Introduction

Conformational isomerism is a foundational concept1,2,3 in classical stereochemistry. Even so-called pure compounds are often a dynamic equilibrium of numerous conformations with different chemistries and reactivities4. Many biochemical processes depend on precise changes in the conformations of nucleic acids and proteins, resulting in highly stereospecific chemical outcomes5. For example, finely tuned conformational adaptations play a key role in defining the catalytic activity of enzymes6,7.

Drawing inspiration from the machine-like behavior of some of these naturally occurring macromolecules, scientists began to contemplate8 the design of wholly synthetic systems through which relative molecular motion could be controlled externally. In the simplest of cases e.g., molecular switches, external stimuli can be used to orchestrate changes in the configuration or conformation of molecules9,10,11,12,13,14,15,16. Structurally, many of these switches are based on mechanically interlocked molecules (MIMs), e.g., bistable catenanes17,18,19,20 and rotaxanes21,22,23,24,25,26,27,28,29,30,31,32. In these switches, the relative positioning of the interlocked components—referred to as coconformations33—can be controlled. An early34, and still popular31,32,35,36 approach to the design of these switches is the controlled relative translation of interlocked rings (and dumbbells in the case of rotaxanes) to encircle different recognition sites driven by a panoply of noncovalent interactions. In these switches, the superstructures corresponding to the threaded rings encircling each recognition site represent a subcategory of co-conformational isomers called translational isomers37,38,39.

The isolation of catenane translational isomers has been reported39 several times in the literature. In 1981, Schill and colleagues37 described the preparation of [3]catenanes consisting of a large ring containing two bulky aromatic units mechanically interlocked with a pair of identical smaller rings. The bulky aromatic units acted as covalent templates for the cyclization of the smaller rings during catenane formation; these sterically bulky components were large enough, however, to prevent ring translation in the final MIMs, resulting in two distinct regions for ring occupancy. The resulting [3]catenanes can exist as three possible translational isomers. One of these translational isomers, in which a single smaller ring occupies each region, was separated chromatographically from a mixture of the two translational isomers in which both of the smaller rings occupy the same region.

In the case of rotaxanes, translational isomers are often encountered in the literature40,41,42,43,44,45,46,47,48,49,50,51,52, their isolation as pure compounds, however, is rarely the focus of study. More commonly, such rotaxanes are designed to undergo interconversion between different translational isomers40,41,42,43,44,45,46,47,48, a concept which is utilized extensively in molecular switches and artificial molecular machines (AMMs). While in principle, many of these isomers could be purified, in practice, stable rotaxanes translational isomers have been isolated only a small handful of times49,50,51,52.

In recent years, artificial molecular pumps (AMPs)—a class of non-equilibrium AMMs which accumulate rings against concentration gradients—have gained in popularity as a useful approach to the precise preparation of rotaxanes53,54,55,56,57,58,59,60. Here, we demonstrate the use of a dual-pump system to direct the formation of pure rotaxane translational isomers. The dual-pump architecture, specifically designed to enhance synthetic precision, offers expanded access to distinct isomeric states while maintaining operational simplicity through redox control. By leveraging radical-based interactions—orthogonal to the driving forces employed in most AMMs—this system establishes a versatile framework for the design of increasingly complex molecular machinery. This study underscores the potential of molecular pumps to orchestrate nanoscale assembly with a high degree of control, advancing their role in the construction of functional artificial systems.

