Stereospecific supramolecular polymerization of nanoclusters into ultra-long helical chains and enantiomer separation

Abstract

During the construction of supramolecular polymers of smaller nanoparticles/nanoclusters bearing hierarchy and homochirality, the mechanism understanding via intuitive visualization and precise cross-scale chirality modulation is still challenging. For this goal, a cooperative self-assembly strategy is here proposed by using ionic complexes with uniform chemical composition comprising polyanionic nanocluster cores and surrounded chiral cationic organic components as monomers for supramolecular polymerization. The single helical polymer chains bearing a core-shell structure at utmost length over 20 μm are demonstrated showing comparable flexibility resembling covalent polymers. A nucleation-elongation growth mechanism that is not dealt with in nanoparticle systems is confirmed to be accompanied by strict chiral self-sorting. A permeable membrane prepared by simple suction of such supramolecular polymers displays high enantioselectivity (e.e. 98% after four runs) for separating histidine derivatives, which discloses a benefiting helical chain structure-induced functionalization for macroscopic supramolecular materials in highly efficient racemate separation.

Understanding the mechanism for supramolecular polymer assembly from nanoclusters bearing hierarchy and homochirality is challenging. Here, the authors report a cooperative self-assembly strategy using ionic complexes with chiral surfactants through a nucleation-elongation growth mechanism with chiral self-sorting and prepare a permeable membrane for the separation of histidine derivatives.

Introduction

Supramolecular polymers (SPs), which are connected by isolated molecular units via noncovalent bonds, perform unique characteristics differing from covalent polymers in fabricating multicomponent and multiscale nanomaterials for various applications^1–3, such as stimuli-response, self-healing, recycling, and resynthesis, etc., due to the dynamic features of intermolecular interactions^4–6. During the construction of SPs for diverse target properties, the supramolecular monomers and intermolecular interactions play crucial roles in controlling the polymerized structure and functions^7,8. Because of the delicately governable feature for selective modulation among multiple interactions, the supramolecular polymerizations of organic small molecular and simple block copolymer units have been deeply elucidated on the mechanisms at monomer level. The equal-K model (isodesmic/step-growth mechanism) and nucleation-elongation model (cooperative/chain-growth mechanism) are accepted for most fabricated supramolecular polymers^6,7. In contrast to past success, the understanding of the polymerizations for micelles/block copolymer nanoparticles, inorganic nanoparticles, and multicomponent complexes analogous to covalent polymerization is still catching up with the emerging class of polymerized systems at the molecular level, although the relevant self-assemblies have been achieved over two decades^9–11. Such discordance can be attributed to the less oriented binding interaction control and the unfavorable stereospecific matching with each other of nanoparticular monomers. Typically, the metal/polymer nanorods/-particles were demonstrated to polymerize into linear assemblies via selective head-to-tail or lateral interaction, providing a conceptual extension to the supramolecular polymerization. The polymerization of crystallizable block copolymers/micells^12,13, colloidal nanoparticles¹⁴, metal nanoparticles^15,16, and nanoclusters¹⁷, have been reported following the same route. However, the structural symmetry of rigid monomers like the collision model and the limitation of locally selected/phase-separated surface modification mostly make these systems only follow the step-growth mechanism. Whether the nucleation-elongation model is applicable and how the functions of used components incorporate into the polymer architectures^18,19 like those in traditional polymers and biopolymeric walls/fibrils bearing helical surfaces^20,21 are still highly desired.

For this purpose, designing representative isotropic monomers maintaining conformation adaptability to different routes of dynamic supramolecular polymerization with structural stability becomes inevitable. From the evolved comprehension point of view, the specifically induced properties will inspire wide potentials of the SPs for applications. In contrast to metal or metal oxide nanoparticles, polyoxometalates (POMs) as a kind of polyanionic nanoclusters bearing uniform nanosize and moderate charges^22,23 show superiorities in the mimicry of self-sorting^24,25 and recognition^26,27 as the analogs of biomacromolecules and match the present motivation with functional extensibility^28–30. Despite the advances in POM-based linear assemblies with the assistance of organic components, the modulated polymerizations with a clear mechanism and nanostructure-induced properties are still on the way^31–34. Fortunately, the combined regulations of POM complexes hold the intrinsic features in serving the cooperative assemblies from organic counterions^35–37. Compared to the sole small molecules, the binding mode of the multicomponent monomers broadens the controlling of supramolecular polymerization with preferential composition and specifically induced properties in material performance^38,39.

In this context, we herein design a type of ionic complexes (iCOMs) comprising a rigid POM cluster core and surrounded chiral cationic amphiphiles⁴⁰ (Fig. 1a) as the monomers via an ion exchange reaction. The direction-free ionic interaction between charged components provides characteristics of structural adaptability for the organic counterpart moving on POM surface⁴¹, while the directional H-bonds by incorporated amide and urea groups enable linear supramolecular polymerization of the spherical nanoclusters. The formed core-shell-structured helical SPs (HSPs) with ultra-high aspect ratios with flexibility like traditional polymers are observed to grow into macroscopic single chains following the nucleation-elongation growth mechanism accompanied by a chiral self-sorting of iCOM monomers. Based on the structural analysis and molecular dynamics (MD) simulation, hydrogen (H)-bonding interactions are deemed to drive the homochiral connection. The shuttlecock-like conformation of the initial iCOM is interpreted to dominate the linear helical packing. Due to the stereospecific restriction, the chirality of the organic component has been amplified in polymerized chains. More importantly, the obtained HSP chain maintains the helical structure during its processing into nanofibrillar membranes, which demonstrates the unexpected capability in the separation of enantiomers with high stereoselectivity.

