Single-molecule level control of host-guest interactions in metallocycle-C60 complexes

Host−guest interactions are of central importance in many biological and chemical processes. However, the investigation of the formation and decomplexation of host−guest systems at the single-molecule level has been a challenging task. Here we show that the single-molecule conductance of organoplatinum(II) metallocycle hosts can be enhanced by an order of magnitude by the incorporation of a C60 guest molecule. Mechanically stretching the metallocycle-C60 junction with a scanning tunneling microscopy break junction technique causes the release of the C60 guest from the metallocycle, and consequently the conductance switches back to the free-host level. Metallocycle hosts with different shapes and cavity sizes show different degrees of flexibility to accommodate the C60 guest in response to mechanical stretching. DFT calculations provide further insights into the electronic structures and charge transport properties of the molecular junctions based on metallocycles and the metallocycle-C60 complexes.

measurements (conductance and thermopower) have been made on C60-functionalized molecules in which it has been employed as a contact group (CA Martin et al, JACS 2008, 130, 13198;E Leary et al, Nano Letts. 2011, 11, 2236SK Lee et al Nano Letts. 2014, 14, 5276). It appears surprising that although molecule 5 does not interact strongly enough with C60 in solution to perturb the electronic spectrum, it does experience a conductance increase when its monolayer on Au is interacted with C60. This appears surprising and some discussion of why this should occur is warranted.
Incidentally, how is the y-axis in Figure 4(a-c) calibrated? The junction extensions of only ca. 0.4 nm seem to suggest that, even allowing for the usual snap-back distance, the molecules are not in fact being fully lifted in the junction (contrary to the cartoon in Fig. 2). Is it not possible that the conductance increase is indeed due to C60 interacting in some way with the molecule in the junction (either in the cavity for 6, or more peripherally for 5), but as the molecule is lifted, the C60 adheres strongly to the gold surface and it is actually this which causes the junctions to break down?
In the theory section of the paper, T(E) curves are used to support the notion of (i) conductance increase on complexation and (ii) conductance decrease on molecular distortion. How well (or otherwise) do the theoretical conductance values for the molecules and their complexes correlate with experiment?
The complex between 6 and C60 is clearly a charge transfer interaction from the electronic spectroscopy data, so: does this result in a Fano resonance near the contact Fermi energy as seen by another of the corresponding authors Lambert (et al) in their recent paper on oligothiophene-TCNE charge transfer complexes (Nanoscale 2015, 7, 18949)? It is hard to tell from the Figures, as these are pyridine-contacted molecules and there is the usual proximity of the LUMO resonances to E(F) seen with this contact group, although there does seem to be an additional peak for the complex with 6. (Incidentally, why is the latter paper not referenced? It would seem to be relevant here).
In summary, this might be a suitable paper for resubmission to Nature Communications with (i) stronger evidence in favor of the mechanism suggested by the authors (ii) additional surface characterization of the molecular monolayers.
Reviewer #2: Remarks to the Author: There has been an increasing level of interest in determining the single molecular conductance of increasingly complex molecular and biomolecular structures and this has included structures that exploit weaker non-covalent interactions, with some of these studies being referenced (references 28-32). This present work extends such investigations to an interesting and chemically rather exquisite supramolecular assembly which involves the insertion of C60. Tang et al show here that conductance of organoplatinum(II) metallocycle hosts can be significant enhanced by the incorporation of a fullerene guest molecule. This is shown using the STM-break junction method (STM-BJ), a very highly cited and respected method for molecular conductance determination which was invented by one of the co-authors (Nongjan Tao) and Bingqian Xu. Conceptually new here is the demonstration that mechanically stretching the metallocycle-C60 junction with this technique can induce the expulsion of the fullerene guest from the metallocycle. This then provides a form of mechanical conductance switch. I support publication but would recommend some additions to underpin the STM measurements section as listed below: (1) The authors should demonstrate that the junction breaking distances are comparable for the with and without C60 cases. I can believe that the junction is fully elongated with the C60 in place as in the cartoon in Figure 2 (top right) but this may not be the case with the empty macrocycles which contain a plethora of possible STM tip attachment points and can also tilt towards the surfaces. This can be straightforwardly resolved by presenting conventional 2D histograms of junction conductance versus stretching distance to show that the empty macrocycle is placed at its break off point as shown the cartoon in Figure 2 (top left).
(2) I would also suggest presenting the 1 (and 2D mentioned above) histograms to show the G(0) peaks from the break junction breaking since this reinforces the data quality.
(3) On page 5 can the implications of the following statement for the STM-BJ results be further clarified "The failure to observe potential CT absorption band from the complexation of 5 with C60 is possibly caused by the relatively weak and shallow interaction between them. This is supported by the DFT calculation results discussed below." If the interaction is weak and shallow why is the host-guest complexation peak seen in figure 3d as just a large peak as for 6 + C60 in Figure 3e?
Reviewer #3: Remarks to the Author: This paper reports on the single-molecule conductance of a host-guest complex. Organoplatnium metallocycles are synthesized and their conductance measured with the STMBJ technique. The molecular conductance can be increased by one order of magnitude when a C60 molecule is encapsulated. Conductance-distance histograms for two metallocycles show two plateaus, which are interpreted as corresponding to the free host (low conductance) and host-C60 complex (high conductance). When increasing the tip distance the C60 molecule is released and the conductance of the free host is recovered. Theoretical calculations aid in the interpretation of measurements. The results are compelling and add to a growing number of papers investigating conductance of host-guest complexes. I recommend publication but some issues need to be addressed first.