Results

Design of a molecular dual pump

In designing rotaxane translational isomers, a minimum of two distinct regions for ring occupancy is required. A recently developed strategy for the preparation of novel rotaxanes is the use of molecular pumps. The most widely used pump cassette, called the Mark II54, comprises a dimethyl pyridinium Coulombic barrier (PY+), redox-switchable viologen recognition site (BIPY2+), and isopropylphenyl steric barrier (IPP). These units work in concert, leveraging radical pairing interactions to draw cyclobis(paraquat-p-phenylene) (CBPQT•4PF6) rings from solution, and accumulate them on a collecting chain during a redox cycle. All the published designs of radically-driven AMPs, however, contain only one single region for ring occupancy53,54,55,56,57,58,59,60. Fortunately, AMPs containing two pumping cassettes are well established. In one design, referred56 to as a dual pump, the first pumping cassette accumulates rings on a central collecting chain, while a pseudopump,61 lacking an IPP steric barrier, releases the rings back into bulk solution. Based on this model, an alternative dual pump has been designed (Fig. 1), incorporating an IPP steric barrier on a second collecting chain, thus providing the necessary two regions for ring occupancy.

Fig. 1: Graphical depiction of the design, and approximate conceptual energy diagrams detailing operation of the dual pump DP•6PF6 through redox cycling.
figure 1

DP•6PF6 comprises two pumping cassettes, composed of a 2,6-dimethyl pyridinium (PY+) Coulombic barrier, viologen (BIPY2+) recognition sight, and an isopropylphenyl (IPP) steric speed bump, serviced by collecting chains 1 and 2. Initially, the cationic DP•6PF6 and CBPQT•4PF6 are repelled. Once reduced, the decreased charge of CBPQT2(•+) allows rings from solution to thread over the PY+ barrier onto the first recognition site to form a trisradical tricationic complex. Oxidation removes the radical stabilization, and Coulombic repulsion from the PY+ unit prevents dethreading of the CBPQT4+, so the ring is compelled to move to the right giving the [2]rotaxane isomer with a ring located on the central collecting chain (R1•10PF6). Reduction of this [2]rotaxane in the presence of free CBPQT•4PF6 results in the acquisition of an additional ring from solution, as the steric bulk of the IPP unit prevents return of the already threaded ring to the first recognition site. Subsequent oxidation, and electrostatic repulsion of the PY+ barriers results in both rings moving to the right, yielding the [3]rotaxane isomer (R3•14PF6) with individual CBPQT4+ rings located on both the central and terminal collecting chains.

Full size image

The mechanism of pump operation is also illustrated in Fig. 1. Initially, the cationic CBPQT•4PF6 rings and pump molecules are repelled. Upon reduction, the decreased charge of the CBPQT2(•+) rings attenuates the repulsion of the PY+ barrier, allowing the rings to thread onto the pump and become captured in a trisradical-tricationic complex62 (BIPY•+⊂CBPQT2(•+)) with the BIPY•+ recognition site of the first pump. Oxidation restores the full repulsive influence of the PY+ unit, preventing the CBPQT4+ dethreading, causing the ring to overcome the IPP unit and come to encircle the central collecting chain. When this rotaxane is reduced, the IPP unit blocks the threaded CBPQT2(•+) from returning to the first pump, so it moves to the right and forms a trisradical complex with the second recognition site. If there are free rings in solution, an additional CBPQT2(•+) is acquired by the first pumping cassette. Oxidation again drives the rings to the right. The second IPP unit allows the centrally located ring to be driven on to the terminal collecting chain by a further redox cycle in the absence of free CBPQT•4PF6.

Synthesis of the dual pump

The synthesis of dual pump DP•6PF6 was achieved (Fig. 2) in two steps from known starting materials54,56,60. In the first step, the terminal collecting chain (S2) was coupled under microwave irradiation to the central pumping cassette (S1•2PF6) in a yield of 51%. The final step, a copper-catalyzed azide-alkyne cycloaddition (CuAAC), gave the dual pump DP•6PF6 in a 74% yield. The pump was characterized by 1H NMR (Fig. 3) and 2D NMR spectroscopy. Electrospray ionization mass spectrometry (ESI-MS) analysis confirms the identity of the dual pump with the unfragmented parent ion [DP − 2PF6]2+ being observed at m/z of 1005.4377 compared to the theoretical 1005.4326. See Supplementary Information.