Fig. 1. Design of ionic complex (iCOM) monomers and their supramolecular polymerization in solution.

a Molecular structures represent inorganic polyanion and chiral cationic amphiphiles forming iCOM monomer into helical supramolecular polymer (HSP) chain. b Schematic drawing for the states of the prepared HSP chains under different conditions.

Results

Formation of HSP chains

The prepared cationic enantiomers (PULG·Cl and PUDG·Cl, Supplementary Fig. 1) bear a pyridinium head, an enantiomeric glutamide spacer with a urea group, and two alkyl chain tails (Supplementary Figs. 2–8). The iCOMs with formulae of (PULG)₄SiW and (PUDG)₄SiW are prepared via an ionic combination of one sphere-like polyanionic cluster [SiW₁₂O₄₀]⁴⁻ (abbreviated as SiW, Supplementary Fig. 9) and four cationic enantiomers according to the charge balance. By comparing to the ¹H NMR spectra of free organic enantiomers, the prominent peak shifting and broadening of protons near pyridinium confirm the electrostatic interaction between the cationic head group and the inorganic nanocluster. The detailed characterizations of elementary analysis, thermal gravimetric analysis (TGA), Fourier-transform infrared (FT-IR), and mass spectra (MS) suggest a definite molar ratio of the two components in the precision charge-compensated iCOMs (Supplementary Figs. 10–14 and Supplementary Table 1).

The ultra-long and fine linear HSP chains with changeable alignment emerge (Fig. 1b) by dissolving iCOM monomers either in chloroform or its mixed solvent with n-hexane. The chains aligning with each other parallelly in chloroform at randomly curled state are observed (Fig. 2a–d). These chains gradually dissociate into an entirely monodispersed state with increasing the volume content of n-hexane to 80% in the mixed solvent (Supplementary Fig. 15). By blotting the solution spreading on Cu grid or silicon wafer along a certain direction, the randomly dispersed HSP chains become straight with slightly twisting at a length exceeding 20 μm, as detected in transmission electron microscopy (TEM) and atomic force microscopy (AFM) images (Fig. 2e–h and Supplementary Fig. 16). In the case of natural evaporation, the polymerized chains appear to be ether entangled or self-cyclized (Supplementary Fig. 17) due to unidirectional surface tension, revealing that they form in solution and possess polymer-like flexibility⁴². By taking enantiomer (PULG)₄SiW as an example, the HSP chains are further identified to bear inorganic cores in the middle dispersing in the form of spot strings, as shown in the images of high-resolution (HR)-TEM and high-angle annular-dark-field (HAADF) scanning TEM (STEM) (Fig. 2i–l). The result demonstrates that iCOMs serve as a monomer to polymerize into single chains in parallelly aligned packing or well-dispersed state rather than conventional lamellar assemblies⁴³. The estimated utmost molecular weight (M_w) reaches about 7.7 × 10⁷ Da based on the size and M_w of iCOM monomer (5398.0 g mol⁻¹). The high-resolution AFM images show the height of super fine HSP chains at ca. 5.1‒5.4 nm (Fig. 2m–p and Supplementary Fig. 18), smaller than the calculated ideal diameter of iCOM in total 7.0 nm of twice the length of stretched PULG and the diameter of SiW cluster. Meanwhile, the isolated single HSP chains of two enantiomeric iCOM monomers display opposite helical morphologies, demonstrating a spatial limitation during the polymerization.

Fig. 2. Morphological characterizations of HSP chains.

a–d Morphological images of the HSP chains in chloroform at room temperature (insets showing local magnification, scale bar 20 nm): a, b TEM and AFM of (PUDG)₄SiW, c, d TEM and AFM of (PULG)₄SiW. e–h Morphological images of single HSP chains in a mixed solvent of chloroform/n-hexane (2:8) by blotting sample solution along a certain direction at room temperature (insets showing local magnification, scale bar 20 nm): e, f TEM and AFM of (PUDG)₄SiW and g, h TEM and AFM of (PULG)₄SiW. i–l HR-TEM and HAADF-STEM images of HSP chains with different alignments from (PULG)₄SiW at room temperature: i, j parallel alignment and k, l discrete state. m–p HSP chains with opposite helical morphologies (insets showing local magnification, scale bar 100 nm): m, o AFM of isolated HSP chains at the edge of (b, d). n, p Height profiles corresponding to (m, o).

Accompanying solvent polarity increase from chloroform/n-hexane to chloroform and then mixed solvent of chloroform with methanol or even little water/DMSO, the initially isolated HSP chains stack into parallel alignment, then rod-like or spherical aggregates and finally helical belts (Supplementary Figs. 19, 20). Such morphologic changes are ascribed to that the increased polarity enhances the hydrophobic interaction of alkyl tails between neighboring HSP chains, and a followed highly ordered packing further evolves into the chiral self-assembly at a larger scale. Upon the concentration increasing in chloroform, more packed chains tend to be parallelly aligned while the solution changes from transparent to flocculent, finally to sol and even gel accordingly (Supplementary Fig. 21), indicating the formation of larger fibers comprising the aggregation of multiple HSP chains. Heating makes HSP chains dissociate, but centimeter-scale filaments of the packed chains regenerate upon progressive cooling (0.2 °C min⁻¹) with strength against gravity from being taken out from the solution (Supplementary Figs. 22, 23). The substitution of SiW with equivalently charged clusters such as [PW₁₁VO₄₀]⁴⁻ generates similar polymer chains (Supplementary Table 2 and Supplementary Fig. 24), indicating the irrelevance of cluster composition. However, the charges of POMs and the number of combined organic countercations have a key impact since complicated assemblies have been detected.

