Introduction

One of ultimate goals in the chemistry of molecular machines is to reproduce motions in the macroscopic world at the level of molecules or their aggregates. A number of excellent examples of molecular machines have been reported to control a variety of motions, including rotation1,2,3,4,5 and translation6,7,8,9,10. However, controlling motion elements such as directionality3,11, speed4,10 and frequency (ON/OFF)9,12 is still a major challenge. For example, Stoddart et al. controlled the translational motion of rings shuttling along the axles by redox or acid/base6. Feringa et al. reported unidirectional molecular rotors based on photo- and thermal-isomerisation of olefins11. In addition to these motions, twisting motion and its inversion motion have recently attracted much attention in terms of motion complexity and asymmetry, leading to advanced molecular machines and chiral memory13,14,15,16,17.

We intuitively understand that the twisting behaviours are highly dependent on the mode of twisting. For instance, a loosely-twisted object will unwind easily, but a very tightly-twisted object may not. Can such movements be reproduced in the nanoscale world? Inspired by biomacromolecules, chemists have synthesised twisted molecules such as helical polymers18,19,20, helicenes21,22 and twisted macrocycles23,24,25, and in some of these examples, the design and control of their twisting motion are being studied. For instance, the rate of helicity inversion was controlled by exploiting the kinetic properties of coordination bonds in twisted metal complexes of synthetic peptides26, macrocycles27 and cryptands28. Another role of metal coordination is to dynamically fix the absolute configuration of the coordinating atoms, including the amine nitrogen atoms, which is usually immediately reversed. Thus, selective synthesis of metal complexes with different modes of twisting resulting in different configurations and/or conformation, or twisted isomers is a promising strategy to control the inversion motion without changing the chemical composition, but such isomers are limited to a few examples29,30,31,32. Thus, it is still challenging to selectively synthesise such twisted isomers.

Molecular motion is described by a combination of rotational and translational motions, but various modes are possible depending on the shape and size of the molecule itself, the flexibility of the structure derived from the bonding modes and the environment surrounding the molecule. Molecular twist is a mode included in many molecular motions of molecular machines. Selective synthesis of isomers with different mode of twisting and control of their motion such as inversion, is an indispensable task to realise more advanced molecular systems.

Herein we report the selective synthesis of two twisted isomers of a trinuclear PdII-macrocycle with a tightly- or loosely-twisted skeleton (Fig. 1). It is particularly important to emphasise that these two isomers have markedly different rates of helicity inversion depending on the mode of twisting with different absolute configurations of the diamine moieties locked by the metal ions. The loosely-twisted isomer exhibited rapid helicity inversion, whereas the helicity inversion in the tightly-twisted isomer was actually undetectable because the inversion process requires absolute configuration inversion of the nitrogen atoms. In other words, the helicity inversion of the twisted macrocycle is configurationally inhibited by twisting more tightly. Thus, this result is an excellent example of twisting motion controlled by the mode of twisting of a single chiral molecule with coordinating atoms of different configurations.

Fig. 1: The concept of molecular helicity inversion controlled by twisting mode due to differences in the absolute configurations of the diamine moieties locked by the metal ions.
figure 1

Colour: Pd yellow, C black, H white, N Purple (non-coordinated), red (R-configuration) and blue (S-configuration).

Results

Selective synthesis of two twisted isomers, 1tight and 1loose (Fig. 2)

Previously, our group synthesised helically-twisted PdII-macrocycles, [Pd3LCl6], from an achiral macrocyclic hexaamine ligand L and 3 equiv. of [PdCl2(CH3CN)2] in CH2Cl2:DMSOā€‰=ā€‰9/1 (v/v)33. The C3-symmetric macrocyclic skeleton with three PdII centres on the same side was helically twisted by intramolecular C-HĀ·Ā·Ā·Ļ€ interactions (Fig. 3i, j). This complex exhibits helicity inversion between the (P)- and (M)-enantiomers with an inversion rate of 14ā€‰sā€“1 at 292ā€‰K in CD2Cl2:DMSO-d6ā€‰=ā€‰9/1 (v/v). In this study, we aimed to significantly modulate the rate of helicity inversion by replacing all two chloride ligands on the three PdII centres with 4,4ā€™-di-tert-butyl-2,2ā€™-bipyridine (tBu2bpy) ligands.