Fig. 2: Synthesis of the dual pump DP•6PF6. The azide56 S1•2PF6 and the bromide60 S2 were dissolved in DMF and heated to 40 °C for 4 h under microwave irradiation to afford S3•3PF6 in 51% yield.
figure 2

The azide S3•3PF6 was subjected to a copper-catalyzed click reaction in DMF with the propargyl ether54,60 S4•3PF6 in the presence of tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA). After workup the dual pump DP6+ was isolated as a yellow solid in 71% yield as the PF6− salt.

Full size image
Fig. 3: Assigned and color-coded 1H NMR (600 MHz, CD3CN, 298 K) spectrum and labeled structural formula of DP•6PF6.
figure 3

The aromatic region is enlarged to clarify the peak assignments.

Full size image

Preparation of the rotaxanes

Using the dual pump DP•6PF6 the translational isomers R1•10PF6 and R2•10PF6 of the [2]rotaxane, and translational isomers R3•14PF6 and R4•14PF6 of the [3]rotaxane were synthesized (Fig. 4). Rings are accumulated on the central collecting chain by a single pump cycle in the presence of free CBPQT•4PF6. Following a purification step, the centrally located CBPQT4+ rings are driven onto the terminal chain by a second pump cycle in the absence of any free CBPQT•4PF6, resulting in an alternative [n]rotaxane translational isomer. It is perhaps worthy of mention that in the case of the [2]rotaxanes, only R1•10PF6 and R2•10PF6 are possible given the availability of a single ring and two collecting chains. In contrast, for the [3]rotaxanes, a third isomer in which both rings are located on the central collecting chain is geometrically plausible. With this particular design of dual pump, however, it is not possible to prepare this third isomer as any pump cycle that recruits a ring from solution will also result in the movement of any centrally located macrocycle to the terminal collecting chain.

Fig. 4: Graphic depiction of the preparation of [2]rotaxane and [3]rotaxane translational isomers using the dual pump.
figure 4

Pumping cycles were performed in acetonitrile (MeCN) solutions under an Argon atmosphere and entail reduction using cobaltocene (CoCp2) at room temperature to thread rings onto the recognition site, followed by oxidation with AgPF6 at 50 °C to allow the CBPQT4+ rings to overcome the IPP barriers. Cycling of DP•6PF6 in the presence of CBPQT•4PF6 yields the M[2]rotaxane R1•10PF6, which, when redox cycled in the absence of free rings is converted to the complimentary T[2]rotaxane translational isomer R2•10PF6. Redox cycling either [2]rotaxane in the presence of CBPQT•4PF6 results in the MT[3]rotaxane R3•14PF6, which can in turn be cycled to yield the TT[3]rotaxane translational isomer R4•14PF6.

Full size image

Pumping was performed under an N2 atmosphere, in acetonitrile solvent, with CBPQT•4PF6 rings, using conditions optimized previously60. Specifically, cobaltocene (CoCp2) and silver hexafluorophosphate (AgPF6) were used as the reductant and oxidant, respectively, and pumping was performed at high concentrations to improve efficiency. Protocols for the preparation of each rotaxane are given in the Supplementary Information.

Initial preparation of R1•10PF6 under these conditions gave a purified yield of 88% after reverse phase chromatography. Threading of the first ring was confirmed by ESI-MS, which revealed the parent ion [R1 − 2PF6]2+ at m/z 1555.4946, closely matching the predicted 1555.4923. The rotaxane R2•10PF6 was prepared by redox cycling R1•10PF6 in the absence of free CBPQT•4PF6, followed by precipitation and counterion exchange. This procedure gave pure R2•10PF6 in 64% yield without the need for chromatography as there was no residual R1•10PF6. This high efficiency is likely a result of preorganization of the ring and pumping cassette in R1•10PF6. ESI-MS analysis found a parent ion [R2 − 2PF6]2+ at m/z of 1555.4936, in good agreement with the theoretical 1555.4923, demonstrating that the product is an isomer of R1•10PF6.