Nucleation-elongation process and stereospecificity

The growth of HSP chains upon aging is characterized by ex-situ AFM (Fig. 3a–d) by blotting excess sample solution of chloroform/n-hexane (2:8 in v/v) on the silicon substrate, which offers a clear size and morphologic evidence of HSP chains upon time of aging. The minimal iCOM aggregates as isolated domains with few primary linear oligomers emerge in 2 h, suggesting that the nucleation takes precedence. Upon prolonging assembly time, the spot-like aggregates decrease while the oligomer chains increase in visual field. Further culturing to 24 h, almost all initial nucleic aggregates disappear, and long linear chains become dominated instead, indicating that the linear chains grow from the initial spot-like nucleation. Owing to the high contrast of inorganic nanocluster, the array of SiWs in polymer chains visualizes the gradual transformation. Under natural evaporation, single chains grow up with a more flexible conformation (Supplementary Fig. 25) upon the subsequent appearance of initial core, circular and short oligomers, later tangled clews with chain ends, and heavily entangled HSP chains finally. The formation process of parallel-aligned HSP chains in chloroform is seen correspondingly (Supplementary Fig. 26), reflecting that the nucleation and parallel packing of chains occur simultaneously. The addition of alcohols into the mixed solvent slows down the growing speed of HSP chains (Supplementary Fig. 27). From the slowed polymerization by methanol, which shows the best result, short oligomers sprouting from initial nucleuses with 2–5 aggregated monomers are captured synchronously (Supplementary Fig. 28).

Fig. 3. Polymerization process of HSP chains from (PULG)₄SiW.

a–d AFM images of nucleation and growth of HSP chains in a mixed solvent of chloroform/n-hexane (2:8 in v/v) at 0.1 mg mL⁻¹ (18 μM) versus time. Samples are prepared by blotting excess solution along a certain direction on the silicon substrate at room temperature (insets scale bars: a–c 50 nm, d 200 nm): a 2 h, b 6 h, c 12 h, and d 24 h. e Diagrams of diffusion coefficient (D) and polymerization degree (D_p) of HSP chains against concentration in deuterated chloroform. f Diagram of hydrodynamic diameter and D_p of HSP chains versus temperature in chloroform at 0.1 mg mL⁻¹ (18 μM). g Normalized degree of aggregation with fitted (blue and red curves) nucleation and elongation regimes corresponding to the ratio of absorbances at ѵ_(N–H) = 3400 cm⁻¹ against temperature from in-situ ATR-FT-IR spectra of HSP chains in chloroform at 0.1 mg mL⁻¹ (18 μM). h Schematic growth from dispersed monomers, through nuclei (minimal aggregates) to oligomers, and final HSP chains coarse-grained (CG) model (Red beads, POM; pink beads, pyridine and methylene group; yellow beads, urea and amide group; blue beads, methylene group).

Dynamic light scattering (DLS) provides statistical evidence for supramolecular polymerization. Within the assembly time prolonging in chloroform/n-hexane (2:8 in v/v) solution, only very small aggregates containing a few monomers indicating nucleation appear in 1 h. The narrow peak in size distribution suggests a uniform majority in the early stage. After that, two distributions by intensity can be found in 4 h and later, indicating the coexistence of both initial aggregates and polymerized chains. The small size distribution has completely disappeared, and the bigger size becomes dominant after 24 h (Supplementary Fig. 29). The instantaneous size distribution in a mixed solvent (chloroform/n-hexane/methanol = 2:8:1 in v/v) shows almost identical dispersion after sonication, which is like that in an extremely dilute sample solution (Supplementary Fig. 30), suggesting the aggregation-induced nucleation. The polymerization degree (D_p) is calculated based on the diffusion coefficient (D) of (PULG)₄SiW via 2D diffusion ordered spectroscopy (DOSY) NMR spectra⁴⁴. The degressive D values upon increasing concentration in deuterated chloroform (Supplementary Figs. 31–33) illustrate a typical concentration-dependence of the supramolecular polymerization⁴⁵. The turning point in NMR plot with a more exhaustive concentration gradient reveals a two-step regime (Fig. 3e), corresponding to the nucleation and elongation from minimal aggregates to matured HSP chains observed above. A contrary tendency of D_p against concentration is calculated by the Stokes-Einstein equation, indicating that the initial aggregates containing 3–5 monomers grow up to the oriented oligomer chain. Moreover, the plots of the hydrodynamic diameter of aggregates and D_p against cooling temperature in chloroform show nonsigmoidal curves via variable-temperature DLS, disclosing an analogous two-stage process from highly discrete monomers to predominant HSP chains (Fig. 3f). Under this condition, the obtained SP chains enduring a cooling period give a D_p of ca. 730 with an average M_w about 3.9 × 10⁶ Da while the chain structure and morphology sustain for months.

