Self-assembly of emissive supramolecular rosettes with increasing complexity using multitopic terpyridine ligands

Coordination-driven self-assembly has emerged as a powerful bottom-up approach to construct various supramolecular architectures with increasing complexity and functionality. Tetraphenylethylene (TPE) has been incorporated into metallo-supramolecules to build luminescent materials based on aggregation-induced emission. We herein report three generations of ligands with full conjugation of TPE with 2,2′:6′,2″-terpyridine (TPY) to construct emissive materials. Due to the bulky size of TPY substituents, the intramolecular rotations of ligands are partially restricted even in dilute solution, thus leading to emission in both solution and aggregation states. Furthermore, TPE-TPY ligands are assembled with Cd(II) to introduce additional restriction of intramolecular rotation and immobilize fluorophores into rosette-like metallo-supramolecules ranging from generation 1–3 (G1−G3). More importantly, the fluorescent behavior of TPE-TPY ligands is preserved in these rosettes, which display tunable emissive properties with respect to different generations, particularly, pure white-light emission for G2.


Re: Response to reviewers' comments for manuscript NCOMMS-17-19577A-Z
We would like to thank all the reviewers for their critical comments and thoughtful suggestions on the manuscript. Enclosed please kindly find the revised manuscript in which we have made our revisions and corrections. The questions, suggestions and comments raised by the reviewers have been addressed in this letter. The detailed summary is as follows: Reviewer #1: 1) The total species in the chelate solution for G1-G3. The authors did not mention anywhere in the manuscript the association constants in various solvents and how might changing solvent composition affect these constants. I almost have the impression that these rosettes are the sole species in solutions, is that true? But I don't think the authors specified in the manuscript. If this is the case, the authors should provide a simple 2D-fluorescence measurement that indicates independence of emission from excitation for G1-G3 and how solvent composition influences the results. Also, a temperature dependent experiment may also help provided that the rosettes dissociate at higher temperatures. Response: In the self-assembly of G2 and G3, hexameric and heptameric rosettes are the sole species in solution. In the case of G1, the self-assembly generated a mixture of macrocycles. Due to the multivalent interactions of multi-armed ligands in G2 and G3, we used G1 as a model system to study the association constants  in different solvents, including THF, CHCl3/MeOH, DMF and DMSO. For example, as shown in Figure R1, the association constants of G1 in CHCl3/MeOH (v/v, 1/3) is 2.9 ± 0.3 × 10 5 M -1. The solvent composition does affect the constants, its value would decreased when increasing the polarity of solvents. In THF, the association constants of G1 is 1.1± 0.2 × 10 5 M -1 , which is close to the value in CHCl3/MeOH (v/v, 1/3). The association constants of G1 were decreased by nearly an order of magnitude in DMF (6.3 ± 1.4 × 10 4 M -1 ) and DMSO (4.0 ± 0.5 × 10 4 M -1 ). We performed 2D-fluorescence study as suggested by reviewer. For instance, as shown in Fig. R2, the luminous range of 2D-fluorescence of G2 was consistent with 1D fluorescence. No other light-emitting specie was observed. The 2D-fluorescence results of G1 (Figs. S67 and S68) and G3 (Figs. S71 and S72) also show the independence of emission from excitation. As such, the supramolecules did not dissociate at the concentration of fluorescence study. Figure R3.Temperature dependence of fluorescence emission spectra of G3 in CH3CN (λex = 320 nm, c = 1.0 µM).
Finally, the temperature dependent study was conducted following reviewer's suggestion. The temperature dependent fluorescence study showed these rosettes are stable at the temperature from -46℃ to 50℃. As for G1 (Fig. S75) and G3 (Figs. R3 and S77), the emission intensity was increased when decreasing the temperature of the system which is consistent with traditional fluorescence process. As for G2 (Fig. S76), the local emission of TPE was decreased while the CT emission was increased when the temperature decreased from 50°C to -46°C, because the intramolecular rotation process is restricted and the electronic communication between TPE and metal becomes dominant (Chem. Sci., 2015, 6, 5347, page 7). Furthermore, variable temperature 1 H NMR spectra of G2 and G3 also show no dissociation of supramolecules (Figs. S53-54). Detailed discussion was added in P13 Paragraph 2 in the manuscript.
2) The interpretation of certain fluorescence results, for instance, the assignment of the shoulder emission at longer wavelength throughout the samples. The authors attribute these peaks to charge-transfer fluorescence while providing no evidence to support it. For example, charge-transfer absorption and fluorescence are typically very sensitive solvent polarity and media rigidity? This should be very easy to prove with just a bit more effort in experiment. Also, the authors explained the enhancement of G1-G3 fluorescence in comparison to L1-L3 in good solvent as inhibition of molecular motions. This is hardly convincing given that the enhancement is not as dramatic. Metal coordination can have a tremendous effect in the electronic transitions of the ligands. I suggest that the author show hard evidence for all the explanations they list in the fluorescence properties section as they did in the characterization part at the beginning of the manuscript, which is nice and clear.  4) P10 L171 "rear" --> "rare"? Response: Revised it as "rare".

