RETRACTED ARTICLE: Mechanically interlocked architecture aids an ultra-stiff and ultra-hard elastically bendable cocrystal

Molecular crystals are not known to be as stiff as metals, composites and ceramics. Here we report an exceptional mechanical stiffness and high hardness in a known elastically bendable organic cocrystal [caffeine (CAF), 4-chloro-3-nitrobenzoic acid (CNB) and methanol (1:1:1)] which is comparable to certain low-density metals. Spatially resolved atomic level studies reveal that the mechanically interlocked weak hydrogen bond networks which are separated by dispersive interactions give rise to these mechanical properties. Upon bending, the crystals significantly conserve the overall energy by efficient redistribution of stress while perturbations in hydrogen bonds are compensated by strengthened π-stacking. Furthermore we report a remarkable stiffening and hardening in the elastically bent crystal. Hence, mechanically interlocked architectures provide an unexplored route to reach new mechanical limits and adaptability in organic crystals. This proof of concept inspires the design of light-weight, stiff crystalline organics with potential to rival certain inorganics, which currently seem inconceivable.

. Here there are many apparently quite large changes in covalent bond lengths, and these appear to be in the authors' mind when noting on p. 11 that there is "weakening of these covalent bonds due to overall lattice perturbations arising from anisotropic mechanical stress". But there is a great deal known about changes in covalent (intramolecular) and noncovalent (intermolecular) separations from high pressure crystallography, where it has become very clear that even at pressures up to 7 GPa the molecules re-orient to reduce so-called void space as much as possible (as this requires much less energy). Only at extreme pressures do covalent bonds change dramatically. Certainly, bending of crystals is in many ways analogous to subjecting them to anisotropic pressures, but they seem unlikely to cause the changes in covalent bonds that are suggested by Table S4.
* The underlying reason for the often large changes to covalent bond distances in Table S4 can be identified in Figure S8: the refinement of anisotropic displacement parameters for heavy atoms in these refinements. The thermal ellipsoids in these figures hint to serious problems such as correlation between ADPs and atomic coordinates (and there appears to be a non-positive definite ADP for one of the caffeine C atoms in the M.i diagram, Figure S8(d)). I cannot understand why the authors chose to refine anisotropic ADPs, given the limitations of their diffraction data. (I am not questioning the quality of those microfocus experiments, or the integration, etc.). I note that in the related study by Worthy et al. (ref. 19) all atoms were treated isotropically. * p. 12 - Figure 4 (5?) contains energy estimates from CrystalExplorer for key intermolecular pairs in the ambient and three regions of a bent crystal. These energies are reported to 0.1 kJ/mol, which is unrealistic; I would never claim them to be meaningful to this degree (at best at least 1 or 2 kJ/mol). But worryingly, these energies are based on the experimental geometries derived from the refinements discussed above. As such, they will incorporate significant uncertainties due to the relatively poorly determined molecular geometries. I doubt that much meaning can be attached to the changes inferred from any apparent trend in these energies. (A further point: CrystalExplorer energies rely on a model of molecular geometries that standardizes bonds to H atoms to typical neutron values. The authors have not commented -in the paper of the supporting information -on whether this was performed).  Figure 3 contains 3 diffraction images, 12 histograms, nine microscope images, several schematic diagrams of unit cells, and a sketch of the relevant regions on a bent crystal). * Based on the above considerations, I cannot agree with the authors that "Molecules in elastically bent crystals are not only related by rotation along the arc but also show significant differences in their internal geometry and intermolecular interactions", p.14. Their own data does not demonstrate what I would call 'significant differences'.

Response:
We thank the reviewer for precise and scholarly evaluation and his/her observation that the strength of our manuscript lies in its technical quality. It is an honor and delight for us to note the reviewer's comment that our manuscript is suitable for the highly reputed journal like Nature Materials.
Reviewer #1 comment 1 -Statements like "These structural features lead to the large E and H values obtained on the (010) face" need to be toned down a bit to something like "These structural features may be responsible".

