6 exhibited the highest rate of TMB oxidation (v0 = 3.7 μM min−1), followed by I− (v0 = 1.5 μM min−1), whereas Cl−, Br− and NO3− showed no or minimal activity (Fig. 3c,d and Supplementary Fig. 63). This specific ion effect trend follows a reverse Hofmeister effect already observed in unbuffered water, suggesting that charge-diffuse anions reduce solvation and enable greater excited-states utilization in photocatalysis. Increasing DSAI from 0.1 mM to 0.4 mM in acetate buffer increased ΦFL about fivefold, from 1.1% to 4.8% (Fig. 3e), consistent with AIE. Correspondingly, the rate of TMB oxidation increases by more than one order of magnitude, from v0 = 0.14 μM min−1 to v0 = 3.3 μM min−1 (Fig. 3f,g and Supplementary Fig. 64), confirming that concentration-driven self-assembly translates into enhanced ROS generation. Salt-induced aggregation was probed by adding NaI to a dilute DSAI solution in acetate buffer. Even at constant chromophore concentration (0.1 mM), the presence of NaI (0.6 mM) increased ΦFL from 1.1% to 2.4% (Fig. 3h) and the ROS production rate from v0 = 0.08 μM min−1 to v0 = 3.9 μM min−1 (Fig. 3i,j and Supplementary Fig. 65). These results highlight the role of electrostatic screening in driving aggregation and suggest that subtle environmental changes can modulate functional output via supramolecular polymerization.

To assess how supramolecular aggregation affects the excited states, we measured fluorescence lifetimes (τavg) of solution and aggregates by time-correlated single photon counting. In all cases, the nanosecond fluorescence decays (Supplementary Figs. 66–70) indicate singlet excited states and motivated further femtosecond and nanosecond transient absorption measurements (TAS). In solution, the monomeric chromophore displayed a transient band at 620 nm that shifted to 545 nm and then decayed with lifetimes of 17 and 79 ps, respectively, consistent with rapid styryl photoisomerization in DSAI (Supplementary Fig. 71) and reported femtosecond transient absorption measurements spectra of 9,10-distyrylanthracene derivatives53. By contrast, DSAI aggregates showed no shift of the excited-state absorption bands (Supplementary Fig. 72). Kinetic analysis identified a fast 21-ps component, similar to the monomer, and a slower 320-ps component, as expected for rigidification-induced suppression of photoisomerization decay and enhanced excited-state stabilization. At 7 ns, the excited-state absorption band at 539 nm was still present, consistent with the lowest triplet excited state; this was confirmed by ns-TAS (Supplementary Fig. 73), although no phosphorescence was observed in the solid state at 77 K (Supplementary Fig. 74). The emission and transient absorption spectra of the aggregates also resemble those of the monomer in rigid media53 (Supplementary Fig. 75), supporting the formation of localized singlet and triplet excited states in the aggregated state. These results show that aggregation enhances photocatalysis by increasing the population of localized excited states. Although weak coupling interactions in slip-stacked molecules cannot be excluded, the large π–π stacking distances (> 5 Å), the absence of notable chromic shifts between solution and aggregated states, and the pump-probe data collectively indicate that DSAI aggregates operate through localized singlet and triplet excited states.

We then investigated the mechanism underlying this aggregation-induced ROS production. Using TMB oxidation as a readout, we evaluated the role of molecular oxygen. The oxidation rate was highest under pure O2 atmosphere, decreased under air, and was nearly abolished under N2, confirming that the photocatalytic process requires oxygen and light (Fig. 4a and Supplementary Figs. 76 and 77). Scavenger experiments51,52 showed that Tiron (3,5-pyrocatecholdisulfonic acid), a superoxide-specific scavenger,considerably suppresses TMB oxidation, whereas d-mannitol (hydroxyl radical scavenger) and l-tryptophan (singlet oxygen quencher) have negligible effect (Fig. 4b and Supplementary Fig. 78). These results point to superoxide (O2•−) as the primary intermediate generated by the photoexcited DSA2+ aggregates. Direct detection of O2•− by spin-trapping electron paramagnetic resonance (EPR) spectroscopy with 5,5-dimethyl-1-pyrroline N-oxide (DMPO) showed the characteristic DMPO–O2•− adduct upon irradiation54 (Fig. 4c and Supplementary Fig. 79). In addition, singlet oxygen emission and anthracene-9,10-dipropionic acid photodegradation experiments ruled out singlet oxygen sensitization (Supplementary Fig. 80). Both singlet and triplet excited states of aggregates are involved in electron transfer to O2 and superoxide formation, as shown by fluorescence quenching (Supplementary Fig. 81) and quenching of the triplet state by ns-TAS (Supplementary Fig. 73).