First, we attempted to synthesise a twisted PdII-macrocycle with Pd(tBu2bpy) moieties in a manner similar to our previous study33. Ligand L was reacted with 3.2 equiv. of [Pd(tBu2bpy)(OH2)2](OTf)2 in CH2Cl2 at room temperature for 4ā€‰h. After removal of the solvent, the residue was recrystallised to afford colourless crystals 1tight in 33% yield (Fig. 2). The composition was determined by elemental analysis to be [Pd3L(tBu2bpy)3](OTf)6Ā·(H2O)4.9Ā·(Et2O)0.15. Single-crystal X-ray diffraction (ScXRD) analysis revealed that [Pd3L(tBu2bpy)3](OTf)6 has a C3-symmetric twisted structure (Fig. 3aā€“c), and the twisting macrocyclic structure was different from that of [Pd3LCl6] (Fig. 3j). Specifically, the three ortho-phenylenediamine moieties folded inside the macrocycle to form a tightly-twisted skeleton, and this complex [Pd3L(tBu2bpy)3](OTf)6 is referred to here as the tightly-twisted isomer (1tight): the other loosely-twisted isomer (1loose) will be discussed below (the definition of the twisted isomers; see the Supplementary Section 2.4). This isomer 1tight has (P)- or (M)-helicity, defined by the direction from the para-phenylene rings toward the inner amine protons (Fig. 2), and both enantiomers crystallised as a racemate. Unlike [Pd3LCl6], 1tight had three amine protons facing outward of the macrocycle and the other three facing inward, forming hydrogen bonds with one triflate ion in the cavity. The absolute configurations of the six nitrogen atoms, the chiral centres in the (P)- and (M)-enantiomers, were therefore all-R- and all-S-configuration, respectively. Intramolecular C-HĀ·Ā·Ā·Pd interactions between one of the protons of para-phenylene and the Pd centre were suggested by the HĀ·Ā·Ā·Pd distance (2.74ā€‰Ć…) and the C-HĀ·Ā·Ā·Pd angle (120Ā°) as one factor stabilising the tightly-twisted structure. These are consistent with typical anagostic interactions (2.3ā€“2.9ā€‰Ć…, 110ā€“170Ā°)34. To confirm the structure in solution, the crystals were then dissolved in acetone or dichloromethane and analysed by 1D 1H and 19F NMR spectroscopies, 2D NMR spectroscopies (1Hāˆ’1H COSY and ROESY), and high resolution-electrospray ionization time-of-flight (HR-ESI-TOF) mass spectrometry (m/zā€‰=ā€‰1100.2498 as [Pd3(Hā€“1L)(tBu2bpy)3(OTf)3]2+) (Supplementary Figs. 4ā€“16). The 19F NMR spectrum showed two separate triflate signals, one of which was assigned to a triflate incorporated within the macrocycle (Supplementary Fig. 6). Two sets of diastereotopic methylene proton signals (Hcā€“f) were observed in the 1H NMR spectrum in acetone-d6, indicating that the structure is chiral (Fig. 3a, d). Notably, the downfield shift of one para-phenylene signal (Hl) to 9.8 ppm even at 300ā€‰K suggested the presence of a C-HĀ·Ā·Ā·Pd interaction (Supplementary Fig. 14). Moreover, an ortho-phenylenediamine proton signal (Hg) was highly upfield shifted to 4.9 ppm due to the shielding effect from the adjacent ortho-phenylenediamine moieties clustered inside the macrocycle. These results suggest that the tightly-twisted structure is maintained in solution. The interactions suggested by ScXRD and NMR analyses were well supported by natural bond orbital (NBO) and noncovalent interaction (NCI) plot analyses after optimising the geometry with density functional theory (DFT) calculations (Supplementary Figs. 118ā€“120).

Fig. 2: Synthesis of two isomeric PdII complexes, 1tight and 1loose, and the absolute configuration of the amine nitrogen atoms.
figure 2

i CH2Cl2, rt, 4ā€‰h, 33%. ii CHCl3, rt, 3ā€‰h, 64%. iii [Pd(tBu2bpy)(OH2)2](OTf)2, CH2Cl2, rt for 3ā€‰h then reflux for 1.5ā€‰h, 49% (31% in two steps).

Fig. 3: Characterisation of the two twisted isomers of 1.
figure 3

a Substructural formula of 1tight. b, c Side and top views of the crystal structure of 1tight with one triflate incorporated via multipoint hydrogen bonding. Hydrogen atoms except amine and methylene moieties are omitted for clarity. d 1H NMR spectrum (500ā€‰MHz, acetone-d6, 300ā€‰K) of 1tight. e Substructural formula of 1loose. f, g Side and top views of the crystal structure of 1loose with one triflate disordered in the inner space. Hydrogen atoms except amine and methylene moieties are omitted for clarity. h 1H NMR spectrum (500ā€‰MHz, acetone-d6, 300ā€‰K) of 1loose. i, j Structural formula and the reported crystal structure (top view) of [Pd3LCl6]33.