Response:
Considering the suggestion, we have now modified the text to "These structural features may be responsible for the large E and H values obtained on the (010) face." Reviewer #1 comment 2 -Some other statements seem a bit like hype. For example, "Notably, the tapes are nearly hydrophilic in the center while the exterior of the tapes is dominated by hydrophobic forces [ Fig. 2 (a)], resembling a composite structure,". The assignment of these structural features as composite-like is not advisable as innumerable organic crystals have features that resemble the separation of polar and non-polar domains. Moreover, the description of these domains as hydrophilic and hydrophobic is incorrect as there is no water here (it's quite common for these words to be misused in this way).

Response:
We thank the referee for this valuable suggestion. We have now changed this text to "Notably, the tapes in the center are populated with polar groups while the exterior is dominated mainly by van der Waals groups [ Fig. 2 (a)]." We have removed the lines related to "composites" and avoided direct description of the cocrystals as composites. However, we retained the text where we mentioned parallels between "composites" and co-crystals" for the following our argument. The properties seen in this cocrystal are drastically different compared to the individual parent compounds, much like what happens in "composites". This is intriguing and unexpected as typically the properties of cocrystals fall somewhere in the middle to that of the properties of parent components. We agree that there have been many earlier instances where crystals showed unexpected extraordinary properties, but not much research has been done to differentiate these from those typical cocrystals, which in the context of crystal engineering is important, for achieving extraordinary properties from simple starting molecular components. In this context, we attempted to draw the parallels between cocrystals and composites, but we will leave it for future study for gathering more evidence from systematic studies for proving this parallel between the cocrystals and the composites. We know the difference between the structural aspects (in terms of order and composition) of the two materials, but there is no reason why we can't compare such cocrystals (with extraordinary properties) with the composites. But we left this matter for future investigation.
The discussion on parallels between co-crystals and composites is as follows in the revised manuscript (same as in the original draft): "However cocrystals, which have been exploited to alter the physicochemical properties of pharmaceuticals, 31 ferroelectrics, 32 optoelectronics and charge transfer materials, 33 have not been perceived as "composite materials". Unlike conventional composites, which typically have at least two sub-lattices, cocrystals have a single lattice with homogenous distribution of constituent molecules. 34 The heterogeneity in cocrystals can be introduced at structural level by carefully exploiting directional and dispersive intermolecular interactions, to access new properties that are remarkably different from individual co-formers." Reviewer #1 comment 3 -Statements like "Please note that these are the best experimental structural models achieved so far from any type of bent molecular crystals." should be avoided.
Let the readers decide.

Response:
We removed this sentence and left the matter to the readers.
Reviewer #1 comment 4 Also, in the conclusion, the authors state "Report by Worthy et al.
suggest that the mechanical interlocking is not necessary for elastic flexibility in organic crystals. This is not entirely correct and needs to be discussed in the right context" contradicts the statement later: Crystal packing without mechanical interlocking and with weak interaction planes may allow some initial elastic flexibility (due to energetic barriers or elastic energy), but the absence of mechanical resistance to slip may eventually assist plastic deformation Response: We now changed this to "Worthy et al. suggested that the mechanical interlocking is not necessary for elastic flexibility in organic crystals. 19,48 This should be seen in the following context.", hence the text now reads well with the text quoted in the later part of referee's comment.
Reviewer #1 comment 5 -Last, to motivate and suggest that these materials will have impact on engineering materials is a bit far-fetched. The observations are interesting enough.

Response:
We feel that our suggestion of these materials for use in engineering applications is not far-fetched. The progress in engineering materials is currently going at a rapid pace (which may not be followed by the crystal engineering community). There have been several publications in recent literature emphasizing the need for such materials with high stiffness, flexibility and light weight (Mecklenburg et al. Adv. Mater. 2012, 24, 3486-3490;Zhang et al. Adv. Mater. 2007, 19, 4198-4201). Hence, one may start looking into these materials for developing further. These two articles are now cited in the revised draft (Refs. 54 and 55).

Reviewer #2 (Remarks to the Author):
Comment: In my opinion, the manuscript can be published Nature Communications.