Considering the non-angle-dependent ionic interaction of organic cations, the oriented packing of iCOMs can be ascribed to the directional H-bonds. Based on the FT-IR spectrum of iCOM (Supplementary Fig. 34) at discrete state obtained from the mixed solvent (chloroform/n-hexane/methanol = 2:8:1 in v/v), the corresponding red-shifts of stretching vibrations of amino group ѵ_(N–H) and carbonyl group (amide-I band) ѵ_(C=O) to 3318 cm⁻¹ and 1635 cm⁻¹, and blue-shifts of C–N bond (amide-III) ѵ_(C–N) and bending mode of N–H bond (amide-II) δ_(N–H) to 1259 cm⁻¹ and 1550 cm⁻¹ of HSP chains in chloroform and its mixed solvents with n-hexane demonstrate the formation of intermolecular H-bonds^19,46. In addition, the delicate variable-temperature attenuated total reflectance (ATR)-FT-IR spectra of HSP chains in solid membrane and the chloroform solution reveal the vectoring of H-bonds to the polymerization against increasing temperature due to the vibration bands weakening and shifting, strongly pointing out the dissociation of H-bands and manifesting that the H-bonding drives the elongation of HSP chains (Supplementary Fig. 35). The normalized degree of aggregation with fitted nucleation and elongation regimes is applied with the modified helical assembly model⁴⁷. The corresponding in-situ changed ratios of absorbances at ѵ_(N–H) = 3400 cm⁻¹ against temperature⁴⁸ reveal an obvious transition state at elongation temperature T_e = 322.7 K (Fig. 3g), consistent with the size change in DLS test above. The enthalpy h_e is −65.3 kJ mol⁻¹, and the equilibrium constant from the inactive to the activated state K_a is 4.1 × 10⁻⁴, indicating a high degree of cooperativity in the nucleation-elongation process. In this model, the number-averaged D_p averaged over all active species at T_e, <N_n(T_e)>, is ca. 13, close to the calculated results by NMR and DLS data above. Upon further cooling to R.T., the <N_n(298 K)> increases to ca. 272, showing the same growing trend. The hydrophobic effect and electrostatic force between components assist the preferable conformation of iCOM toward linear growth (Fig. 3h). The amphiphilic feature drives the random aggregation for nucleation at the initial stage. The later phase separation contributes to the orientated array of hydrophilic POM clusters along the chain axis since the hydrophobic alkyl chains are observed to distribute outside in weak polar solutions from negatively stained TEM image (Supplementary Fig. 36). The much higher wavenumbers of ѵ_asym (–CH₂–) at 2922 cm⁻¹ and ѵ_sym (–CH₂–) at 2852 cm⁻¹ point out the less ordered state of the alkyl chains, showing no additional contribution to the orientated growth¹⁹ except helping in shielding hydrophilic cores. For each homogeneous enantiomeric cation alone, the time-dependent circular dichroism (CD) spectra show mirror positive and negative signals at about 240 and 280 nm (Fig. 4a). In contrast, only a whole positive or negative Cotton signal of enantiomeric iCOMs appears as a broad band at around 274 nm (Fig. 4b), which is ascribed to the combination of organic component and induced dichroism of SiW in HSP chains (Supplementary Figs. 37, 38). Considering SiW bears an absorption band at the same region as organic cation, the induced chirality is verified by replacing SiW with a reduced cluster [PMo₁₂O₄₀]⁴⁻ (rPMo) with the same architecture and charges but different absorption band at 730 nm (Supplementary Table 3 and Supplementary Fig. 39)⁴⁹. Importantly, the intensity of Cotton signals for both (PULG)₄SiW and (PULG)₄rPMo gradually enhances by prolonging assembly time, proving the chiral amplification effect upon the homochiral polymerization⁵⁰. As a result, the supramolecular polymerization can be concluded to happen under a directional interaction of successive intermolecular H-bonds like those observed in π-scaffolds for the helical packing⁵¹.

Fig. 4. Chiral effect and self-sorting of HSP chains during supramolecular polymerization.

a CD spectra of enantiomeric PULG·Cl and PUDG·Cl in chloroform. b CD spectra of HSP chains upon the polymerization of (PUL/DG)₄SiW in chloroform (0.1 mg mL⁻¹) versus time (0.5, 1.0, 2.0, 4.0, 6.0, and 8.0 h). c Variable-time DLS plots of HSP chains in chloroform at the concentration (1.0 mg mL⁻¹) for (PULG)₄SiW and (PUDG)₄SiW, and identical/twice concentration of their racemate in total (half and identical for each). d–f AFM images of HSP chains at the same condition with c in chloroform with insets corresponding macroscopic images: d (PULG)₄SiW, e (PUDG)₄SiW and f their racemate. g Magnified AFM image of HSP chains in a diluted racemate solution (0.01 mg mL⁻¹).

These results reveal that molecular chirality and the conformation of iCOM play a decisive role in the stereospecific polymerization process. To evaluate the stereospecificity of active position, the hydrodynamic diameter of HSP chains is monitored by DLS versus real-time. Under the same concentration, the observed size changes of two enantiomeric monomers by real-time DLS referring to polymerization display similar tendency with time while their mixture at an equivalent ratio does not reach the same polymerizing scale and smaller hydrodynamic sizes appear instead (Fig. 4c). For the total concentration of racemate equivalent to above sole enantiomers, the formed much shorter HSP chains indicate that the concentrations effect of two enantiomers are independent from each other (Fig. 4d–f). Meanwhile, large amounts of opaque floccule generate in two sole enantiomeric iCOM solutions while their racemate solution retains clear at the same aging time (insets of Fig. 4d–f). According to the chiral recognition effect⁵², such a difference between racemate and sole enantiomers during supramolecular polymerization is just in accord with the stereo-adaptivity-induced kinetic reduction in the nucleation-elongation process²⁶. The opposite-handed chains sourcing from two enantiomeric iCOM monomers in diluted racemate solution, visualized from AFM images, prove the strong homochiral recognition restriction in the polymerization (Fig. 4g). This chiral self-sorting indicates that the monomer adopts a unique spatial conformation accompanying supramolecular polymerization.