To synthesise another isomeric trinuclear PdII-macrocycle with a skeleton similar to [Pd3LCl6], we optimised the reaction conditions and found that a dinuclear PdII-macrocycle is a key intermediate for the selective synthesis of the other isomers (Fig. 2). When ligand L was reacted with 1.6 equiv. of [Pd(tBu2bpy)(OH2)2](OTf)2 in CHCl3 at room temperature, a dinuclear [Pd2L(tBu2bpy)2](OTf)4 (2) complex precipitated. Its meso-twisted skeleton with two intramolecular C-HĀ·Ā·Ā·Ļ€ interactions was deduced from the crystal structure of a PtII-analogue, 1D and 2D NMR and HR-ESI-TOF mass analyses (m/zā€‰=ā€‰689.2654 as [Pd2(Hā€“2L)(tBu2bpy)2]2+) (Supplementary Figs. 34ā€“45). Dinuclear complex 2 was then reacted again with 1.2 equiv. of [Pd(tBu2bpy)(OH2)2](OTf)2 in CH2Cl2, and the product was recrystallised to afford colourless crystals 1loose, [Pd3L(tBu2bpy)3](OTf)6Ā·(H2O)4, in 31% total yield. ScXRD analysis revealed that the crystals were composed of another C3-symmetric trinuclear complex with a twisted skeleton similar to [Pd3LCl6] (Fig. 3eā€“g, i, j). Unlike 1tight, the three ortho-phenylenediamine moieties were not folded much and located outside the macrocyclic structure, instead, the inside was filled with para-phenylene and three methylene moieties, forming C-HĀ·Ā·Ā·Ļ€ interactions between them. Thus, this complex [Pd3L(tBu2bpy)3](OTf)6 (1loose) can be regarded as a loosely-twisted isomer of 1tight. This isomeric complex also formed racemic crystals consisting of (P)- and (M)-enantiomers. In contrast to 1tight, all six amine protons of 1loose were located outside of the macrocyclic structure, resulting in an alternating R- and S-absolute configuration of nitrogen atoms. Therefore, helicity inversion between the (P)- and (M)-1loose preserves the alternate absolute configurations (alt-R/S) of the nitrogen atoms. One triflate ion was found to be disordered in the space surrounded by three bipyridine portions. Analyses of 1loose dissolved in acetone by 1D and 2D NMR spectroscopies and HR-ESI-TOF mass spectrometry (m/zā€‰=ā€‰1100.2410 as [Pd3(Hā€“1L)(tBu2bpy)3(OTf)3]2+) revealed that the structure of 1loose in solution is consistent with the crystal structure (Supplementary Figs. 19ā€“29). For instance, one methylene signal (Hc) was upfield shifted to 1.9 ppm in 1H NMR, which supported the C-HĀ·Ā·Ā·Ļ€ interactions observed in the crystal structure (Fig. 3f, g). This interaction was also supported by NBO and NCI plot analyses after geometry optimisation using DFT calculation (Supplementary Figs. 126 and 127).

1loose was stable in acetone-d6 at room temperature for two weeks, as evidenced by 1H NMR analysis (Supplementary Fig. 32). In contrast, 1tight was less stable and slowly isomerised to 1loose in acetone-d6 while also producing non-assignable by-products. The isomerisation rate of 1tight to 1loose in acetone-d6 at 293ā€‰K was estimated to be (5.7ā€‰Ā±ā€‰0.4) Ɨ 10ā€“6 sā€“1 by the time course 1H NMR analysis, assuming the isomerisation to be a pseudo first order reaction (Supplementary Section 3). This result indicates that 1tight and 1loose are the kinetic and thermodynamic products in acetone, respectively, and is consistent with the fact that 1tight is obtained kinetically under mild conditions in CH2Cl2. As mentioned above, dinuclear complex 2, which preferentially precipitated when reacted with less than 2 equiv. of PdII salts in CHCl3, is an important intermediate for the selective synthesis of 1loose. This result is consistent with the fact that the absolute configuration of the four PdII-coordinated amine nitrogen atoms in 2 is the same as that of 1loose and no configurational change is required when 2 is converted to 1loose. In contrast, to convert to 1tight with an all-R- or all-S-configuration, 2 requires a configurational inversion of the two PdII-coordinated amine nitrogen atoms. Thus, the conversion from 2 to 1tight is kinetically undesirable, resulting in a selective conversion to 1loose (Fig. 2).

The difference in the absolute configuration of the PdII-coordinated amine nitrogen atoms is thus a notable structural difference between 1tight and 1loose. It is noteworthy that the two diastereomers can also be regarded as in/out-isomers if we focus on the nitrogen-containing cyclic structures, which is usually applied to bridged bicyclic compounds35. Here, the two nitrogen atoms of one ortho-phenylenediamine moiety are the bridgehead atoms of the bicyclic structure bridged via PdII. According to the definition of in/out-isomers, 1tight and 1loose correspond to the in,out- and out,out-isomers, respectively, from the direction of the N-H moieties.

Estimation of helicity inversion rate by exchange spectroscopy (EXSY)

The inversion rates of 1loose and 1tight were then investigated. The helicity inversion rate (k) between (P)- and (M)-1loose was evaluated using EXSY with varying mixing time. The rate constant (k) was calculated using the ratio of integrations of chemical exchange signals (e.g., Ho and Hv for different bipyridine moieties) produced by helicity inversion, which was 3.31ā€‰Ā±ā€‰0.02ā€‰sā€“1 at 300ā€‰K in acetone-d6 (Fig. 4aā€“c, Supplementary Sections 4.1, 4.3). In the EXSY analysis, the rate constant (k) obtained was defined as the sum of the rate constants of the helicity inversion from (P)- to (M)-1loose (kPM) and from (M)- to (P)-1loose (kMP) (Fig. 4a)36. The Gibbs free energy (Ī”Gā€”), enthalpy (Ī”Hā€”) and entropy (Ī”Sā€”) of activation at 300ā€‰K were estimated to be 70.6ā€‰Ā±ā€‰1.3ā€‰kJ/mol, 86.1ā€‰Ā±ā€‰0.9ā€‰kJ/mol and 51.9ā€‰Ā±ā€‰3.1ā€‰J/(molĀ·K), respectively, by the Eyring plot based on the inversion rate at temperatures varying in the range from 280 to 310ā€‰K (Supplementary Figs. 50ā€“52). The large positive entropy of activation is probably due to the formation of lower symmetry structures in the transition state and the release of anions and solvent molecules. The inversion rate obtained here is nearly equivalent to that of [Pd3LCl6] with a loosely-twisted conformation (kā€‰=ā€‰14ā€‰sā€“1 at 292ā€‰K in CD2Cl2/DMSO-d6)33, although the ligands and conditions are different. The relatively fast inversion rate of the loosely-twisted complexes is consistent with the argument, discussed earlier, that the helicity inversion does not require configurational inversion of the amine nitrogen atoms.