The increase in conductance due to C60 seems convincing in view of the STMBJ histograms of figure 3. New conductance peaks are observed for the [5+C60] and [6+C60] complexes. The large cavity size of 7 suggests no sufficient host-guest interaction. The paper mentions mechanical control of hostguest interactions as an attractive strategy however the conductance switch described by the authors seems to take place in one direction only (release of C60). Can encapsulation be induced?
Control experiments with C60 solution showed no conductance peak, contrary to reports in the literature for this molecule, why?
Theoretical calculations demonstrate the potential to modulate host-guest interactions by changing host size. Transmission calculations show that C60 increases the conductance of the host providing an extra current pathway. Theoretical conductance is smaller than experimental values, in contradiction to the results of DFT calculations in the literature. This should have been discussed but was not mentioned. The enhancement of conductance by the C60 host is expected and is demonstrated by the calculations. However, there seems to be a fundamental disagreement between calculations and experiment where calculations show less conductance, not more.
The effect of donor-acceptor interactions in the metallocycle-C60 calculations needs some clarification. Host-guest charge transfer shifts the frontier orbitals of C60 (eg. through the intermediate configurations in figure 6). The shift in C60 HOMO is very large however the LUMO is unchanged. The LUMO of the C60 acceptor in the host-guest is not filled, why not?
A 100 times increase in transmission was obtained when the Au-N bond was almost perpendicular to the plane of the pyridine ring. Although intriguing, without any additional information it is not possible to judge whether this geometry is in any way relevant or energetically favorable. Is such a change in conductance observed in the STMBJ experiments?
Further analysis and discussion are necessary.

Reviewer# 1
The field of molecular electronics was revived by the development of reliable methods for making and characterizing metal-single molecule-metal junctions about 15 years ago (including a scanning probe microscopy-based method pioneered by one of the corresponding authors of this submission). Since then, the science of single molecule junction electrical, thermal and spintronic properties has advanced greatly. Probing the single molecule behavior of supramolecular species is an avenue that is ripe for exploration and is now developing. In this submission, the electrical properties of junctions with some supramolecular Pt organometallic assemblies and their host-guest complexes with C60 are examined. The authors claim that on junction extension, strain causes the molecule: C60 complexes to dissociate, resulting in a conductance decrease. This would be suitable for publication in a general interest journal such as Nature Communications if the data were suitably strong. Unfortunately there are some doubts that this is so, and publication is not recommended in the present form, because of the weaknesses identified below.
The supramolecular coordination complexes are well-characterized. Their interaction with C60 in solution are examined; molecule 5 with the smallest cavity does not significantly interact with C60 in solution because there is no perturbation of the electronic spectrum indicative of a charge transfer interaction, but complex 6 does interact because a new intermolecular charge transfer band appears in the spectrum; the formation constant is high (over 105). Complex 7 in contrast is too large and again, with C60 there is no significant interaction in solution.
The first issue is that the molecules are all coordination complexes of Pt(II) with aryl ligands, phosphines, and pyridines. The latter in particular might represent a weak point when such molecules interact with a bare gold surface. For a paper making a high impact claim like this, some independent evidence that the molecules remain intact upon adsorption to gold surfaces should be obtained (e.g. surface vibrational spectroscopy, XPS and/or NEXAFS etc.) so that one can be confident that junctions are indeed forming with intact molecules and that the higher conductance values are indeed due to C60 complexes.
1. Reply: We thank the reviewer for these constructive suggestions, and have performed XPS and AFM experiments to prove that the metallocycles remain intact on Au surfaces. XPS results of metallocycle monolayers on Au surfaces agree well with those of metallocycle powders (shown in Table. 1-2, Figs. 1-4, and as added to SI in Table. S1-S2, Figs. S24-S27 ), suggesting that the metallocycles are intact after they are absorbed onto Au surfaces. Moreover, the height changes of the resulting films are consistent with the size of the metallocycles in AFM images (Fig. 5,added to SI as Fig. S28), providing further evidences of the formation of metallocycles functionalized Au surfaces. We have discussed these results in the main article as follows:

XPS characterizations.
To prove that the metallocycles remain intact on Au surfaces, XPS characterizations were carried out on monolayers on Au surfaces and corresponding powders of 5 and 6. As shown in Figs. S24-S27, the selected signals of Pt4f, P2p, N1s and F1s are observed in both powders and monolayers. According to the further quantitative analysis (Tabs. S1-S2), the atomic percentages of the elements of monolayers agree well with the powders compositions. For example, the Pt/P/N/F atomic ratio of 5 monolayer (1.0/1.4/3.0/2.6) is consistent with that of the powder sample (1.0/2.1/2.7/2.7) within experimental error.
AFM imaging. AFM images of the bare Au surface and those modified with 5, [5 + C 60 ], 6, [6 + C 60 ] are shown in Fig. S28, with the root mean square roughness (rms) of 4.51, 1.29, 0.69, 0.70 and 0.55 nm, respectively. This suggests that the Au surface was slightly smoothed after being functionalized with metallocycles and metallocycle-C 60 complexes. The height changes of the films are in accordance with the size of metallocycles. Moreover, after the incorporation of C 60 , the films of the [5 + C 60 ] and [6 + C 60 ] complexes exhibit almost the same height changes relative to corresponding free metallocycles. These results not only suggest a perpendicular configuration of metallocycles but also provide insights into the formation of host-guest complexes on Au surfaces.  The break-off distances (q.v.) do not seem consistent with the likely full length of the molecules (although this could well be due to them failing to be fully lifted in the junction).