Given that backwards motion is prevented for rings on both chains, R3•14PF6 could be prepared from either R1•10PF6 or R2•10PF6. Starting from R1•10PF6 was more convenient as it could be produced in a single pump cycle. Thus, R3•14PF6 was synthesized by redox of R1•10PF6 and CBPQT•4PF6 in acetonitrile, followed by reverse phase chromatography to achieve a purified yield of 83%. Mass spectrometry confirmed the threading of a second ring, where the [R3 − 3PF6]3+ ion was at m/z of 1355.3846, in line with the expected 1355.3798. In the same vein as for the preparation of R2•10PF6, it was expected that the preparation of R4•14PF6 may occur at very high efficiency. After filtration and precipitation in water to remove redox byproducts, however, the resulting material was determined to be R4•14PF6, with significant R3•14PF6 contamination. According to NMR integration this pump cycle was 85% efficient, typical for the Mark II pumping cassette54. The loss of efficiency can most likely be attributed to Coulombic repulsion of the terminal ring, which implies that the maximum capacity of this length of collecting chain has been reached. The preparation of R4•14PF6 was repeated, but using purification by reverse phase chromatography; the product was isolated as the pure rotaxane in 60% yield. Product identity was confirmed by ESI-MS, which found the [R4 − 4PF6]4+ ion at m/z of 980.2930, in agreement with the calculated value of 980.2936.

Analysis of the rotaxane translational isomers

All rotaxanes were characterized by 1D and 2D NMR in addition to the ESI-MS analysis discussed above. Complete assigned spectral data for each is available in the Supplementary Information. Examination of the 1H NMR spectra of the products (Fig. 5) clearly reveal the formation of rotaxane translational isomers. In particular, changes in chemical shifts of the collecting chain protons, caused by shielding as a result of ring currents in the threaded rings, are clear evidence of CBPQT4+ units encircling the pump molecule.

Fig. 5: 1H NMR of the rotaxane translational isomers.
figure 5

Comparison of the 1H NMR spectra of (a) DP•6PF6 and its rotaxane translational isomers, (b) R1•10PF6, (c) R2•10PF6, (d) R3•14PF6, and e R4•14PF6. Spectra were recorded in CD3CN at 298 K and a field strength of 600 MHz. Key shifts demonstrating the mechanically interlocked nature of these compounds, and the changes between pumping cycles, are labeled. In particular, the presence of a CBPQT4+ ring on the central collecting chain, as in R1•10PF6 and R3•14PF6, is demonstrated by the upfield shift of H-16 of the linker and H-17 of the triazole unit from the aromatic region to 3.5–4.5 ppm. Rings on the terminal chain; as in R2•10PF6, R3•14PF6, and R4•14PF6; are evidenced by the upfield shifting of the H-42 to H-49 alkane resonances. The additional ring density of R4•14PF6 results in upfield shifting of H-35 of the second IPP unit.

Full size image

For example, the rotaxanes with a CBPQT4+ on the central collecting chain (R1•10PF6 and R3•14PF6) can be readily distinguished from pristine DP•6PF6 and the rotaxanes which lack a central ring (R2•10PF6 and R4•14PF6), by the ___location of signals of H-16 (belonging to the linker between the first IPP and the triazole unit) and H-17 of the triazole unit itself (shown in pink). In DP•6PF6, R2•10PF6, and R4•14PF6, the peaks for H-16 and H-17 are near their expected chemical shifts of 5.2 and 7.9 ppm, respectively. Conversely, for R1•10PF6 and R3•14PF6 these peaks are moved significantly upfield to 3.4 and 4.6 ppm. On the other hand, the presence of a ring on the terminal chain can be determined by the appearance of its associated peaks (primarily H-42 to H-49) in the very highfield region (<1 ppm). In the absence of a terminal ring (as is the case for DP•6PF6 and R1•10PF6), the signals for H-42 to H-49 appear as an overlapping multiplet between 1.4 and 1.3 ppm. In the presence of a single CBPQT4+ ring encircling the terminal chain, the resonances for H-42 to H-49 become resolved into individual peaks between 1.1 and −1.0 ppm. Addition of another ring to this chain results in movement further upfield, with peaks for the H-39 to H-51 protons now observed in the region of 1.1 to −1.87 ppm.