Fig. 4: Mechanistic insights into ROS generation by irradiation of DSA2+ nanostructures.The alternative text for this image may have been generated using AI.

a, Light-driven oxidation rate (5.1 ± 0.4, 3.1 ± 0.3 and 0.73 ± 0.29 μM min−1) of TMB probe (134 μM) by DSAI nanostructures (DSAI 0.3 mM) with white light (LED, 180 mW cm−2) under different atmospheres (O2, air or N2). Statistics are from two independent groups. b, Light-driven oxidation rate (3.2 ± 0, 2.9 ± 0.1, 0.0 ± 0 and 2.8 ± 0 μM min−1) of TMB probe (134 μM) by DSAI nanostructures (DSAI 0.3 mM) with white light (LED, 180 mW cm−2) in presence of ROS scavengers. Statistics are from two independent groups. c, In situ EPR spectra for DSAI nanostructures (DSAI 0.5 mM) and DMPO trap in EDTA (0.1 M, pH 6) under dark and irradiation conditions. d, H2O2 light-driven production over time by irradiation of DSACl nanostructures (DSACl 0.5 mM, NaCl 0.9 M) in MOPS (0.1 M, pH 7.0) with 415 nm (LED, 140 mW cm−2). Statistics are from three independent groups. e, Control experiments for H2O2 produced by DSACl aggregates (2.9 ± 0, 0.15 ± 0, 0.0 ± 0, 0.0 ± 0 and 0.0 ± 0 mmol g−1). Statistics are from three independent groups. f, Proposed photocatalytic mechanism for H2O2 production by the DSA2+ nanostructures (SD is the sacrificial donor, specifically MOPS in this case).

Source data

In the absence of TMB probe or ROS scavengers, photo-generated O2•− can undergo disproportionation or a further one-electron reduction to yield hydrogen peroxide (H2O2)55,56. We therefore monitored the production of hydrogen peroxide (H2O2) under continuous illumination. H2O2 production was linear over time, reaching 2,900 μmol g−1 after 4 h (Fig. 4d and Supplementary Fig. 82). Sustained activity requires a sacrificial electron donor, here the buffer 3-(N-morpholino)propanesulfonic acid (MOPS), which reduces the oxidized chromophore (DSA3+) and regenerates its ground state, thereby closing the photocatalytic cycle (Supplementary Figs. 83 and 84). Control experiments confirm these requirements: removing DSA aggregates, light, oxygen or the sacrificial donor resulted in no or negligible H2O2 production (Fig. 4e). On the basis of these results, we propose that photoexcited DSA aggregates (DSA2+*) transfer an electron to O2 to form O2•−, which undergoes disproportionation or a second one-electron reduction to yield H2O2, while the oxidized DSA3+ is reduced by a sacrificial donor, to regenerate the ground state and complete the photocatalytic cycle (Fig. 4f). Crucially, aggregation enables this reactivity by stabilizing the excited states and suppressing non-radiative decay through dye rigidification.

The ability to generate superoxide and H2O2 under ambient conditions, without metal co-catalysts, highlights the potential of AIP-based supramolecular systems for sustainable photocatalysis. We also demonstrated this by using DSA2+ in light-driven oxidation of organic substrates such as 4-(methylthio)phenol to the corresponding sulfoxide in the presence of oxygen (Supplementary Fig. 85). We also investigated the oxidation of glycerol, an abundant feedstock that can be upgraded to value-added chemicals57, and found that DSA2+ aggregates (DSAI 0.1 mM, NaI 25 mM) exhibit excellent activity for its selective conversion to glyceraldehyde under white-light irradiation (Supplementary Fig. 86).