2. Reply: In this experiment, a linear voltage amplifier was used because of its high accuracy, but its conductance range is limited to ~4 orders of magnitude and gold quantum point contact conductance (G 0 ) falls out of the measurable range of the linear amplifier. To obtain the most accurate conductance measurement with the setup, we thus adjusted the amplifier gain and bias voltage such that the current in the metallocycles and their host-guest complexes is in the middle of the amplifier range (1.410 -7 G 0 ~1.410 -3 G 0 for 5 and 6, and 710 -8 G 0 ~710 -4 G 0 for 7). To study the configuration of metallocycles or host-guest complexes on Au surface, we performed AFM analysis as stated. The AFM analysis shows the preferable perpendicular geometry of the metallocycles and their host-guest complexes.
The very large size of the molecules, and the fact that they all contain two meta-oriented contact groups, with resulting destructive quantum interference, means that low conductance values are expected, as is observed. It is claimed that the conductance increases by an order of magnitude or so on C60 exposure for molecules 5 and 6, but not for 7. Moreover, on junction extension, molecules 5 and 6 often revert to the low-conductance values characteristic of the free molecules, particularly at longer tip retraction distances.
The claim that molecular strain on junction extension causes C60 dissociation is potentially the most interesting observation in the paper, but to justify this significant claim, stronger evidence is needed. Some 'pull and hold' experiments might help here; does the complex persist in the high conductance state when the junction is not strained by pulling, just maintained at constant tip height in the junction, or does it in fact stochastically switch (as might be the case if, for instance, the C60 is being exchanged with the neighboring gold surface)? In that respect it is worth noting that, although the authors apparently found no evidence of junction formation using gold decorated with C60 alone, other workers have found that C60 interacts strongly with Au, forming self-assembled monolayers from solution (Yoshimoto et al., Langmuir 2002 18, 8518), and moreover, single molecule electrical measurements (conductance and thermopower) have been made on C60-functionalized molecules in which it has been employed as a contact group (CA Martin et al, JACS 2008, 130, 13198;E Leary et al, Nano Letts. 2011, 11, 2236SK Lee et al Nano Letts. 2014, 14, 5276). It appears surprising that although molecule 5 does not interact strongly enough with C60 in solution to perturb the electronic spectrum, it does experience a conductance increase when its monolayer on Au is interacted with C60. This appears surprising and some discussion of why this should occur is warranted.
3.1 Reply: We have performed the suggested 'pull and hold' experiments for [5 + C 60 ] and [6 + C 60 ] complexes. As shown in Fig. 6 (added to SI in Fig. S17), the higher conductance states of [5 + C 60 ] and [6 + C 60 ] last for more than 0.1 s and 0.2 s, respectively. This is different from the conventional break junction measurements of these complexes, where the high conductance states last for less than 0.03 s (0.59 nm at 19.5 nm/s retraction rate). These results suggest that the conductance switching from higher to lower level is mechanically dependent. We have added the following description to the revised manuscript: 'Pull and hold' experiments were also performed for [5 + C 60 ] and [6 + C 60 ] complexes. As demonstrated in Fig. S17, the higher conductance states of [5 + C 60 ] and [6 + C 60 ] last for more than 0.1 s and 0.2 s, respectively. This is different with the conventional break junction measurements of these complexes, where the high conductance states last for less than 0.03 s. These results suggest that the conductance switching from higher to lower level is mechanically dependent. 3.2 Reply: We apologize for the typo in main text. The concentration of C 60 in our . We agree with the reviewer that a discussion is necessary for the absence of analytical features of the conductance measurement of C 60 modified Au surface. We emphasize that the concentration of C 60 solvent, the coverage of C 60 on Au surface was too low to be measured by break junction methods (ref as in 45-50). To verify the assumption, we modified the Au surface using a higher concentration of C 60 solution (7 mM), and removed the excess C 60 by transferring the surface to 1,1,2,2-tetrachloroethane shortly (ref 48). Fig. 7 (added to SI as Fig. S19) shows the conductance histogram of C 60 modified on Au surface, featured by two conductance peaks at around 0.2 G 0 and 0.5 G 0 , which is in agreement with the reported values (ref 45, 49).

Fig. 7.
Conductance histogram of C 60 , where the Au substrate was incubated with C 60 (7 mM) in 1,1,2,2-tetrachloroethane for 2 hours, and transferred to pure 1,1,2,2-tetrachloroethane for 1 minute, then rinsed with DI water. The blue arrows mark the conductance peaks from single C 60 molecules, and the red arrows marks the G 0 peak.