Comparing each rotaxane translational isomer, between R1•10PF6 and R2•10PF6 in addition to the changes outlined above, the H-P resonance is seen to move upfield, while those of H-α and in particular H-β are moved downfield. These observations indicate a change in shielding between the two isomers, which can be explained by the presence of the aromatic triazole group. In R1•10PF6, the CBPQT4+ ring encircles the triazole, and because of similar ring currents impacting the collecting chain peaks, H-α and H-β are shielded. This shielding moves their peaks upfield, indicating that these protons are oriented in a face-to-face geometry with the triazole, while the H-P resonances are deshielded because of their side-on orientation with the triazole moving them downfield. When the ring is pumped onto the terminal chain, which lacks a triazole (as in R2•10PF6), the shifting of their peaks is negated, resulting in the different peak positions in the two isomers. This behavior is corroborated by the chemical shifts of protons in R3•14PF6, which contain the CBPQT4+ peaks associated with both R1•10PF6 and R2•10PF6.

More significant alterations occur between R3•14PF6 and R4•14PF6. The H-35 peak of the second IPP unit is shifted out of the aromatic region (7.1 ppm in R3•14PF6) upfield to 3.95 ppm. Such a large movement indicates close proximity of the second CBPQT4+ ring (located at the left in the figure) to the second IPP. Moreover, the CBPQT4+ peaks of R4•14PF6 are much broader than in the other rotaxanes. This phenomenon was observed, but not explained, in the case of previous CBPQT4+ [3]rotaxanes produced using AMPs53,54,60. Given the nature of the compounds for which these broad peaks have been observed, and the fact that broadening occurs for R4•14PF6 and not R3•14PF6, it can be reasoned that the broadness of the peaks is a result of the nanoconfinement of multiple tetracationic cyclophanes. In 1H NMR, broad peaks are often the result of exchange processes. The aromatic protons associated with the broad peaks would not be anticipated to undergo chemical exchange in the absence of a strong base. Given that the resonances in question are associated with equally sized, rigid CBPQT4+ units, the interchange (i.e., physical exchange) of the rings with each other is unlikely on geometric grounds. In addition, no interchanging of the different rings was observed in related hetero[3]rotaxanes60. The most probable explanation then, is that there are multiple conformations of R4•14PF6 which undergo exchange in solution at ambient temperature. Other factors, including loss of symmetry, may contribute to the observed broadening. Additional work is needed, then, before any conclusion can be reached on the origin of this peak broadness. In the interim, the 1H NMR results give strong evidence for the formation of both [2] and [3]rotaxane translational isomers using the dual pump.

Beyond 1H NMR, 2D NMR gives further insight into the structure of the rotaxanes. Diffusion-Ordered Spectroscopy (DOSY) reveals that in all rotaxane samples, the peaks observed in 1H NMR, including the broad peaks of R4•14PF6, come from pure individual compounds (Fig. 6). In decreasing order of diffusion constant (D) are DP•6PF6 (6.4 × 10−6 cm2s−1), R1•10PF6 (5.1 × 10−6 cm2s−1), R2•10PF6 (4.8 × 10−6 cm2s−1), R3•14PF6 (4.4 × 10−6 cm2s−1), and R4•14PF6 (4.2 × 10−6 cm2s−1). Further, the results demonstrate that the addition of a CBPQT4+ ring to DP•6PF6 to form R1•10PF6, or to R2•10PF6 to form R3•14PF6, causes a decrease in the diffusion constant, in line with the increased molecular weight of the compound. Moreover, the data also indicates that the diffusion rate is lower when the CBPQT4+ rings are located on the terminal chain when compared to the central chain, i.e., for R2•10PF6 compared to R1•10PF6, and R4•14PF6 compared to R3•14PF6. The cause of this decrease in diffusion by rings on the terminal chain is uncertain, however, it may be that Coulombic repulsion between centrally located rings and the pumping cassettes enforces a more linear, and therefore more rapidly diffusing geometry on the rotaxanes.