5) What is the potential use of white emission in solution since the authors seem to emphasize the concept a lot?
Response: In the past few years, white organic light-emitting materials, devices, and processes attracted great attention due to their fundamental importance and practical application. However, most light-emitting materials reported so far were based on the combination of multi-components with emission color covering the entire visible range. Single-component white light emitters are expected to exhibit superior performance improved stability, good reproducibility, and simple device fabrication procedure but without phase segregation and color aging compared to these combined emitters (B. Z. Tang et.al. Nat Commun. 2017, 8, 416). We herein provide an alternative strategy to construct a single-component white light emitter in this study.

5) I am just curious what the purpose is for choosing Cd (II) over other metals such as Zn(II)?
Response: 2,2′:6′,2″-Terpyridine (tpy) ligands are widely used as building blocks in supramolecular and macromolecular chemistry. An extremely interesting aspect of this tridentate ligand is the different binding strengths of tpy and transition metal ions which order as Ru(II) > Fe(II) > Ni(II) > Zn(II) > Cd(II) in the complex of [M(tpy)2] 2+ . So the reversibility of Zn(II) is not as good as Cd(II). ESI-MS showed many byproducts were obtained in addition to target assemblies when ligands self-assembled with Zn(II). By contrast, we are able to obtain discrete assemblies G2 and G3 using Cd(II).  Response: In order to elucidate luminescence process, the lifetime measurements have been repeated in degassed solutions but nearly no influence of oxygen was observed. The lifetimes still in the ns scale, which indicates the lifetimes are not sensitive to oxygen (Fig. R5). It indicates the luminescence is fluorescence process. The additional experimental results were depicted in Figure S85.

6) Measured lifetimes in the ns range do not necessarily mean fluorescence for G1-G3. If the quantum yield is only half a percent in air, then the intrinsic lifetimes are in the sub micron scale, which could involve metal in
Reviewer #2: 1) Since the structures of ligands L1-L3 are very complicated, the detailed 1H NMR assignments should be done more carefully. For example, the two sets of tpy signals should be distinguishable by NOE experiments. In addition, to ensure high purity of ligands, it is suggested the authors provide the whole spectrum of L1-L3 instead of the isotope patterns, or their elemental analysis.
Response: According to reviewer's suggestion, L1 and L2 have been further characterized by NOESY (L1) (Figs. S13 and S14) and ROESY(L2) (Figs. S21 and 22). All proton include tpy signals of L1-L3 have been well assigned carefully characterized by 1 H, 2D COSY, 2D ROESY and NOESY NMR. In addition, we have provided the whole MALDI-TOF spectra of L1-L3 (Figs. S15, S23 and S31) in the SI as suggested by reviewer.  2) The complexation of L1 Fig. R8. At lower concentration, we did not observe the formation of heptamer due to entropy-driven self-assembly in ESI-MS. However, no dimer was detected with the concentration from 0.5 mg/mL to 3.0 mg/mL perhaps because of the large ring constraint for dimer. We also conducted variant-concentration fluorescence study of   Response: Because of the labile interaction of metal-ligand coordination, we had to use very mild electrospray ionization condition to maintain the integrity of supramolecules, including low desolvation temperature and low ESI cone voltage. Therefore, supramolecules were readily to form solvent adducts or salts adducts in ESI-MS. Fujita and other groups reported similar results, particularly in the characterization of large metallo-supramolecules using cold-spray ionization mass spectrometry (J. Mass Spectrom. 2003, 38, 473;Nat. Chem. 2012, 4, 330;Angew. Chem. Int. Ed 2016, 55, 445). Furthermore, ESI-MS still cannot totally prevent the dissociation of metallo-supramolecules althought it is a soft ionization technique. The dissocation could induce the formation of the minor peaks in Figure 3a. We conducted varianttemperature NMR experiments (from 293K to 333K) according to reviewer's suggestion. As shown in Fig.  R10, the peaks in 1 H NMR of G2 were getting much shaper when increasing temperature. It indicates that the broad 1 H NMR peaks were attributed to slow tumbling motions of rigid skeleton of rosettes. Some signals of impurities at 8.8 and 8.4 ppm and from 2.4 to 4 ppm in the NMR spectrum of G2 come from introducing solvents incautiously. We have re-performed all NMR ( 1 H NMR, 13 C NMR, COSY, DOSY and NOESY) of G2 again which are pure without impurities. All the spectra were updated in manuscript and SI. Also we have recalculated and revised the yield of G2 and the "quantitative yield" has been removed from text following your suggestion. The additional experimental results have been added in Supporting Information.