Incidentally, how is the y-axis in Figure 4(a-c) calibrated? The junction extensions of only ca. 0.4 nm seem to suggest that, even allowing for the usual snap-back distance, the molecules are not in fact being fully lifted in the junction (contrary to the cartoon in Fig. 2). Is it not possible that the conductance increase is indeed due to C60 interacting in some way with the molecule in the junction (either in the cavity for 6, or more peripherally for 5), but as the molecule is lifted, the C60 adheres strongly to the gold surface and it is actually this which causes the junctions to break down? 4. Reply: In Figs.4a-c, the y axis is the conductance, and the x axis is the relative stretching distance. Figs.4a-c are generated by detecting the beginning and end of conductance plateaus and aligning the beginning of each curve at the distance origin. Those parts before the junction formation and after the junction breakdown are excluded from the 2D histogram. As described in our reply to question #1, since G0 is out of the measurable conductance range, a conventional stretching distance, measured from the point contact of electrodes, is not available. As a result, the plateau length on conductance-distance traces represents the distance of the molecular junction being stretched from formation to breakdown. Fig. 8 (added to SI as Fig. S18) shows that the plateau length is the longest for 6, and the host-guest complexes demonstrate comparable, but slightly shorter plateau lengths than the free metallocycles. This is intuitively reasonable, because 6 is expected to have the largest flexibility when connected with electrodes, and the incorporation of C 60 decreases the junction flexibility. In the theory section of the paper, T(E) curves are used to support the notion of (i) conductance increase on complexation and (ii) conductance decrease on molecular distortion. How well (or otherwise) do the theoretical conductance values for the molecules and their complexes correlate with experiment? 5. Reply: We thank the reviewer for this point.
(i) Upon complexation with C 60 , the theoretical conductance values show an increase in conductance compared with the bare metallocycles. This trend correlates well with our experiments and is insensitive to the binding configuration of the molecule to the electrodes.
To illustrate this last point, we have provided new results in Fig. 9 and 10 below (added to the SI as Figs (ii) Regarding the conductance decrease due to distortion, as noted in the text above, and below figure 5 in main manuscript, the conductance of the distorted host 5' is lower than that of the undistorted host 5, whereas the conductance of the (slightly) distorted host 6' is a little higher than that of the undistorted host 6. Therefore, we conclude that distortion affects the conductance of the host, but the shift could be either to higher or lower conductance. The distorted geometries are shown in more detail in Fig. 11 (added to SI as Fig. S35) below. The slightly higher conductance of 6' compared with 6 is due to a combination of a number of small geometric effects. For example, the shorter distance 1.9 nm for 6' between the two N atoms connected to gold electrodes compared to 2.1 nm for 6 indicated by the vertical yellow lines and the smaller spacing between the -Pt-P(CH 3 ) 3 groups (indicated by the two dashed ellipses) in 6' could account for the slight increase. The shorter distance and smaller spacing are due to the attraction of C 60 in [6+C 60 ] complex before removing C 60 . In contrast, the lower conductance of 5' compared to 5 in Fig. 5g is due to the severe deformation of the planar geometry from the lateral view. In this case N-N distances differ significantly (they are 1.81 nm and 1.78 nm for 5 and 5' respectively) and C 60 cannot enter the host fully as shown by Fig.  5b of the main script. In the experiment, the distortion effect cannot be isolated from the effect of guest binding, because distortion of the host occurs at the same time as binding of the C 60 .
To address this point, we have modified the text above Fig. 5 of the main manuscript as follows: After removing C 60 from the host-guest complex junctions, while freezing the metallocycle, the transmission functions of the distorted hosts 5' and 6' (shown by the yellow curves in Fig.  5g, h) are lower than those of [5 + C 60 ] and [6 + C 60 ]. This indicates the presence of an extra current path in the complexes due to the presence of C 60 . Fig. 5g reveals that the transmission of 5' is also significantly lower than that of the undistorted 5, which shows that the distortion of the host (shown in Fig. 5b) can also have a significant effect on the conductance. On the other hand, Fig. 5h reveals that the transmission of 6' is much closer to that of 6, which shows that the smaller distortion of this host (shown in Fig. 5d) has a less significant effect on the conductance. Since the transmission of 5' is lower than 5, while the transmission of 6' is slightly higher than that of 6, we conclude that the distortion of the host can either increase or decrease the conductance and the effect of distortion is more pronounced in 5' than in 6'.
More geometric details are shown in Fig. S35.   The complex between 6 and C60 is clearly a charge transfer interaction from the electronic spectroscopy data, so: does this result in a Fano resonance near the contact Fermi energy as seen by another of the corresponding authors Lambert (et al) in their recent paper on oligothiophene-TCNE charge transfer complexes (Nanoscale 2015, 7, 18949)? It is hard to tell from the Figures, as these are pyridine-contacted molecules and there is the usual proximity of the LUMO resonances to E(F) seen with this contact group, although there does seem to be an additional peak for the complex with 6. (Incidentally, why is the latter paper not referenced? It would seem to be relevant here).
6. Reply: We thank the reviewer for highlighting this point. In contrast with the oligothiophene-TCNE charge transfer complexes discussed in the above Nanoscale paper, charge transfer does not result in Fano resonances, because Fano resonances occur when a pendant orbital sits orthogonal to the current path and weakly couples to the current-carrying backbone. In our host-guest complexes, the bound states on the C 60 form part of the current path and therefore lead to Breit-Wigner resonances, rather than Fano resonances. On the other hand, as discussed in the above Nanoscale paper, charge transfer can cause energy levels on the donor or acceptor to move towards the Fermi level and therefore the resonance (Fano or otherwise) tends to appear near E F . This effect is clearly present in the red curves of Fig. 5 of the main manuscript, where the resonances associated with the C 60 LUMO are pinned close the DFT-predicted Fermi energy.