Fig. 4: Estimation of the helicity inversion rate.
figure 4

a Scheme of the helicity inversion between (P)- and (M)-1loose. b EXSY spectrum (500ā€‰MHz, acetone-d6, 300ā€‰K, mixing time = 0.3ā€‰s, 0.19ā€‰mM) of (P/M)-1loose. The chemical exchange signals between Ho and Hv (ex o-v), flamed in blue, were used to estimate the inversion rate. c Partial structural formula of 1loose. d Scheme of the helicity inversion between (P)- and (M)-1tight. e EXSY spectrum (500ā€‰MHz, acetone-d6, 300ā€‰K, mixing time = 0.3ā€‰s) of (P)- and (M)-1tight. No chemical exchange signals between Ho and Hv were observed, as shown in the blue dotted boxes. f Partial structural formula of 1tight.

In contrast, no chemical exchange signals between Ho and Hv were observed in the EXSY spectra of 1tight in acetone-d6 at 300ā€‰K (Fig. 4dā€“f). This suggests that the inversion rate of 1tight is too slow to be evaluated by EXSY. Compared to 1loose, the helicity inversion of 1tight requires configurational inversion of all amine nitrogen atoms, as described above, which slows the inversion rate. The triflate incorporated into the interior of the macrocycle may also contribute to stabilising the absolute configuration of the inward amine nitrogen atoms via hydrogen bonding. Therefore, we next attempted to synthesise enantio-enriched 1tight using a chiral auxiliary and estimated its racemisation rate as the helicity inversion rate.

Asymmetric synthesis of 1tight

Asymmetric synthesis of enantio-enriched 1tight was investigated using chiral sulfoxides as additives (Fig. 5). Among several chiral sulfoxides examined, (R/S)-mesityl methyl sulfoxide ((R/S)-3) was the best in terms of product yield and optical purity (Fig. 5c, d and Supplementary Fig. 96). Specifically, 3.2 equiv. of [Pd(tBu2bpy)(OH2)2](OTf)2 was mixed with an excess of (S)-3 and reacted with L in CH2Cl2 at ā€“70ā€‰Ā°C for 4ā€‰h (Fig. 5a). After washing (S)-3, enantio-enriched 1tight was obtained in 15% yield. The enantiomeric excess of the product was determined to be 25% ee by 1H NMR in CD2Cl2 containing Ī”-TRISPHAT tetrabutylammonium salt (Ī”-4) as a chiral shift reagent (Fig. 5e, f and Supplementary Fig. 90). Ī”-4 had no effect on the enantiomeric ratios of the products, as evidenced by time-course 1H NMR analysis of racemic and enantio-enriched 1tight (Supplementary Figs. 114ā€“117). The product was pure enough for further analysis and was not recrystallised to prevent changes in enantiomeric excess. When analysed by circular dichroism (CD) spectroscopy, this product exhibited a negative Cotton effect at 325ā€‰nm in CH2Cl2 (Fig. 5b (red line)). Using enantiomer (R)-3, the other enantio-enriched 1tight with opposite chirality was synthesised in the same way, and the product showed a positive Cotton effect at 325ā€‰nm in the CD spectrum (Fig. 5b (blue line)). The mirror image of the CD spectra indicated that the asymmetric synthesis was successfully achieved by chiral 3.

Fig. 5: Asymmetric synthesis of 1tight.
figure 5

a Synthesis of (M)-enantio-enriched 1tight. b CD spectra (CH2Cl2, 293ā€‰K, lā€‰=ā€‰1.0ā€‰cm) of (P)- and (M)-enantio-enriched 1tight (blue and red lines, respectively). Structural formula of (R)-3 c, (S)-3 d and Ī”-TRISPHAT tetrabutylammonium salt (Ī”-4) e. f Partial 1H NMR spectra (500ā€‰MHz, CD2Cl2, 300ā€‰K) of the (M)-enantio-enriched 1tight with Ī”-4. The upper spectrum (day 0) was recorded by dissolving the as-synthesised product in CD2Cl2 containing Ī”-4. The lower spectrum (day 3) was obtained by dissolving the product in acetone-d6, allowing it to stand at 293ā€‰K for 3 days, evaporating the solution at room temperature and then redissolving in CD2Cl2 containing Ī”-4. Day 0 and day 3 enantiomer excesses were evaluated by the integral of the signals after deconvolution. g 1H NMR analysis (500ā€‰MHz, acetone-d6, 300ā€‰K) of the (M)-enantio-enriched 1tight without Ī”-4 dissolved in acetone-d6 after 0 day (upper) and 3 days (lower). Signals of 1loose are indicated by orange triangles.