Mechanism analysis and MD simulation

The stereospecific chain growth is further analyzed from energetic consideration through MD simulation. Here, a coarse-grained (CG) MD simulation is employed for the reaffirmation of the chain-growth mechanism through the capture of the nucleation-elongation transition in the HSP evolution process (Supplementary Figs. 40, 41). The snapshots of the simulation box in prolonged time distinctly illustrate the growth from discrete monomers through nuclei (minimal aggregates) to oligomers and final HSP chains (Figs. 3h, 5a). The potential energy of the system against simulation time reveals the final stable state of HSPs, and the increasing trend of the averaged number of iCOMs in HSPs throughout the process statistically indicates the different stages of nucleation and elongation (Supplementary Figs. 42, 43). For further inquiry on the conformation and packing mode of the iCOMs in HSP structure, as well as the H-bonding mode between the monomers within the polymerization, three preferential conformations, (i) tetrahedral, (ii) planar, and (iii) shuttlecock-shaped states, are proposed as the possible monomer status in the all-atomic MD simulation. The conformation of the flexible organic part is optimized into the final steady state in a pre-aligned straight HSP chain model through 2000 ps simulations (Supplementary Figs. 44–46). The number density of cations and chloroform around POM surface suggests the preference of organic component at the periphery. The interaction energy evolutions of POM-countercation/-chloroform, organic moiety, and organic moiety-chloroform, individually, achieve the structure optimization driven by Coulomb interaction, and Lennard-Jones potential reaches the stable states with energy minimization. The density distribution curves by planes along different directions depict the continuous chain structure consistent with the observed morphologies and each proposed model of conformation i, ii, and iii. The conformation iii bears the lowest calculated total interaction energy, in accord with the quantitative possession on maximum number and highest proportion (82.8%) of intermonomeric H-bonds among three models (Fig. 5b, Supplementary Table 4 and Supplementary Figs. 47–49). Therefore, the conformation iii provides the best combination energy of supramolecular polymerization and facilitates the consistent stereospecific helicity along the HSP chain. The statistical angle (θ) by the vector of HSP chain axis and molecular long-axis of countercation is in a range of 46–65°, depicting a shuttlecock shape of iCOM as a stable conformation packing into linear assembly (Fig. 5c and Supplementary Table 5).

Fig. 5. Preferential helical model of HSP chain and molecular dynamics (MD) simulation.

a The snapshots of locus diagram against time for supramolecular polymerization of iCOM monomers in CG-MD simulation: (i) 0 ns (frame 0), (ii) 1 ns (frame 10), (iii) 5 ns (frame 50), (iv) 10 ns (frame 100), (v) 100 ns (frame 280) and (vi) 300 ns (frame 482). b Intermolecular H-bonds between the iCOM monomers in HSP chains in the all-atomic MD simulated model. c Simplified helical model of HSP chains and averaged twisted angle between monomers. d SAXS curves of HSP chains at different states.

At conformation iii, the averaged rotation angle (α) between two adjacent monomers is ca. 3.2° based on one pitch ca. 40.2 nm from AFM sections. The detected shrinking for height/width of HSP chain can be explained by both alkyl tail disorder and angle constraint of the shuttlecock shape of monomers in packing. The calculated average distance between parallel-aligned chains from the electron density profile of SiW array (Supplementary Fig. 50) is ca. 4.0 nm, which is far from the full length of iCOM but fits well to the shuttlecock-shaped model. The alignment of iCOMs in HSP chains which performs sensitively to the solvent polarity is further identified by small angle X-ray scattering (SAXS) with cross-interchain distance d_interchain and intercluster distance d_intercluster along the chain as two decisive parameters of the structural model (Fig. 5d). In chloroform, the d_interchain ca. 3.9 nm from the peak of q_interchain is in good agreement with the calculated value of tight stacking chains by TEM. While in a weaker polar solvent of chloroform/n-hexane (2:8 in v/v), much more decreased height (ca. 3.3 nm) of HSPs (Fig. 2f, h) with almost vanished helical morphology verifies the proposed model upon countercations shrinking to minish surface energy. Meanwhile, the nonpolar condition weakens the hydrophobic interaction between abreast chains, which explains the dissociation of parallel-aligned HSP chains into an entirely monodispersed state. In a solid state, the parallel alignment of chains turns to be more regular and tighter with a narrower and stronger intensity of q_interchain. The calculated d_intercluster from q_intercluster maintains almost the constant value of ca. 1.4 nm.