To clarify this point, the text has been included in the discussion below Fig. 6 of the main manuscript: In contrast with the oligothiophene-TCNE charge transfer complexes 56 , the host-guest charge transfer does not result in Fano resonances, because Fano resonances occur when a pendant orbital weakly couples to a current-carrying backbone and sits orthogonal to the current path. In our host-guest complexes, the bound states on the C 60 form part of the current path and therefore lead to Breit-Wigner resonances, rather than Fano resonances. However, as discussed in oligothiophene-TCNE complexes 56 , charge transfer can cause energy levels on the donor or acceptor to move towards the Fermi level and therefore the resonance (Fano or otherwise) tends to appear near E F . This effect is clearly present in the red curves of Fig. 5, where the resonances associated with the C 60 LUMO are pinned close the DFT-predicted Fermi energy. In contrast, the resonances just below the shaded regions (between -1.2 and -0.8 eV for [5 + C 60 ] and between -1.0 and -0.5 eV for [6 + C 60 ]) are due to states located on the host. Therefore, these HOMO resonances are sensitive to the gating effect of the negatively charged C 60 guest. Consequently, they move down in energy as the C 60 is systematically moved away from the host.
In summary, this might be a suitable paper for resubmission to Nature Communications with (i) stronger evidence in favor of the mechanism suggested by the authors (ii) additional surface characterization of the molecular monolayers.
7. Reply: As stated in the point-by-point reply above, we have presented additional evidence in favor of the suggested mechanism. The molecular monolayers were further studied by XPS and AFM techniques.

Reviewer# 2
There has been an increasing level of interest in determining the single molecular conductance of increasingly complex molecular and biomolecular structures and this has included structures that exploit weaker non-covalent interactions, with some of these studies being referenced (references 28-32). This present work extends such investigations to an interesting and chemically rather exquisite supramolecular assembly which involves the insertion of C60. Tang et al show here that conductance of organoplatinum(II) metallocycle hosts can be significant enhanced by the incorporation of a fullerene guest molecule. This is shown using the STM-break junction method (STM-BJ), a very highly cited and respected method for molecular conductance determination which was invented by one of the coauthors (Nongjan Tao) and Bingqian Xu. Conceptually new here is the demonstration that mechanically stretching the metallocycle-C60 junction with this technique can induce the expulsion of the fullerene guest from the metallocycle. This then provides a form of mechanical conductance switch.
I support publication but would recommend some additions to underpin the STM measurements section as listed below: (1) The authors should demonstrate that the junction breaking distances are comparable for the with and without C60 cases. I can believe that the junction is fully elongated with the C60 in place as in the cartoon in Figure 2 (top right) but this may not be the case with the empty macrocycles which contain a plethora of possible STM tip attachment points and can also tilt towards the surfaces. This can be straightforwardly resolved by presenting conventional 2D histograms of junction conductance versus stretching distance to show that the empty macrocycle is placed at its break off point as shown the cartoon in Figure 2 (top left).
1. Reply: We thank the reviewer for these constructive suggestions. As suggested, we generated the 2D conductance histograms of the free 5, 6 and 7 (Figs Fig. S18). As a result, the plateau length on conductance-distance traces represents the distance of the molecular junction when stretched from formation to breakdown. The analysis shows that the plateau lengths of the free metallocycles (5 and 6) are comparable to but slightly larger than those of the host-guest complexes ([5+C 60 ] and [6+C 60 ]), suggesting that at the breakdown point, the free metallocycles are placed as shown in main text Fig.2 (top left). (2) I would also suggest presenting the 1 (and 2D mentioned above) histograms to show the G(0) peaks from the break junction breaking since this reinforces the data quality.
2. Reply: As described in the response to Reviewer# 1's question, a linear voltage amplifier was used here to achieve high accuracy of conductance measurement, but its conductance range is limited to ~4 orders of magnitude and gold quantum point contact conductance (G 0 ) falls out of the measurable range of the liner amplifier..
(3) On page 5 can the implications of the following statement for the STM-BJ results be further clarified "The failure to observe potential CT absorption band from the complexation of 5 with C60 is possibly caused by the relatively weak and shallow interaction between them. This is supported by the DFT calculation results discussed below." If the interaction is weak and shallow why is the host-guest complexation peak seen in figure 3d as just a large peak as for 6 + C60 in Figure 3e?
3. Reply: The following discussion has been added in the revised manuscript: Although the absorption spectral analysis implies that the interaction between 5 and C 60 is weaker compared to that between 6 and C 60 , the metallocycle 5 can still hold C 60 with a shallower host-guest configuration according to our DFT calculations. This allows the complexed molecular wire, either in the case of [5 + C 60 ] or [6 + C 60 ], to open up the C 60related current path to enhance the conductance (see transmission calculation below).

Reviewer# 3
This paper reports on the single-molecule conductance of a host-guest complex. Organoplatnium metallocycles are synthesized and their conductance measured with the STMBJ technique. The molecular conductance can be increased by one order of magnitude when a C60 molecule is encapsulated. Conductance-distance histograms for two metallocycles show two plateaus, which are interpreted as corresponding to the free host (low conductance) and host-C60 complex (high conductance). When increasing the tip distance the C60 molecule is released and the conductance of the free host is recovered. Theoretical calculations aid in the interpretation of measurements. The results are compelling and add to a growing number of papers investigating conductance of host-guest complexes. I recommend publication but some issues need to be addressed first.