4)
In the case of G3, again the broad 1H NMR peaks and only a few signals in the 13C NMR spectrum were observed. There are even more noisy signals shown in the ESI-MS spectrum. On the basis of those data, the structure of the uncommon heptameric macrocycle cannot be established unambiguously. At least, the whole ESI-MS spectrum of G3 should be provided and carefully assigned to exclude the possibility of forming other species. Figure R11: Variable temperature 1 H NMR spectra (500 MHz) of G3 in CD3CN (from 293 K to 333 K)

Response:
We have carried out variant-temperature NMR experiments of G3 (from 293K to 333K). As shown in Fig. R11, the peaks in 1 H NMR of G3 were getting much shaper when increasing temperature as G2. It indicates that the broad 1 H NMR peaks were attributed to slow tumbling motions of rigid skeleton. It's a challenge to achieve high quality of 13 C NMR spectra of G2 and G3 with high molecular weight (14,498 Da and 27,599 Da, respectively) due to the poor solubility and sensitivity. Generally, 13 C DEPT NMR has better sensitivity than normal 13 C NMR. To obtain better quality of 13 C NMR spectra of G2 and G3, we have performed 13 C DEPT 45° NMR of G2 and G3. For example, as shown in Fig. R12, all aliphatic carbon signals were collected and several aromatic carbon signals of G3 were also observed.
As mentioned in the ESI-MS of G2, we typically used very mild ionization condition for metallosupramolecules to minimize the fragmentation. However, due to the high molecular weight of G3, we had to increase the ionization voltage to improve the ionization efficiency and increase the sensitivity. Such operation created more fragments in ESI-MS spectrum. The whole ESI-MS spectrum of G3 (Fig. S2) has been provided in SI and the minor signals were assigned carefully. In addition, the traveling-wave ion mobility mass spectrometry (TWIM-MS) displays a narrowly distributed band of signals suggesting the formation of a rigid and discrete assembly. The additional experimental results have been added in Supporting Information. Figure R13. Powder XRD result of G2 nanotube.

Response:
We conducted electron diffraction for the tubular nanostructures but did not get satisfactory diffraction signals perhaps due to the small size of nanostructures. We then used powder x-ray diffraction to characterize G2, which gave broad signals as shown in Fig. R13. The highest peak is corresponding to 4Å spacing. Unfortunately, we were unable to get more conclusive evidence from powder x-ray diffraction. Compared to the study Aida and coworkers (Science 2014, 344, 499), we speculate that the alignment of tubular nanostructures to form more uniform sample for either electron or x-ray diffraction is the major issue we need to solve in the ongoing study.