HSP membrane for enantiomer separation

The flexibility and chiral surface of ultra-long helical chain provides the processibility resembling conventional polymers in membrane separations. Several preparation methods have been tried (Supplementary Fig. 51 and Fig. 6a). A free-standing fibrillar membrane with flexibility and toughness is prepared via upper volatilization from two phases, but it is not suitable for direct utilization. Although the jelly state at the increased concentration can be used for membrane preparation on solid substrate, the generated wrinkles after tearing off become difficult to eliminate later. Alternatively, the suction filtration method is adopted by filtrating the solution of parallelly aligned HSP chains through polytetrafluoroethylene (PTFE) substrate with a pore size of 0.22 μm for direct use, which ensures the scalability and intactness of the nanofibrillar membrane according to the examination of polarized optical microscopy, AFM, and scanning electron microscopic (SEM) (Supplementary Fig. 52). The membrane thickness is controlled by adjusting the sample volume and concentration. Typically, the membrane-based separation exhibits advantages of high capacity, scalability, and continuous processing⁵³. The permeable membranes for enantioseparation of small biological molecules and drugs perform a broad prospect, which are mainly divided into polymer-based ones like polymers with chiral main chain or side chain^54–57, addition of chiral selectors^58–63, and molecularly imprinted membranes⁶⁴; polycrystalline like MOF-based membranes^65,66; hybrid ones such as mixed matrix membranes^67–71 and 1D/2D carbon materials^72–76. Considering the chiral surface feature of every single HSP chain sourcing from the helical array of iCOMs, the prepared fibrous membranes enable the enantiomer separation efficiently, which is different from those chiral polymers⁷⁷ or common polymeric composites mixed with chiral additives⁷⁸. As indicated in UV–vis spectra, the membranes sustain a satisfied stability in acetonitrile for a week (Supplementary Fig. 53).

Fig. 6. Permeable membrane preparation and enantiomer separation.

a Schematic preparation routes of permeable membranes. b HPLC profiles of racemic permeants (left) and permeation solution (right) in enantioseparation of L-type membrane for histidine derivatives via concentration-driven permeation. c ¹H-NMR for binding sites of Fmoc-His-OMe on membrane through adsorption in solution (CD₃CN:CDCl₃ = 1:1 in v/v, 500 MHz, 25 °C). d Schematic enantioseparation process via the chiral channel sourcing from contorted grooves by helically close-packed HSP chains.

A racemic mixture of Fmoc-protected histidine ester derivatives (Fmoc-L/D-His-OMe) in acetonitrile with a concentration of 0.05 mg mL⁻¹ is separated as feed solution in a cell device with two chambers partitioned by the membrane of (PULG)₄SiW (L-type membrane) at a thickness of ca. 60 μm (Supplementary Figs. 54–57). The profile of blank solution from real-time chiral high-performance liquid chromatography (HPLC) demonstrates the superior enantioselectivity of Fmoc-L-His-OMe with e.e. value 82.1 ± 1.4% (mean averaged by 3 parallel experiments) and a flux 5.3 ± 0.4 mmol m⁻² h⁻¹ in initial 30 min (Fig. 6b and Supplementary Fig. 58). For a continuous separation, the superimposed separation raises the enantioselectivity⁷⁹ to e.e. 98% after 4 cycles (Supplementary Fig. 59). Correspondingly, the D-type membrane exhibits a very close enantioseparation efficiency (Supplementary Fig. 60). These results confirm that the HSP chains perform characteristics of chiral polymers by spontaneous formation of micro- even macro- scale fibers. Several factors, such as permeation time, membrane thickness, concentration of feed solution, and membrane working area, affect the separation efficiency. Upon prolonging permeation time, the enantioselectivity declines gradually and the flux of substrate Fmoc-D-His-OMe from the feed solution to the blank solution significantly decreases after 60 min due to the decreased concentration gradient in two chambers (Supplementary Fig. 61). In contrast, the permeation of Fmoc-L-His-OMe maintains at a relatively low but constant level, indicating a less influence of the mirror enantiomer Fmoc-D-His-OMe. The membrane thickness change in the region of 20‒80 μm does not show much impact on the e.e. value, suggesting the satisfied condition for an optimized efficiency. The 75% decrease in membrane thickness only causes about 10% of e.e. value decrease, but the permeation flux goes up drastically (Supplementary Fig. 62). On the other hand, the increase in concentration causes a less than 10% decrease in the enantioselectivity while the flux performs a large ascending (Supplementary Fig. 63), which is in accord to the general permeation regularity^53,80. The results of the changed working area of the membrane show that the enantioselective efficiency almost does not change in the separation (Supplementary Fig. 64).

Selective adsorption experiments are performed by soaking L-/D-type membranes in racemic solution. The remaining solution and eluent from leaching the membrane are confirmed almost equivalent adsorption for Fmoc-D/L-His-OMe substrates (Supplementary Fig. 65). Therefore, the transmembrane diffusion is proposed to follow the facilitated-transport (diffusion-enantioselective) mechanism⁵⁵. The binding site in membranes is analyzed by comparing ¹H NMR spectra of two enantiomers after mixing with isolated organic cations and SiWs (Fig. 6c)⁸¹. When PULG·Cl is used, both enantiomers show little change of chemical shift, indicating no observable binding position on the organic component alone. Since inorganic clusters do not dissolve in the used solvent, dioctadecyldimethylammonium bromide (DODA)-combined SiW is used instead. The chemical shifts of polar groups in two substrates move nonenantioselectively when mixing with (DODA)₄SiW, suggesting the interaction between the polar region of substrates and SiW where POM serves as the jumping site for a favorable diffusion process. Thus, the high enantioselectivity can be attributed to the generated discrepancy of diffusion kinetics in transmembrane due to the different matching degree of chiral substrates to the interior asymmetric pores deriving from contorted grooves of close-packed HSPs (Fig. 6d). The synergy of appropriate volume (size effect) with stereoselective to match the asymmetric channels and the groups amiable to polar clusters are further verified to be critical factors by tracing more chiral substrates (Supplementary Table 6). The substrates like methyl phenyl sulfoxide that are much smaller than the averaged pore size of the channel (see below) in membranes show nearly no enantioselectivity even though the polar groups performing amiable to cluster cores have grafted. Upon the increase of molecular sizes in both dimensions closing to Fmoc-D/L-His-OMe, the separation efficiency gets increased, and bifendatatum receives a separation efficiency of e.e. value ca. 41.2%. Meanwhile, for the substrates with similar sizes to Fmoc-D/L-His-OMe yet bear stronger or weaker polar groups, the separation efficiency decreases greatly. The substrates with the same structure of Fmoc-D/L-His-OMe but in the state of hydrochlorate or acid show greatly decreased separation, indicating degressive efficiency. Similarly, by replacing imidazole with a phenyl group, the separation efficiency also becomes low, suggesting the necessity of a polar group. For the substrates having both longer size and strong polar groups, no separation is observed. We speculate that the changed diffusion rate deriving from weak H-bond and polar interaction from oxygen-enriched/electronegative surface of POMs may narrow the distinction in diffusivity of two enantiomeric permeants, resulting in a lower enantioselectivity (Supplementary Fig. 66).