Time dependent-DFT calculations [M06-D3/def2svp for Pd, 6-31G(d) for other atoms] were then performed to determine the absolute structures of both enantio-enriched 1tight. The calculated CD spectrum of the optimised (P)-1tight qualitatively reproduced the experimental spectrum of the enantio-enriched 1tight synthesised with (R)-3 (Supplementary Fig. 121). For example, positive Cotton effects in the low energy region were found in both the experimental and calculated spectra. Similar results were obtained in the calculations of CD spectra with other functionals or basis sets (Supplementary Figs. 122ā€“125). These support that the (P)- and (M)-enantio-enriched 1tight were synthesised with (R)- and (S)-3, respectively.

Helicity inversion versus twist loosening observed in 1tight (Fig. 6)

The helicity inversion rate was evaluated using (M)-enantio-enriched 1tight. An acetone-d6 solution of the (M)-enantio-enriched 1tight (P:Mā€‰=ā€‰37:63) was allowed to stand at 293ā€‰K for 3 days (Fig. 5g). After the solvent was removed, the enantiomeric ratio was examined using Ī”-4 in CD2Cl2, and its enantiomeric ratio (P:Mā€‰=ā€‰38:62) was almost the same as that of the starting material. This indicates that either the helicity inversion is too slow to be detected or that the inversion of 1tight does not occur under this condition (Fig. 5f, Supplementary Sections 7.1ā€“7.2). Similarly, the inversion rate in CD2Cl2 was also examined at 293ā€‰K, but no inversion was observed in 10 days (Supplementary Figs. 112 and 113). Besides, 1tight gradually isomerised to 1loose as described above, and the isomerisation to 1loose was observed during these analyses. These results suggest that the rate of isomerisation from 1tight to 1loose (twist loosening) (5.7 Ɨ 10ā€“6 sā€“1) is faster than that between (P)- and (M)-1tight (helicity inversion). The faster isomerisation from 1tight to 1loose can be explained from the number of amine nitrogen atoms whose absolute configuration inverts. That is, in the case of (all-R) or (all-S)ā€‰ā†’ā€‰(alt-R/S) (twist loosening), only three of the six amine portions need to be inverted, but in the case of (all-R) ā‡„ (all-S) (helicity inversion), all six amine portions must be inverted. Since configurational inversion of the amine moieties involves dissociation of the Nā€“Pd or Nā€“H bonds, the number of nitrogen inversion sites may affect the isomerisation rate. This consideration is also applied to understanding that the helicity inversion of 1loose (1.38ā€‰sā€“1 at 293ā€‰K, estimated by the Eyring plot, Supplementary Fig. 52) is much faster than that of 1tight. This is because the helicity inversion of 1loose does not require the configurational changes of amine nitrogen atoms ((alt-R/S) ā‡„ (alt-R/S)). On the other hand, the inversion of the all-R or all-S configuration in 1tight probably needs to occur in a stepwise manner via the intermediary alt-R/S configuration, but the intermediate corresponding to thermodynamically stable 1loose is no longer isomerised to the all-S or all-R configuration, respectively (Fig. 7).

Fig. 6: Helicity inversion versus twist loosening of 1tight.
figure 6

Undetectable helicity inversion between (P)- and (M)-1tight, twist loosening from 1tight to 1loose and fast helicity inversion between (P)- and (M)-1loose.

Fig. 7: Possible isomers and their isomerisation pathways of 1.
figure 7

All possible diastereomers and their isomerisation pathways for 1 are shown in this scheme. The black circles indicate the helically twisted macrocyclic structures of 1. The notions RR, SS and SR above the black circles indicate pairs of the absolute configurations of the amine nitrogen atoms of the ortho-phenylenediamine moiety observed in the crystal strictures. The isomerisation pathways to the RS configuration are excluded in this scheme, as a pair of the amine nitrogen atoms with the RS configurations, with the two amine protons pointing towards the inside of the macrocycle, was not observed experimentally. The green arrows indicate the direct isomerisation pathways from 1tight to 1loose, and the dotted arrows indicate other possible isomerisation pathways.

Finally, 1tight was found to be more stabilised when the solvent was substituted from acetone-d6 to CD2Cl2. The stability of (M)-enantio-enriched 1tight was examined by 1H NMR spectroscopy and it was found that in acetone-d6 twist loosening and degradation began within 1 day at 293ā€‰K, whereas in CD2Cl2 such changes were significantly slower even after 6 days (Supplementary Figs. 109 and 112). Of note, the addition of Ī”-4 in CD2Cl2 also markedly inhibited the tight-to-loose isomerisation. 1H NMR spectra of (M)-enantio-enriched 1tight in the presence of Ī”-4 (17 equiv.) showed that most of 1tight remained without significant isomerisation or degradation at 293ā€‰K in CD2Cl2 even after 14 days (Supplementary Fig. 116). One possibility is that the Ī”-TRISPHAT anion associates with the cationic 1tight with a similar symmetry to increase the stability, which was supported by the shift and splitting of the 1H NMR and 19F NMR signals of the encapsulated triflate (Supplementary Figs. 89 and 115). Thus, a very slow or no inversion can be switched to a faster inversion by isomerisation from a tight to a loose state, and the switching speed can be adjusted by additives or solvents.