6) In the manuscript, the emission at shorter wavelength was attributed to the LE state of TPE and the one at longer wavelength was assigned to ICT from TPE to tpy units. The authors should provide experimental
evidence to say so. In addition, since the TPE in G1 has higher flexibility, it would be expected to observe the emission at around 430 nm from TPE at higher solvent polarity due to AIE, but the result showed the longer wavelength emission was dominant. Only the dual emission for G2 was tunable. However, the detailed mechanistic discussion was not found for why the LE of TPE emission was quenched and how the molecular conformation or intermolecular interactions affect the emission. More insights into these issues should be given. Response: To better understand the fluorescence process, we have done 2D-fluorescence, variantconcentration fluorescence experiments as well as fluorescence in different solvents with various polarity. As mentioned above, the emission maxima and the shape of the emission bands of G1-G3 rely strongly on the polarity of solvents, suggesting that the emission at longer wavelength should be derived from ICT. Actually, the LE emissions of G1 (Fig. R14) and G3 (Figure 6a) do exist, although the intensity is low, due to the flexibility of the TPE backbone (G1) or relatively weaker intensity compared with ICT emission (G3). Furthermore, as shown in Figs. R15-R17, the overlapped emission spectra and the absorption spectra of G1−G3 are prone to the energy transfer (ET) process, and quench the LE emission. Benefited from the relative rigid skeleton (compared with G1) and medium ICT effect, dual emission was found in G2 system Intensity Intensity Intensity and tunable. In addition, the strongest ICT effect of G3 among three rosettes leads to the longer wavelength emission dominant.  The authors did a great job addressing my previous questions by conducting more thorough experiments. The responses are professional and adequately sufficient; I would recommend for its timely publication in Nature C ommunications.
Reviewer #2 (Remarks to the Author): The structural characterization of ligands and complexes has been improved, but there are a few comments for the authors to consider.
(1) Since a better 1H NMR spectrum of G2 was obtained, Figure 2a should be replaced by the better one.
(2) The whole ESI-MS spectrum of G3 with the identified fragment signals ( Figure S2) should be presented in Figure 3c instead. In addition, tandem MS experiments may help clarify the source of fragments, which the authors attributed to the harsh ionization conditions.
(3) Molecular simulation may provide insights into why a heptamer was formed instead of a hexamer in the complexation reaction of L3 by comparing their energy states.
(4) The association constants of G1 in various solvents were calculated by UV/vis titration experiments. However, there are many species generated from the complexation reaction between L1 and C d(II). In addition, in the presence of excess metal ions (L:M = 1:24.5), the assembled structure may further changed to a bisterpyridine -metal (1:2) adduct. The authors did not specify which equilibrium they looked into and which method and model they used for the curve fitting. The coefficients of determination were pretty bad in Figures S59 and S60. Since this is a complicated dynamic system, the calculation should be done more carefully.
I support this manuscript to be published in Nature C ommunications after the mentioned issues are properly addressed.

Re: Response to reviewers' comments for manuscript NCOMMS-17-19577B
To whom it may concern: We would like to thank all the reviewers for their critical comments and thoughtful suggestions on the manuscript. Enclosed please kindly find the revised manuscript in which we have made our revisions and corrections. The questions, suggestions and comments raised by the reviewer 2 have been addressed in this letter. The detailed summary is as follows: Reviewer #2: (1) Since a better 1 H NMR spectrum of G2 was obtained, Figure 2a should be replaced by the better one.

Response:
The 1 H NMR spectrum of G2 in Fig.2a was updated.
(2) The whole ESI-MS spectrum of G3 with the identified fragment signals ( Figure S2) should be presented in Figure 3c instead. In addition, tandem MS experiments may help clarify the source of fragments, which the authors attributed to the harsh ionization conditions. Response: The Figure 3c was updated with labeled fragments for the whole ESI-MS spectrum of G3. We have performed tandem MS using collisionallyinduced dissociation (CID) following Reviewer's suggestion. In CID, however, supramolecules were prone to losing different number of neutral PF5 instead of forming large fragments (Angew. Chem. Int. Ed., 2010, 49, 6539). To clarify the source of fragment, we further performed ESI-MS experiments of G3 with different ESI cone voltages. As shown in Figure R1, the signals of fragments increased when increasing the ESI cone voltage. Trace fragments could be observed when the ionization voltage is 2 kV ( Figure R1c). The fragments were further increased especially at low m/z region when cone voltage up to 4 kV ( Figure R1a). The further discussion on the fragment signals of G3 was added in P7 Paragraph 2 in the manuscript and the additional experimental results was added in Figure S3.