The helical structure of single chains in packing state demonstrates the superiority in the enantioseparation of scale-matched permeants since the membranes prepared from unassembled monomers or spheric assembly show poor e.e. values (Supplementary Fig. 67). Meanwhile, the parallel alignment of HSP chains plays a critical role since the membrane prepared from entangled chains exhibits a relatively lower e.e. value (60.2%) under the same condition. An average pore size of 1.3 nm and more uneven pore size distribution in the membrane of entangled chains affect enantioseparation performance according to the microporosity of two nanofibrillar membranes from CO₂ sorption isotherm experiment at 273 K by nonlocal density functional theory method (Supplementary Fig. 68). The intactness and stability of permeable membranes are characterized by morphological observations and SAXS before and after permeation (Supplementary Fig. 69). The dynamic feature from noncovalent bonds allows the assembled membranes easily recycled by polar good solvents and reprepared into new membranes to implement more permeations with relatively constant enantioselectivity (Supplementary Fig. 70). The systematic comparisons of the present results with those published results (Supplementary Table 7) through detailed measurements reveal the synergy of pore size and asymmetric channel for the selected substrates (Supplementary Figs. 71, 72). Moreover, the small change of molecular structure, such as the distance prolonging of the urea group from the POM, improves the surface properties of HSP chains and results in the changes of enantioseparation (Supplementary Figs. 73–76 and Supplementary Tables 8–10). As a result, the inversion of separation selectivity may originate from the space-effect-induced orientation and angle changes of iCOM packings.

Discussion

In this research, polyanionic nanoclusters have been covered by amphiphilic organic cations to get a kind of ionic complexes via ionic interaction. Different from the nanoclusters that are encapsulated with simple organic cations, these nanocluster complexes serving as the particle-like monomers polymerize into supramolecular polymers with ultra-long supramolecular chains structure. Through adjusting solvent polarity, the formed supramolecular polymers change their existing state either in monodispersed single chain or parallelly aligned state. The dynamic process of supramolecular polymerization shows unexpected nucleation-elongation growth that is known for small molecules but has not yet dealt with nanoparticle monomers. The charge delocalization at the surface of POM provides the structural basis for the configuration adaptability of iCOM monomers, and aggregation-induced anisotropic phase separation of amphiphilic multicomponent complexes in solution brings about the asymmetry into the system. For the iCOM monomers, the mobility of organic component makes this part phase-separated surrounding the surface of nanocluster driven by hydrophilic/hydrophobic effect, which triggers the nucleation by forming small aggregates. As a result, the aggregation enables the H-bonding interaction between iCOMs, which causes the prevailing conformation at a shuttlecock-like state as the activated position for initiating polymer chain growth. The speculation is well supported by the experiment data and computational simulation. Due to the chiral site in organic component, the tight-packing-caused spatial restriction drives the combination of homochiral monomers, leading to the chirality amplification in the growth of HSPs via H-bonds and to the self-sorting in racemate solution. The packed chain structure and helical morphology of such cluster-based supramolecular polymer generate unique material properties. The permeable membranes composed of parallel-aligned helical chains are applicable for enantioselective and substrate-selective separation, like Fmoc-protected histidine ester derivatives Fmoc-His-OMe bearing ca. 98% of e.e. after 4 cycles. Both helical surface of supramolecular chains and nanocluster core and organic component are proposed to contribute to the precise stereoselectivity separation. As a natural prolonging of the present research, it can be envisioned that other nanoparticular monomers, like metal nanoparticles, should have similar characteristic and synergistic functions only if the structure and property of modified components match with particle cores. We believe that this methodology of nanoparticular chiral supramolecular polymers extends the functionalization of macroscopic soft materials.

Methods

Materials

All the commercial chemicals were used as received. The synthesis procedures of chiral organic cations (PULG·Cl and PUDG·Cl) are described in Supplementary Fig. 1. The enantiomeric monomer iCOM with the chemical composition of (PULG)₄SiW and (PUDG)₄SiW are described in Supplementary Fig. 9.

Preparation and characterization of supramolecular polymers

In a typical synthesis, 5 mg (0.9 mM) of enantiomeric iCOM monomer was dispersed in 5 mL chloroform in an enclosed sample vial, and the mixture solution was heated until the solid sample was fully dissolved. The solution was then cooled to room temperature, aging for hours to form supramolecular polymers. For the measurement, the solution was dropped onto the solid substrate by a pipette at room temperature with oriented blotting or natural evaporation to obtain parallel-aligned HSP chains. When the sample solution stood for days, the cross-linked flocculent precipitations occurred. As for the mixed solvents, the iCOM monomers were first dissolved in chloroform, and then the calculated amounts of other solvents were added. The mixed solution was stood by need. The different statuses of the HSP chains were obtained following similar procedures.