Discussion

In this study, we have succeeded in selectively synthesising two twisted isomers of trinuclear PdII-macrocycles with markedly different rates of helicity inversion. In the tightly-twisted isomer 1tight, the three ortho-phenylenediamine moieties were folded inside the macrocyclic skeleton and the absolute configurations of the amine nitrogen atoms were all-R or all-S, while in the loosely-twisted isomers 1loose, the three ortho-phenylenediamine moieties were folded outside the skeleton and the absolute configurations of the amine nitrogen atoms were alt-R/S. In stark contrast to 1tight, which shows a very slow or no inversion, 1loose exhibits fast helicity inversion (1.38ā€‰sā€“1 at 293ā€‰K in acetone-d6). Moreover, the inversion kinetics can be controlled by isomerisation in a range from 1tight, where no inversion is detected, to 1loose, where inversion is fast. Our approach to the control of helicity inversion motion by the twisted isomers resulting from the configurational locking with metal ions is quite different from conventional approaches that require chemical substitutions or additives to control twisting motions. The new strategy of controlling molecular motions by the mode of twisting with coordinating atoms of different configuration is expected to be applicable to a variety of systems and can be expanded to the design of more sophisticated molecular machines.

Methods

Synthesis of 1tight

A CH2Cl2 solution (1.0ā€‰mL) of L (10.0ā€‰mg, 15.9ā€‰Āµmol, 1.0 equiv.) was mixed with a CH2Cl2 solution (2.0ā€‰mL) of [Pd(tBu2bpy)(OH2)2](OTf)2Ā·(H2O)2 (38.2ā€‰mg, 51.3ā€‰Āµmol, 3.2 equiv.), and then stirred at room temperature for 4ā€‰h. During the reaction, the colour of the solution was changed from pale yellow to purple in a few minutes. The reaction mixture was filtered to remove the precipitate and the filtrate was evaporated. The resulting solid was washed with CHCl3 and the residue was dried up under reduced pressure. The solid was recrystallised from CH2Cl2 by vapour diffusion of Et2O. The obtained plate crystals were washed with a small amount of CHCl3 and dried up under reduced pressure to afford tightly-twisted 1tight, [Pd3L(tBu2bpy)3](OTf)6Ā·(H2O)4.9Ā·(Et2O)0.15, (14.3ā€‰mg, 5.19ā€‰Āµmol, 33% yield) as a colourless solid.

Mp: > 257ā€‰Ā°C (decomp.). 1H NMR (500ā€‰MHz, acetone-d6, 300ā€‰K): Ī“ 9.81 (d, Jā€‰=ā€‰7.0ā€‰Hz, 3H), 9.09 (d, Jā€‰=ā€‰6.0ā€‰Hz, 3H), 9.00 (d, Jā€‰=ā€‰6.0ā€‰Hz, 3H), 8.95 (d, Jā€‰=ā€‰1.5ā€‰Hz, 3H), 8.88 (d, Jā€‰=ā€‰2.0ā€‰Hz, 3H), 8.85 (d, Jā€‰=ā€‰3.5ā€‰Hz, 3H), 8.10 (dd, Jā€‰=ā€‰5.0, 1.0ā€‰Hz, 3H), 8.03 (dd, Jā€‰=ā€‰6.0, 1.5ā€‰Hz, 3H), 7.98 (d, Jā€‰=ā€‰8.5ā€‰Hz, 3H), 7.80 (d, Jā€‰=ā€‰5.5ā€‰Hz, 3H), 7.64 (t, Jā€‰=ā€‰8.0ā€‰Hz, 3H), 7.41 (d, Jā€‰=ā€‰10.5ā€‰Hz, 3H), 7.10 (t, Jā€‰=ā€‰7.5ā€‰Hz, 3H), 6.94 (d, Jā€‰=ā€‰7.0ā€‰Hz, 3H), 6.40 (d, Jā€‰=ā€‰7.0ā€‰Hz, 3H), 5.09 (d, Jā€‰=ā€‰14.5ā€‰Hz, 3H), 5.05 (d, Jā€‰=ā€‰15.0ā€‰Hz, 3H), 4.89 (d, Jā€‰=ā€‰7.5ā€‰Hz, 3H), 4.87 (d, Jā€‰=ā€‰13.5ā€‰Hz, 3H), 4.50 (dd, Jā€‰=ā€‰14.0, 11.5ā€‰Hz, 3H), 1.56 (s, 27H), 1.50 (s, 27H). 13C NMR (126ā€‰MHz, acetone-d6, 301ā€‰K): Ī“ 168.9, 168.7, 157.8, 157.3, 152.2, 150.4, 142.0, 141.8, 134.5, 134.2, 133.4, 133.1, 132.1, 131.9, 131.2, 129.2, 128.6, 126.9, 126.6, 126.5, 123.3, 123.2, 122.9, 120.8, 118.2, 63.0, 61.3, 37.1, 37.0. The 13C signals of tert-butyl groups were overlapped with those of acetone-d6 and could not be identified. The TfO anion has four 13C signals but only three signals were observed due to the low S/N ratio. 19F NMR (471ā€‰MHz, acetone-d6, 300ā€‰K): Ī“ ā€“75.4, ā€“76.1. IR (ATR, cmā€“1): 3144 (br), 2967, 1618, 1417, 1248, 1155, 1028, 810, 636. UV-vis (CH2Cl2, 293ā€‰K, 87.3ā€‰ĀµM): Ī»max (nm) (Īµ (Mā€“1 cmā€“1)) = 309.8 (3.98 Ɨ 104). HRMS (ESI-TOF): m/zā€‰=ā€‰1100.2498 as [Pd3(Hā€“2L)(tBu2bpy)3](OTf)3]+ (calcd 1100.2462). Anal. Calcd for C102.6H125.3F18N12O23.05Pd3S6 {[Pd3L(tBu2bpy)3](OTf)6Ā·(H2O)4.9Ā·(Et2O)0.15}: C 44.82, H 4.58, N 6.09; found: C 44.81, H 4.58, N 6.09.