Fig. 6: DOSY NMR (600 MHz, CD3CN, 298 K) spectra of DP•6PF6, R1•10PF6, R2•10PF6, R3•14PF6, and R4•14PF6, labeled with their diffusion constants.
figure 6

A trend of decreasing diffusion constant with increasing rings, and particularly their presence on the terminal chain is evident.

Full size image

Conclusive evidence of the formation of the rotaxane translational isomers is found in the nuclear Overhauser effect spectroscopy (1H-1H NOESY) data, which display the through space interactions of the protons associated with the peaks in the 1H NMR spectrum. For the [2]rotaxanes, the 1H-1H NOESY spectra (see Fig. 7 for partial spectra containing the key crosspeaks, complete spectra are given in the Supplementary Information) clearly reveal that the CBPQT4+ ring encircles a different collecting chain for each isomer. For R1•10PF6, crosspeaks are evident between the H-α, H-β, and H-P of the ring, and H-15 of the first IPP unit (green), the H-16 linker resonance, the H-17 triazole proton, and H-18 to H-20 of the central collecting chain. In contrast, for R2•10PF6, couplings are apparent between the H-α, H-β, and H-P of the ring, and H-40 to H-51 of the terminal collecting chain, in addition to H-53 of the S unit.

Fig. 7: Partial 1H-1H NOESY spectra (600 MHz, CD3CN, 298 K) of the [2]rotaxane translational isomers R1•10PF6 and R2•10PF6, which have been made with the dual pump.
figure 7

Peaks of interest are labelled on the projected 1D spectra. Key crosspeaks between the CBPQT4+ rings with the central collecting chain (H-15 to H-20) in the case of R1•10PF6, and the terminal collecting chain (H-40 to H-51 and H-53) for R2•10PF6 are indicated in boxes. These crosspeaks demonstrate the threading and ___location of the rings in these [2]rotaxane translational isomers.

Full size image

Similarly, the 1H-1H NOESY confirms the formation of separate isomers in R3•14PF6 and R4•14PF6. For R3•14PF6 (partial spectra are given in Fig. 8a and complete spectra in the Supplementary Information), the resonances of the CBPQT4+ ring on the middle chain (denoted by subscript M) have crosspeaks with H-15 of the first IPP unit, H-16 of the linker, H-17 of the triazole, and H-18 to 21 of the central collecting chain. In contrast, the ring on the second chain of the same rotaxane (subscript T) exhibit couplings with the terminal alkyl protons of H-40 to H-51.

Fig. 8: 1H–1H NOESY spectra of the [3]rotaxanes.
figure 8

Partial 1H-1H NOESY spectra (600 MHz, CD3CN, 298 K) of the [3]rotaxane translational isomers (a) R3•10PF6, and (b) R4•14PF6. Key peaks on the 1D projections are annotated in the complete spectra given in the Supplementary Information. The graphical depictions are labeled to present the subscripts denoting the different CBPQT4+ rings. a Subscript M denotes the ring on the middle chain, T the terminal chain. Crosspeaks between CBPQT4+ and both the central (H-15 to H-20) and terminal (H-41 to H-50) collecting chains indicate that the rings are located in distinct regions of the DP6+ pump. b Subscript IPP denotes the ring closest to the IPP, and S the ring closest the stopper. Crosspeaks between CBPQT4+IPP with H-35 and H-37 of the second IPP unit, and the H-38 to H-41 alkane protons reveal that this ring is located on the first half of the terminal collecting chain. Crosspeaks between CBPQT4+S with the H-46, H-47, H-50, H-51, and H-53 resonances indicate that this ring is positioned on the second half of the terminal collecting chain.