Preparation of permeable membrane

The permeable membranes for structural characterizations and separation experiments were prepared through a simple filtration of the HSP chains in a cross-linked flocculent state onto a PTFE support by controlling the sample concentration and volume. The free-standing membranes were obtained through natural volatilization of the sample solution on the supporting phase. The other tentative preparation methods are shown in Supplementary Fig. 51.

Experiment of enantioselective permeation

The above-obtained permeable membrane with a diameter of 2.3 cm was clamped in the diaphragm position of an electrolytic cell. A certain quantity of racemic permeants dissolving in acetonitrile was placed in one chamber as the feed solution, while the blank solvent was placed in another chamber. The permeation solution is monitored by the chiral HPLC at the set intervals and the enantioselectivity and flux corresponding to the enantioseparation were then calculated.

Measurement of NMR and mass spectra

The ¹H-NMR, 2D DOSY NMR spectra were performed on a Bruker Avance 500 MHz spectrometer (Germany) with methanol-d₄, DMSO-d₆, chloroform-d_, or acetonitrile-d₃ as the solvents. Tetramethylsilane (TMS) was used as an internal reference (s = singlet, br = broad, d = doublet, t = triplet, q = quartet, m = multiplet). Matrix-assisted laser desorption/ionization (MALDI) time of flight (TOF) mass spectrometry was recorded on a Bruker Autoflex^TM speed TOF/TOF equipped with a nitrogen laser (337 nm, 3 ns pulse). Trans-2-[3-(4-tert-Butylphenyl)-2-methyl-2-propenylidene] malononitrile (DCTB) was used as a matrix. The m/z range during datum acquisition was recorded from 200‒1000 Da for reflection positive mode and 5‒20 K Da for linear positive mode.

Measurement of FT-IR and ATR-FT-IR spectra

FT-IR spectra (KBr pellet) of samples in varied mixed solvents were collected on a Bruker Vertex 80 V spectrometer equipped with a deuterated triglycine sulfate (DTGS) detector (32 scans) in a resolution of 4 cm⁻¹. Variable-temperature ATR-FT-IR spectra of samples in membrane or solution state were collected on a Bruker Vertex 80V-ATR spectrometer with the germanium (Ge) ATR module.

Elemental and TG analysis

Organic elemental analysis (C, H, N) was carried out on a Vario microcube (Elementar). Thermal gravimetric analysis (TGA) was recorded on a PerkinElmer Diamond TG/differential thermal analysis instrument using high-purity nitrogen as the carrier gas with a heating rate of 10 °C min⁻¹ in the temperature range of 28–800 °C.

TEM, AFM, and SEM characterizations

TEM, HR-TEM, and HAADF-STEM characterizations were carried out on a JEOL JEM-2100F field emission transmission electron microscope at an accelerating voltage of 200 KV with/without negatively staining by uranyl acetate aqueous solution (0.2–1.0 wt%). AFM images were recorded on a Dimension FastScan^TM atomic force microscope on a silicon wafer as substrate in the air under ambient conditions with the tapping mode. Scanning electron microscopic (SEM) measurements were performed on a JEOL JSM-6700F (Japan) field emission scanning electron microscope.

DLS plots

DLS curves were performed on a Zetasizer NanoZS instrument. Variable-temperature measurements were finished at a cooling rate of 1 °C 5 min⁻¹.

UV–vis and CD spectra

UV–vis spectra were carried out on a Varian CARY 50 Probe spectrometer. CD spectra were performed on a Bio-Logic MOS-450/AF-CD and JASCO J-1700 spectropolarimeter with a step wavelength of 0.5 nm and speed of 0.5 nm s⁻¹ at 25 °C.

SAXS measurements

SAXS measurements were performed on an Anton Paar instrument SAXSess mc2 with an X-ray wavelength of 1.542 Å and carried out using a 1D detector at room temperature with an exposure time of 2 min.

Polarized optical microscopy

The optical microscope images were obtained using a Zeiss Axioskop 40 microscope.

Contact angle measurements

Static contact angle measurements were performed on an OCA20 contact angle measurement system (Data-physics) at ambient temperature.

Chiral high-performance liquid chromatography

The enantioselectivity and flux in enantiomers separation were determined by HPLC using a SHIMADZU LC-20A equipped with a chiral OD-H column (4.6 × 250 mm) sourcing from Daicel Chemical Industries Ltd by using a 2-propanol/n-hexane mixture as mobile phase. The eluent of Fmoc-His-OMe separation analysis was set at a ratio of 2-propanol/n-hexane = 4/6 in v/v (conditions: flow rate 1 mL min⁻¹, detector 275 nm). The separation conditions of other enantiomeric permeants are summarized in the corresponding Supplementary Information.

MD simulations

Simulations were performed using the GROMACS software package. The force fields and corresponding method parameters of the CG-MD and all-atomic MD simulations are supplemented in the Supplementary Information.

Author contributions

L.X.W. and M.H.L. conceived and designed the project. Z.X.Z. synthesized and characterized the materials, and performed the analysis with G.H.Z. on structures. Z.X.Z., L.X.W. and B.L. wrote and revised the final manuscript.

Peer review

Peer review information

Nature Communications thanks Fatima Garcia 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 experimental procedures and supplementary data generated in this study are provided in the Supplementary Information, and the Source Dataset is accessible from 10.6084/m9.figshare.25139810. All relevant data are available from the corresponding author upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-52402-6.