Crystal data for tightly-twisted Pd3L(tBu2bpy)3Ā·(OTf)5.08Ā·(H2O)6.95Ā·(CH2Cl2)1.27 (missing triflates were not observed due to severe disorder): C102.36H116.55Cl2.55F15.25N12O22.20Pd3S5.08, Fwā€‰=ā€‰2732.41, crystal dimensions 0.131 Ɨ 0.081 Ɨ 0.031 mm3, trigonal, space group R-3, aā€‰=ā€‰23.0889(2), cā€‰=ā€‰41.6772(6) ƅ, Vā€‰=ā€‰19241.3(4) ƅ3, Zā€‰=ā€‰6, Ļcalcdā€‰=ā€‰1.415ā€‰gā€‰cmā€“3, Ī¼ā€‰=ā€‰53.78ā€‰cmā€“1, Tā€‰=ā€‰93ā€‰K, Ī»(CuKĪ±)ā€‰=ā€‰1.54187ā€‰Ć…, 2Īømaxā€‰=ā€‰144.478Ā°, 39889/8311 reflections collected/unique (Rintā€‰=ā€‰0.0578), R1ā€‰=ā€‰0.0845 (Iā€‰>ā€‰2Ļƒ(I)), wR2ā€‰=ā€‰0.2630 (for all data), GOFā€‰=ā€‰1.113, largest diff. peak and hole 1.414/ā€“1.123 eƅā€“3. CCDC deposit number 2190130.

Synthesis of 1loose

A CHCl3 solution (2.2ā€‰mL) of L (13.2ā€‰mg, 20.9 Āµmol, 1.0 equiv.) was mixed with a CHCl3 solution (4.7ā€‰mL) of [Pd(tBu2bpy)(OH2)2](OTf)2Ā·(H2O)2 (25.2ā€‰mg, 33.8 Āµmol, 1.6 equiv.), and then stirred at room temperature for 3ā€‰h. During the reaction, a pink solid was precipitated. The resulting precipitate was collected by filtration and washed with CHCl3 to obtain a dark pink solid whose main component was dinuclear metallocycle 2 (21.3ā€‰mg, 10.7 Āµmol, 64%), which was then suspended in CH2Cl2 (10ā€‰mL). To the suspension was added a CH2Cl2 solution (5ā€‰mL) of [Pd(tBu2bpy)(OH2)2](OTf)2Ā·(H2O)2 (9.7ā€‰mg, 13.0 Āµmol, 1.2 equiv. to the dinuclear metallocycle). The reaction mixture was stirred at room temperature for 3ā€‰h, and then heated at reflux for 1.5ā€‰h. During heating, a colourless solid was precipitated. The resulting precipitate was collected by filtration and dried under reduced pressure. This solid was recrystallised from acetone by vapour diffusion of Et2O to afford loosely-twisted 1loose, [Pd3L(tBu2bpy)3](OTf)6Ā·(H2O)4, (14.2ā€‰mg, 5.21 Āµmol, 31% in total) as colourless plate crystals.