The increase in conductance due to C60 seems convincing in view of the STMBJ histograms of figure 3. New conductance peaks are observed for the [5+C60] and [6+C60] complexes. The large cavity size of 7 suggests no sufficient host-guest interaction. The paper mentions mechanical control of host-guest interactions as an attractive strategy however the conductance switch described by the authors seems to take place in one direction only (release of C60). Can encapsulation be induced? 1. Reply: We thank the reviewer for highlighting this point. Due to the limit of junction stability and kinetics of the encapsulation step, statistical measurement and analysis of the reencapsulation process is difficult. However, to elucidate the mechanical effect of conductance switching, we performed 'pull and hold'experiments, i.e., to stop the tip movement once a junction formation is detected. As shown in Fig. 6 (as shown in SI Fig. S17), the higher conductance states of [5+C 60 ] and [6+C 60 ] persist for more than 0.1 s and 0.2 s, respectively. This is in contrast with the conventional break junction measurements (19.5 nm/s retraction rate) where the high conductance states persist for less than 0.03 s (0.59 nm at 19.5 nm/s retraction rate) in any case. The above results confirm the assumption that the conductance switching from higher to lower level is mechanically dependent.
Control experiments with C60 solution showed no conductance peak, contrary to reports in the literature for this molecule, why?
2. Reply: The concentration of C 60 substrate with copious amounts of solvent, the coverage of C 60 was likely too low (ref as in 45-50). To verify this, we modified the Au surface using a higher concentration of C 60 solution (7 mM) and removed the excessive C 60 by transferring the surface to 1,1,2,2tetrachloroethane shortly (ref 48). Fig. 7 (added to SI as Fig. S19) shows the conductance histogram of C 60 modified on Au surface, featured by two conductance peaks at around 0.2 G 0 and 0.5 G 0 , which agrees with the reported values (ref 45, 49). As the reply of question 3.2 of Reviewer#1.
Theoretical calculations demonstrate the potential to modulate host-guest interactions by changing host size. Transmission calculations show that C60 increases the conductance of the host providing an extra current pathway. Theoretical conductance is smaller than experimental values, in contradiction to the results of DFT calculations in the literature. This should have been discussed but was not mentioned. The enhancement of conductance by the C60 host is expected and is demonstrated by the calculations. However, there seems to be a fundamental disagreement between calculations and experiment where calculations show less conductance, not more.
3. Reply: We thank the review for this point. Whereas the increase in conductance due to host-guest binding is independent of the shape of the electrodes, the predicted conductance values are not. Due to the large-scale nature of these simulations, it is only possible to sample a small number of tip configurations theoretically. Nevertheless, to address this point, we have carried out new simulations, in which one of the pyramidal electrode tips (used to obtain the results in Fig. 4 of the manuscript) is replaced by a flat electrode. Figs 9 and 10 show the new transmission curves and predicted conductances. Comparison between these results and Fig. 4 of the manuscript shows that although the conductance increase upon guest binding is resilient, the precise values of the conductances are sensitive to the shapes of the electrodes. The conductances with a flat substrate and pyramidal tip are closer to the experimental values. Therefore, we conclude that for our experiments a combination of a flat substrate electrode and pyramidal tip is encountered with high probability.
To address this point, the following text has been added to the manuscript in the paragraph preceding the conclusion and the new figures included in the SI as Figs. S36 and S37).
Whereas the increase in conductance due to host-guest binding is independent of the shape of the electrodes, the predicted conductance values are not. Due to the large-scale nature of these simulations, it is only possible to sample a small number of tip configurations theoretically. Nevertheless, to illustrate this point, we have carried out new simulations, in which one of the pyramidal electrode tips (used to obtain the results in Fig. 5) is replaced by a flat electrode. Figs. S36 and S37 show the new transmission curves and predicted conductances.
Comparison between these new results and Fig. 5 shows that although the conductance increase upon guest binding is resilient, the precise values of the conductances are sensitive to the shape of the electrode. The conductances with a flat substrate and pyramidal tip are closer to the experimental values. Therefore, we conclude that for our experiments a combination of a flat substrate electrode and pyramidal tip is a closer representation of the experimental setup.
The effect of donor-acceptor interactions in the metallocycle-C60 calculations needs some clarification. Host-guest charge transfer shifts the frontier orbitals of C60 (eg. through the intermediate configurations in figure 6). The shift in C60 HOMO is very large however the LUMO is unchanged. The LUMO of the C60 acceptor in the host-guest is not filled, why not? 4. Reply: We thank the reviewer for highlighting this point. As discussed in reply 6 to reviewer#1, in common with other charge transfer complexes, such as those discussed in [Nanoscale 2015 7, 18949], charge transfer causes energy levels of the donor or acceptor to move towards the Fermi level and therefore a resonance tends to appear near E F . This effect is clearly present in the red curves of Fig. 5 of the main manuscript, where the resonances associated with the C 60 LUMO appear close the DFT-predicted Fermi energy, just above the shaded regions. The LUMO of the C 60 (rather than the HOMO) moves towards E F , because the C 60 accepts a fraction of an electron from the host and becomes partially filled. Consequently the Fermi energy is located in the low energy tail of the lifetime-broadened LUMO. As an example from the literature (Phys. Rev. B, 63, 121104), studied a C 60 in direct contact with Au electrodes and showed that electrons were transferred into the triply degenerate LUMO of the C 60 . In other words, the LUMO is partly filled. In our case the C 60 is not in direct contact with the gold and we find that 0.62 (0.48) of an electron is transferred from the gold into the triply-degenerate LUMO of the C 60 when located in host 6 (5) respectively.