Aggregation enhances photocatalytic hydrogen evolution

The aggregates show promising electron transfer properties for photocatalysis, as also supported by their calculated excited-state oxidation and reduction potentials relative to commonly used photocatalysts (Supplementary Figs. 87 and 88). The reduction potential of DSA2+ is nearly unchanged from the molecular to the aggregate state (Ered = –1.01 V versus –1.09 V versus NHE, respectively), as evidenced by electrochemical measurements (Supplementary Fig. 89) and calculations (Supplementary Figs. 90–92). This could be an advantageous feature, because strong hydrophobic interactions and molecular orbitals mixing between monomers often lead to large changes in redox potentials58. Thus, preserving redox and excited-state potentials across monomer and aggregate states simplifies predictive design. To assess whether AIP extends beyond oxygen photoreduction, we investigated the ability of DSA2+ aggregates to drive hydrogen evolution (Fig. 5a). We first explored photoinduced electron transfer to methyl viologen (MV2+), a well-established redox probe (E°′ = −0.45 V versus NHE)59. Upon irradiation of DSA aggregates we observed an increase in the UV–visible absorption bands of the MV•+ radical cation (Fig. 5b), whose formation was confirmed by its characteristic EPR spectrum (Fig. 5c). The ability of the DSA2+ to photoreduce MV2+ implies that the aggregates could also drive proton reduction (E°′ = −0.35 V versus NHE at pH 6).

Fig. 5: Light-driven hydrogen evolution by DSA2+ nanostructures.The alternative text for this image may have been generated using AI.

a, Proposed photocatalytic mechanism for methyl viologen (MV2+) reduction or hydrogen evolution in presence of a co-catalyst (PtNPs) and sacrificial donor (AA). b, UV–visible absorption spectra over time of a mixture of DSAI (1.0 mM) and MV2+ (10 mM) under 450 nm (LED, 700 mW cm−2) irradiation. c, EPR spectra over time of a mixture of DSAI (1.0 mM) and MV2+ (10 mM) under 450 nm (LED, 700 mW cm−2) irradiation. d, H2 light-driven evolution over time by irradiation of DSAI solution (yellow trace) or nanostructures (NaI 25 mM, purple trace) in AA (1.0 M, pH 4.0), in presence of PtNPs (8% mol) with white light (LED, 100 mW cm−2). The inset shows DSAI samples in AA without (left) and with (right) NaI under daylight and UV light. Statistics are from three independent groups. e, Control experiments for H2 produced by DSAI (10.9 ± 0.1, 0.00 ± 0, 0.00 ± 0, 0.00 ± 0 and 0.00 ± 0 mmol g−1). Statistics are from three independent groups. f, Absorbance over time by irradiation of DSAI solution or DSAI nanostructures in AA. The inset shows irradiated DSAI samples in AA without (left) and with (right) NaI under daylight and UV light. Statistics are from two independent groups. g, Emission spectra of DSAI 0.1 mM dissolved in AA 1.0 M (pH 4.0) and aggregated in AA 1.0 M (pH 4.0) with NaI 25 mM (λexc = 395 nm, at isoabsorbance). h, Transient absorption kinetic traces at 420 nm of an aqueous solution of aggregates of DSAI 0.05 mM and 2 mM of NaI (yellow trace) and after addition of methyl viologen ([MV2+] = 7 mM (green trace) and 20 mM (purple trace). λexc = 355 nm. A355 nm = 0.31, 1 cm optical path, 4.4 mJ per pulse. The inset shows the Stern–Volmer analysis. i, DSAI 0.1 mM with NaI 0.1 M emission quenching with different MV2+ concentration (λexc = 470 nm at isoabsorbance). The inset shows the Stern–Volmer analysis. SD, sacrificial donor.

Source data

We then investigated light-driven hydrogen evolution in presence of Pt nanoparticles (PtNPs, pre-made 3-nm colloidal dispersion) as co-catalysts and ascorbic acid (AA) as sacrificial electron donor (Fig. 5a). Interestingly, adding the DSAI to an AA solution dissolved the powder, probably because DSA2+ becomes solubilized through counterion exchange with solvated ascorbate anions. This enabled direct comparison of hydrogen evolution in the molecularly dissolved and aggregated states, the latter obtained by adding NaI under otherwise identical conditions. In the absence of NaI, when DSA2+ is dissolved, no H2 evolution is observed (Fig. 5d and Supplementary Fig. 93). By contrast, NaI-induced charge screening and aggregation increased hydrogen production to over 10,900 μmol g−1 after 4 h of white-light irradiation (Fig. 5d). Control experiments confirmed that all components (DSAI, AA, PtNPs and light) are essential for the observed activity (Fig. 5e and Supplementary Figs. 94–97). The apparent quantum yield for photocatalytic H2 evolution was 0.2%, with the photocatalytic activity matching the absorption profile of the aggregate (Supplementary Fig. 98), consistent with reported values for nanostructured chromophore amphiphiles25. Further gains in efficiency could probably be achieved by improving electron-transfer kinetics and interfacial coupling between the aggregates and co-catalysts; for example, in situ photodeposition of PtNPs can promote more efficient electron transfer and a better co-catalyst distribution than impregnation with pre-formed, ligand-capped PtNPs4.