Response:
We have conducted molecular simulation for hexamer and heptamer assembled by L3 and Cd 2+ . As shown in Figure R2, the optimum structure of hexamer was highly distorted. In contrast, the structure of heptamer exhibited more extended skeleton with lower distortion. The torsion energy of heptamer (499.27 kcal/mol) is lower than hexamer (567.91 kcal/mol). Therefore, the self-assembly preferred the formation of heptamer rather than hexamer. Detailed dis cussion was added in P7 Paragraph 2 in the manuscript.

(4) The association constants of G1 in various solvents were calculated by UV/vis titration experiments.
However, there are many species generated from the complexation reaction between L1 and Cd(II). In addition, in the presence of excess metal ions (L:M = 1:24.5), the assembled structure may further changed to a bisterpyridine-metal (1:2) adduct. The authors did not specify which equilibrium they looked into and which method and model they used for the curve fitting. The coefficients of determination were pretty bad in Figures  S59 and S60. Since this is a complicated dynamic system, the calculation should be done more carefully. I support this manuscript to be published in Nature Communications after the mentioned issues are properly addressed.
Response: Thank you for Reviewer's suggestion. The bisterpyridine-metal (1:2) adduct was not taken into account when added excess amount of Cd 2+ in our first response letter.
Scheme R1. (I) Less than 1 equv. Cd(NO3)2; (II) more than 1.0 equv. Cd(NO3)2 binding with L1 We recalculated the association constants of G1 on 1:1 binding model (Scheme R1I) by Benesi-Hildebrand method (Equation1) according the reported literature (Spectrochim. Acta A, 2012, 90, 40-44). For example, as shown in Figure R3, the association constants of G1 in THF is 1.24 × 10 5 M -1 , which is close to the value (1.1 × 10 5 M -1 ) calculated before. The measured absorbance [1/(A-A 0 )] varied as a function of 1/[Cd 2+ ] in a linear relationship (R = 0.997), which is consistent with 1:1 complex formation. Similarly, the association constants in CHCl3/MeOH (v/v, 1/3), DMF and DMSO were all recalculated with satisfactory coefficients by this method. Note that in Benesi-Hildebrand plot, we only used the absorbance with L:M ratio slightly larger than 1:1 to calculate Ka. The association constants in various solvents have been updated in the manuscript and Supporting Information as Figure S60-63 in page S82-85.
The association constant (Ka) of L1 with Cd 2+ was determined using the Benesi-Hildebrand equation as follows: where A and A0 represent the absorbance of L1 in the presence and absence of Cd 2+ , respectively, Amax is the saturated absorbance of L after the addition of excess amount of Cd 2+ ; [Cd 2+ ] is the concentration of Cd 2+ ion added. Ka is the association constant. I hope that the revision will satisfy our reviewers' requirements. Again, I would like to thank the reviewers for their thoughtful and careful comments and appreciate very much for your kindly handling this manuscript.

Reviewer #2 (Comments to the Author):
The Benesi-Hildebrand method the authors used for calculating association constants is a specific approach for one-to-one complex systems (M + L -> ML  Figure R1, the calculated association constant in CHCl3/MeOH (1/2) is 4.17 ×10 10 M -2 . There is no significant difference for association constants in different solvents. It should be noted that the stability of multi-armed ligands L1-L3 binding with Cd 2+ should higher than monoterpyridine compound 16.
1 H, 13 C NMR and high resolution ESI-TOF mass spectrometry information of compound 16 were provided in Page S19. The further discussion on the determination of association constants of compound 16 binding with Cd 2+ was added in Page S83 and the additional experimental results were added in Supplementary Figures 65-67.
Scheme R1. Equilibrium for binding Cd 2+ to compound 16, where K1, K2 are the first and the second binding association constants, respectively, K is the overall association constant .