This pinning of the C 60 LUMO to the vicinity of the Fermi energy is why the LUMO does not shift through the intermediate configurations in Fig. 6. On the other hand, as discussed in the main text below Fig. 6, the resonances just below the shaded regions (between -1.2 and -0.8 eV for 5+C 60 and between -1.0 and -0.5 eV for 6+C 60 ) are due to states located on the host. Therefore these HOMO resonances are sensitive to the gating effect of the negatively charged C 60 guest, which increases their energy through electrostatic repulsion. Consequently they move down in energy as the C 60 is systematically moved away from the host.
This point is addressed by the following modified text in the main manuscript below Fig. 6.
….. as discussed in oligothiophene-TCNE complexes 56 , charge transfer can cause energy levels on the donor or acceptor to move towards the Fermi level and therefore the resonance (Fano or otherwise) tends to appear near E F . This effect is clearly present in the red curves of Fig. 5, where the resonances associated with the C 60 LUMO are pinned close the DFTpredicted Fermi energy. In contrast, the resonances just below the shaded regions (between -1.2 and -0.8 eV for [5 + C 60 ] and between -1.0 and -0.5 eV for [6 + C 60 ]) are due to states located on the host. Therefore, these HOMO resonances are sensitive to the gating effect of the negatively charged C 60 guest. Consequently, they move down in energy as the C 60 is systematically moved away from the host.
A 100 times increase in transmission was obtained when the Au-N bond was almost perpendicular to the plane of the pyridine ring. Although intriguing, without any additional information it is not possible to judge whether this geometry is in any way relevant or energetically favorable. Is such a change in conductance observed in the STMBJ experiments? 5. Reply: Thanks very much for raising this point. The "around a 100 times increase" mentioned in the paragraph before the conclusion is a typo and should be replaced by "around a 10 times increase." The geometry shown in Fig. S34, in which the Au-N bonds is almost perpendicular to the plane, is one of many that could be encountered during the evolution of a junction pulling curve. Since the electrodes are closer together in Fig S34, compared with the electrodes in Fig.  4 of the main text, we expect the former to be encountered before the latter during junction pulling. As the junction evolves from the former to the latter, we therefore expect the conductance to drop by a factor of approximately 10. Decreases of this order of magnitude are indeed present in the pulling curves of Fig. 2 of the main script.
To address this point, the typo has been corrected and the below text has been added to the paragraph above the conclusion.
…..Since the electrodes are closer together in Fig S34, compared with the electrodes in Fig. 5, we expect the former to be encountered before the latter during junction pulling. As the junction evolves from the former to the latter, we therefore expect the conductance to drop by a factor of approximately 10. Decreases of this order of magnitude are indeed present in the pulling curves of Fig. 2.
Further analysis and discussion are necessary.
6. Reply: Further analysis and discussion are presented as above point-by-point reply.

Reviewers' Comments:
Reviewer #1: Remarks to the Author: The authors have addressed all of the comments and suggestions of the referees in a satisfactory way.
In particular, the additional characterization data is useful and convincing. Although the question of why molecule 5 does not significantly complex C60 in solution, but does appear to do so in the junctions, is still an open one, the paper should now be published in its revised form. One caveat; the additional text in the revised version contains some grammatical errors and could benefit from a careful re-reading by the authors and/or editorial attention before final publication.
Reviewer #2: Remarks to the Author: I did not find the additional experiments added after review to be sufficiently convincing. In terms of the XPS the atomic percentages could approximately remain constant even after partial fragmentation on the surface. The AFM imaging is insufficient and would use best flat single crystal gold as a starting place rather than rough. The pull and hold experiments on junction stability are a nice addition but not decisive. Proper 2D histogram measured from a point contact of electrodes are not available and the 2D stretching ones of Figure 4 are not a substitute in terms of assessing whether the molecule is properly pulled in the junction. These data need strengthening before publication can be recommended.
Reviewer #3: Remarks to the Author: In this version some of the issues raised have been answered however other questions related to the calculations are not addressed.
While calculations taken on their own establish that complexation with C60 increases conductance, I am not convinced they represent the experimental situation.
The authors propose a combination of flat substrate electrode and pyramidal tip because they are closer to the experimental values. However it has been argued that pyridine groups bind preferentially to adatoms and not to flat substrates (Nature Nanotechnology volume 4, pages 230-234 (2009)).
Here there are no binding energies presented to show that the proposed binding is realistic. To make matters more confusing, the data of gold-molecule-gold calculations in the paper still have two pyramidal electrode tips despite the combination of flat substrate electrode and pyramidal tip favored by the authors. Similar analysis is missing also for the geometry in which the Au-N bond is almost perpendicular to the plane, what is the binding energy? There are no arguments supporting the claim that the adsorption geometry on a flat surface, or other geometries, are stable or favorable.
The paper does not really discuss how well the theoretical conductance correlates with experiment. In their reply, the authors just state that the precise values of the conductances are sensitive to the shape of the electrode, which is hardly a new result. The suggested geometries seem to be suggested solely to achieve a better agreement with experiment with no further analysis.