Mp: > 272ā€‰Ā°C (decomp.). 1H NMR (500ā€‰MHz, acetone-d6, 300ā€‰K): Ī“ 8.99 (d, Jā€‰=ā€‰6.0ā€‰Hz, 3H), 8.69 (s, 3H), 8.62 (d, Jā€‰=ā€‰1.5ā€‰Hz, 3H), 8.47 (d, Jā€‰=ā€‰1.5ā€‰Hz, 3H), 8.10 (d, Jā€‰=ā€‰8.0ā€‰Hz, 3H), 8.00 (brs, 3H), 7.91 (dd, Jā€‰=ā€‰5.5, 1.0ā€‰Hz, 3H), 7.82 (brs, 3H), 7.68 (t, Jā€‰=ā€‰7.5ā€‰Hz, 3H), 7.62 (brs, 3H), 7.55 (m, 6H), 7.47 (brs, 3H), 7.19 (d, Jā€‰=ā€‰5.0ā€‰Hz, 3H), 6.94 (d, Jā€‰=ā€‰5.0ā€‰Hz, 3H), 6.75 (brs, 3H), 4.78 (d, Jā€‰=ā€‰13.0ā€‰Hz, 3H), 4.23 (d, Jā€‰=ā€‰13.0ā€‰Hz, 3H), 3.21 (d, Jā€‰=ā€‰11.5ā€‰Hz, 3H), 1.88 (brs, 3H), 1.54 (s, 27H), 1.43 (s, 27H). 13C NMR (126ā€‰MHz, acetone-d6, 300ā€‰K): Ī“ 168.6, 166.9, 158.3, 156.4, 151.4, 150.0, 145.6, 141.7, 135.1, 135.1, 134.8, 134.0, 133.8, 133.6, 131.3, 131.3, 127.3, 125.8, 125.7, 123.8, 123.2, 122.8, 120.7, 118.1, 63.1, 60.4, 36.9, 36.7, 30.8, 30.2. 1H NMR signals of para-phenylene moieties were not fully assigned at 300ā€‰K because the rotation of the para-phenylene moieties was so fast that extra cross signals derived from chemical exchange processes appeared in the ROESY spectrum and disturbed the assignment of the signals. So, the 1H NMR signals of para-phenylene moieties were assigned by 2D 1Hāˆ’1H COSY and ROESY NMR analyses at 270ā€‰K where the rotation of para-phenylene moieties was slow enough. IR (ATR, cmā€“1): 3459 (br), 2975, 1620, 1417, 1240, 1156, 1027, 844. UV-vis (CH3CN, 293ā€‰K, 82.3ā€‰ĀµM): Ī»max (nm) (Īµ (Mā€“1 cmā€“1)) = 311 (3.69ā€‰Ć—ā€‰104). HRMS (ESI-TOF): m/zā€‰=ā€‰1100.2410 as [Pd3(Hā€“1L)(tBu2bpy)3](OTf)4]+ (calcd 1100.2462). Anal. Calcd for C102H122F18N12O22Pd3S6 {[Pd3L(tBu2bpy)3](OTf)6Ā·(H2O)4}: C 45.01, H 4.52, N 6.18; found: C 45.04, H 4.52, N 6.17.

Crystal data for loosely-twisted Pd3L(tBu2bpy)3Ā·(OTf)6Ā·(C3H6O)0.375Ā·(H2O)0.5: C103.12H116.25F18N12O18.88Pd3S6, Fwā€‰=ā€‰2679.38, crystal dimensions 0.221ā€‰Ć—ā€‰0.150ā€‰Ć—ā€‰0.087ā€‰mm3, trigonal, space group P-3c1, aā€‰=ā€‰23.8678(5), cā€‰=ā€‰29.5387(5) ƅ, Vā€‰=ā€‰14572.9(7) ƅ3, Zā€‰=ā€‰4, Ļcalcdā€‰=ā€‰1.221ā€‰gā€‰cmā€“3, Ī¼ā€‰=ā€‰44.34ā€‰cmā€“1, Tā€‰=ā€‰93.15ā€‰K, Ī»(CuKĪ±)ā€‰=ā€‰1.54178ā€‰Ć…, 2Īømaxā€‰=ā€‰134.130Ā°, 30304/8396 reflections collected/unique (Rintā€‰=ā€‰0.0431), R1ā€‰=ā€‰0.1399 (I >ā€‰2Ļƒ(I)), wR2ā€‰=ā€‰0.3524 (for all data), GOFā€‰=ā€‰1.473, largest diff. peak and hole 6.128/ā€“2.917 eƅā€“3. CCDC deposit number 2190129.

Asymmetric synthesis of (M)-enantio-enriched 1tight from (S)-3

A CH2Cl2 solution (0.5ā€‰mL) of (S)-3 (96% ee, 93.5ā€‰mg, 513 Āµmol, 65 equiv.) was added to a CH2Cl2 solution (1.0ā€‰mL) of [Pd(tBu2bpy)(OH2)2](OTf)2Ā·(H2O)2 (19.0ā€‰mg, 25.5 Āµmol, 3.2 equiv.), and this reaction mixture was then cooled to ā€“70ā€‰Ā°C. To this solution was added a CH2Cl2 solution (1.0ā€‰mL) of L (5.0ā€‰mg, 7.93ā€‰Āµmol, 1.0 equiv.) at ā€“70ā€‰Ā°C. This reaction mixture was stirred at ā€“70ā€‰Ā°C for 4ā€‰h. After bringing the reaction mixture to room temperature, Et2O was added and the reaction mixture was filtered. The filtrate was evaporated under reduced pressure to afford (S)-mesityl methyl sulfoxide (87.9ā€‰mg, 482ā€‰Āµmol, 94% recovery yield, 95% ee). The precipitate was dissolved in CH2Cl2 and filtered to remove insoluble residue. The filtrate was evaporated and dried in vacuo. After the resulting solid was washed with CHCl3, (M)-enantio-enriched 1tight was obtained as a pale yellow solid (3.2ā€‰mg, 1.19ā€‰Āµmol, 15% yield, 25% ee). The enantiomeric excess was estimated using Ī”-4 as a chiral shift reagent.