After photocatalysis, the supramolecular aggregates were recovered by filtration. The recovered solid remained active, whereas the filtrate showed no activity under identical conditions (Supplementary Figs. 99 and 100), confirming that the photocatalytic function resides in the aggregated state. Consistent with this, aggregated DSAI was also more photostable than the solution species under prolonged illumination, which underwent steady photobleaching (Fig. 5f and Supplementary Fig. 101). Aggregation therefore both enables photocatalysis and stabilizes the chromophore against non-radiative decay and degradation.

Fluorescence experiments revealed that salt-induced aggregation markedly increased emission, with ΦFL going from 0.8% to 14% (Fig. 5g and Supplementary Fig. 102 for excitation–emission maps), whereas ns-TAS showed the emergence of a triplet state in the aggregated material (Fig. 5h). To quantitatively determine which excited state (singlet or triplet) dominates the productive electron-transfer step under catalytic conditions, we performed quenching experiments (Fig. 5h,i and Supplementary Fig. 103). These experiments revealed that both the excited states of the aggregated DSAI are quenched with increasing concentration of the acceptor MV2+, with the singlet (kSVS = 222 M−1 from fluorescence titration) prevailing over the triplet state (kSVT = 32 M−1 from ns-TAS titration). Altogether, these results show that aggregation enhances not only oxygen reduction but also proton reduction. Both reactivities arise from a common mechanism in which photoexcited DSA2+* aggregates undergo electron transfer to an external acceptor, either molecular oxygen or PtNPs. The higher fluorescence quantum yield of the aggregate indicates a greater abundance of excited states, increasing the probability of productive electron transfer either through singlet- or triplet-state reactivity. Aggregation also confers marked photostability, as shown by suppressed photobleaching under continuous irradiation. These features (higher excited-state utilization, enhanced reactivity and improved stability) further define the scope and advantage of the AIP approach.

Photocatalytic performance depends on aggregation pathway

The DSA nanostructures described so far were obtained immediately upon sample preparation in water (or buffer) and are therefore considered kinetic aggregates. Supramolecular systems, however, often exhibit multiple assembly pathways, and structures formed under kinetic or thermodynamic control can differ markedly in packing, photophysical properties and functional output2,60. We therefore investigated how kinetically trapped or thermodynamically controlled aggregates differ in structure, emissive behaviour and photocatalytic function.

To study the assembly pathway of DSAI (total concentration, cT = 50 μM) in NaI solution (5.0 mM), we monitored variable-temperature UV–visible absorption. Upon heating, the spectral changes indicate solubilization of the supramolecular polymer into monomers, whereas cooling shows transition from monomeric DSAI to its aggregate state (Fig. 6a and Supplementary Fig. 104), which we plotted as degree of aggregation (αagg)61,62,63. The non-sigmoidal cooling transition and thermal hysteresis between the critical temperatures of heating (Te = 351 K) and cooling (Te′ = 306 K) indicate a nucleation-elongation process61,62,63. This is further supported by Te′ shifting to higher temperature as the cooling rate was decreased progressively from 2.0 to 1.0 K min−1 (Supplementary Fig. 105). Upon dilution of DSAI, the elongation temperature (Te′) decreased linearly in the van’t Hoff plot, from which the standard enthalpy (ΔH°) and entropy (ΔS°) were estimated as −67 kJ mol−1 and −139 J mol−1 K−1, respectively (Supplementary Fig. 106). Furthermore, the lag time required to reach the thermodynamic equilibrium enabled supramolecular polymerization and crystal growth to be initiated by seeding. Seeding thermally annealed solutions with increasing amounts of pre-formed nanoribbons at 30 °C accelerated formation of extended crystalline structures and shortened the time to equilibrium (Fig. 6b). A schematic free energy diagram illustrates the higher energy barrier for direct nucleation than for seeded growth (Fig. 6c), consistent with the lag phase in unseeded samples. Microscopy confirmed that these different pathways lead to distinct morphologies: kinetically trapped aggregates appear as small particles (Fig. 6d and Supplementary Figs. 107–109), whereas seeded and thermodynamic aggregates yield