I am not convinced the theoretical calculations represent the geometries encountered in the experiments and I therefore cannot recommend publication.

Reviewer #1:
The authors have addressed all of the comments and suggestions of the referees in a satisfactory way. In particular, the additional characterization data is useful and convincing. Although the question of why molecule 5 does not significantly complex C60 in solution, but does appear to do so in the junctions, is still an open one, the paper should now be published in its revised form. One caveat; the additional text in the revised version contains some grammatical errors and could benefit from a careful re-reading by the authors and/or editorial attention before final publication.
Reply: We gratefully thank Reviewer #1 for his/her recommendation for the publication of our work. we have checked and corrected the grammatical errors in the manuscript.
Reviewer #2: I did not find the additional experiments added after review to be sufficiently convincing. In terms of the XPS the atomic percentages could approximately remain constant even after partial fragmentation on the surface. The AFM imaging is insufficient and would use best flat single crystal gold as a starting place rather than rough. The pull and hold experiments on junction stability are a nice addition but not decisive. Proper 2D histogram measured from a point contact of electrodes are not available and the 2D stretching ones of Figure 4 are not a substitute in terms of assessing whether the molecule is properly pulled in the junction. These data need strengthening before publication can be recommended.
Comments from reviewer#2 last review: "The authors should demonstrate that the junction breaking distances are comparable for the with and without C60 cases. I can believe that the junction is fully elongated with the C60 in place as in the cartoon in histograms of junction conductance versus stretching distance to show that the empty macrocycle is placed at its break off point as shown the cartoon in Figure 2 (top left)." 2 Reply: We thank the reviewer for suggesting the additional control experiment. We have now completed the experiment and determined 1D and 2D histograms with conductance ranging from electrode quantum point contact to the noise level with a logarithmic current-voltage converter. These histograms (for 5, 6, [5 + C 60 ] and [6 + C 60 ]) are shown in Fig. 1, and Fig. S21 of SI. As summarized in Table 1 (Table S1 of SI), the conductance values measured by logarithmic amplifier and linear amplifier are in agreement within 13.5% (which is within the expected difference between logarithmic and linear amplifiers).
The relative displacement distributions from electrodes' point of contact to junction break are shown in Fig. 2 (Fig. S22 of SI). After adding the Au-Au snap-back distance We also thank the reviewer's suggestion about characterizing the monolayers on a flat single crystal gold substrate. We have re-measured AFM images of 5 and 6 on Au(111) surfaces. As shown in Fig. 3 (Fig. S30 of SI), we were unable to clearly resolve the metallocycles due to the limited resolution of AFM and imperfect Au (111) surface. However, the heights of the monolayers on Au (111) Chem. Soc. 2012, 134, 2292−2304. On the other hand, these studies ignore important dispersion forces. When these are included (see J. Phys. Chem. C, 2013, 117, 4470) the binding energy to a flat gold (111) surface increases substantially to around 0.78 eV, which is higher than the binding energy claimed for the extended junctions in Nat. Nanotech. 2009, 4, 230-234. We entirely agree with the results of these studies and are delighted to refer to them.
To further address this point, we have calculated the binding energies to each electrode of the two geometries below, using the LDA(CA) exchange-correlation functional and counterpoise method (Mol. Phys. 1970, 19, 553-566) which eliminates basis set superposition errors (BSSE). Binding energies are calculated using the formula . All binding energies are significantly greater than k B T at room temperature, and therefore the junctions are stable (Fig.4, Fig. S41 of SI). To address this issue, we have included the following section in the SI: Binding energies to gold electrodes.   And the following sentences have been added in main manuscript.
The binding energy calculation shows -1.08 eV for pyramidal tip Au-N, -0.6 eV for atop Au-N with flat gold electrode, -0.53 eV for Au-N bond out of pyridine plane, which are significantly greater than k B T at room temperature and therefore the junctions shown in Fig. S37, S39, S40 are stable (see details in Fig. S41 of SI degrees between a N-Au bond and the pyridine plane are constructed for 5 and 6 molecules or [5 + C 60 ] after C 60 falls out and [6 + C 60 ] after C 60 falls out. The corresponding transport properties are shown in Fig. 6 (below). Red curves show the transmission functions of complexes, the blue depicts that of each optimal metallocycle and the yellow shows that of the distorted metallocycle obtained from optimal complexes by removing C 60 . This shows that the transmission function within the gap shifts to higher values compared to those of the fully extended junctions shown in Figure S33. The conductances reach to and respectively in a large range of Fermi energies (denoted by blue shaded regions) in Figure 6b, which are in great agreement with experimental data. The red curve is higher than the yellow one which reveals the additional current path due to the complexation of C 60 with the metallocycles. For molecule 6, the yellow curve is higher than the bl ue curve, which is mainly due to the smaller angle between the N-Au bond and pi orbitals, while for molecule 5, the yellow curve is close to the blue one, which is due to the large geometrical deformation caused by the complexation of C 60 and the different contact angles during the pulling process. With the above structures, all the trends of the measured transport properties are obtained and the values of the predicted conductances are improved compared with the previous estimates obtained for fully extended junctions. Initial structure corresponding to high conductance of [5 + C 60 ]. b. Optimal structure after geometrical relaxation. c. Initial structure of 5 or [5 + C 60 ] after C 60 falls out. d.