Full size image

In the same fashion, in the 1H-1H NOESY spectra for R4•14PF6 (Fig. 8b), the leftmost CBPQT4+ ring (denoted by subscript IPP) has clear crosspeaks with H-35 and H-37 of the second IPP unit, and with H-38 to H-41 belonging to the first section of the terminal chain. The peaks of the CBPQT4+ closest to the S, although less clear, are evident between H-βS with alkane resonances H-46 to H-47 and H-50 to H-51. In addition, H-PS has visible crosspeaks with H-50 and H-51, all located on the second half of the terminal chain.

Discussion

In this study, we have demonstrated the ability of a dual molecular pump to achieve the controlled assembly of co-conformationally pure translational isomers of [2]rotaxanes and [3]rotaxanes. By leveraging sequential redox-driven pumping cycles, we successfully directed CBPQT4+ rings along a defined molecular axle, enabling their precise localization on either the central or terminal collecting chain. This process resulted in the isolation of two pairs of distinct translational isomers in high yields, with each characterized comprehensively through 1D NMR, 2D NMR, and mass spectrometry. The 1H-1H NOESY spectra provided unambiguous evidence of the translational states of the CBPQT4+ rings, revealing spatially resolved crosspeaks that distinguished the isomers. DOSY measurements further complemented this analysis, offering quantitative insights into the hydrodynamic properties of the rotaxanes, which vary systematically with the number and ___location of the interlocked rings. The combined analyses underscore the ability to generate and isolate isomers with exceptional spatial accuracy and structural integrity. The successful isolation of translational isomers highlights the versatility and precision of molecular pumps in guiding nanoscale assembly. This work not only advances our understanding of co-conformational control in MIMs but also paves the way for future applications of molecular pumps in the creation of functional, positionally defined architectures. By demonstrating the power of these systems to achieve unparalleled geometric control, we have opened avenues for the development of molecular machines and devices capable of performing increasingly complex tasks at the molecular level.

Methods

Materials

All reactions of air and water-sensitive compounds were carried out under an N2 atmosphere in a glovebox. Chemical redox cycling was performed under a dry Argon atmosphere using a glovebox. All compounds were purchased from commercial sources and used without additional purification. All reagents were purchased from Sigma-Aldrich. Diethyl ether (Et2O), dimethylformamide (DMF), methanol (MeOH), and acetonitrile (MeCN) were dried over 3 Å molecular sieves. MeCN used in redox cycles was dried and degassed using the freeze-pump-thaw method. Other solvents, including acetone (Me2CO) and those used in extractions and chromatography were used as received. Detailed synthetic procedures and characterization of the precursors, pump, and rotaxanes are given in the Supplementary Information.

Synthetic instruments

Microwave assisted syntheses were carried out in Biotage initiator+ microwave reactor.

Column chromatography was performed using a CombiFlash 300 (Teledyne ISCO) for reverse-phase (RediSep Rf Gold Reversed-phase C18) separations.

NMR analysis

Nuclear magnetic resonance (NMR) spectra of the precursor were recorded on Bruker Avance III 500 spectrometer with a working frequency of 500 MHz for 1H and 126 MHz for 13C nuclei. NMR Spectra of the pump (DP•6PF6) and rotaxanes were taken on a Bruker Avance III HD 600 fitted with a TCI helium cryoprobe, with working frequencies of 600 MHz and 151 MHz for 1H and 13C nuclei, respectively. Chemical shifts are reported in ppm relative to the residual non-deuterated solvent signal CD3CN: δH = 1.94 ppm, δC = 118.3 ppm).

MS analysis

nESI-HRMS experiments were performed using an LTQ Orbitrap XL (Thermo Scientific, Waltham, MA). Positive voltages of 0.8–1.0 kV were applied to the nESI emitters in relation to the capillary entrance to the mass spectrometer. nESI emitters with tip sizes of 2.0–3.0 μm were used.