Response:
We thank the referee for the positive assessment of our work.
Comment: I think that the manuscript is a bit too long and can be shorten, especially the Conclusions part.

Response:
We have now shortened the draft, particularly the conclusion part, which is now significantly short. We have also reduced the panels in Figure 3.
Comment: I also recommend to improve the English of the manuscript (see my comments in the attached file).

Response:
We thank the referee for the suggestions in the pdf version of the manuscript. We have corrected the text in all the places as per the suggestions, as detailed below.

Page No, Line No
As per Reviewer's suggestion, the text is changed from to (in the revised version)

Reviewer #3
In this paper the authors argue that the CAF:CNB:methanol 1:1:1 mixed crystal has an "exceptional mechanical stiffness" and "unprecedentedly high hardness". Further, they argue that this is due to the apparent mechanical interlocking nature of the hydrogen bond network CAF and CNB. These claims are based on nanoindentation studies and microfocus single crystal XRD experiments on a bent crystal.
These claims are certainly novel, and have the potential to be of considerable interest to researchers focusing on elastic, plastic and bent molecular crystals. However, I have numerous concerns with this submission, and I will address these through dot points, in the order in which they appear in the manuscript:

Response:
We thank the reviewer for the critical assessment of our manuscript and positive feedback. stands out as an alternative approach for achieving the high stiffness in crystals. However, we agree to delete change the sentence on "typical range" of 0.1-15 GPa for molecular crystals as it seems the upper limit of "typical range" would be somewhere close to 50 GPa (based on the articles by Haussül). We reorganized this section as below.
Deleted the following text: "For instance, typical elastic modulus (E) of molecular materials is 0.1 to 15 GPa 8 whereas even for low density metals such as magnesium (42 GPa experiments, subject to many important assumptions, but reported without any estimate of an experimental uncertainty. Only a range seems to be given, but surely there must be a reasonably rigorous way to estimate an uncertainty in these -apparently very important -quantities.

Response:
We have now defined the quantities, E and H, as follows, "E (elastic modulus, which is defined as a measure of resistance to elastic deformation)" ….. "H (hardness, which is defined as the measure of resistance to plastic deformation)." in the introduction part.
We agree that we gave only the range of E and H values for each of the crystal faces, which is also common in the nanoindentation literature. However, we now provided the standard deviations (see Table 1) and the process of estimating the same as detailed below.
From the nanoindentation experiments, the quantities E and H are derived from the load displacement curves using the well-established Oilver-Pharr method (which is given in detail in the Supplementary Information Section S1). Standard deviation is calculated using the following equation: Standard deviation or σ = ∑( Reviewer #3 comment 4: * p. 6 -The term "hydrophobic forces" is a commonly used misnomer. These are not 'forces' in the scientific sense, and 'hydrophobic' doesn't make sense here either. A more accurate and precise term is needed.

Response:
We now changed the sentence to "Notably, the tapes in the center are populated with polar groups while the exterior is dominated mainly by van der Waals groups [ Fig. 2 (a)]." Reviewer #3 comment 5: * pp. 6 and 7 -What does "knuckling" mean on these pages? Do the authors mean 'buckling'?

Response:
The double-sided comb-like tapes are arranged in a zipper-like architecture and 'knuckling' here refers to the pushing of the teeth of one comb-like tape towards the teeth of the adjacent tape. As shown in Figures 2 and 5, the torsion between the CAF and CNB molecules increases. We have drawn the analogy to the "knuckling" of fingers of both hands when folded, crossing each other.
Reviewer #3 comment 6: * p. 7 -"Please note that these are the best experimental structural models achieved so far from any type of bent molecular crystals". This sort of sentence is unhelpful in the paper -and I would argue that it is also inaccurate. On p. 9 the authors highlight  Table S4. And it definitely cannot be claimed that the (CNB) C-O bond elongates, if actual esds are quoted: 1.28 (3) to 1.31(3) Å. It is this sort of lack of rigor that significantly weakens much of the underpinning of the claims made in this paper.

Response:
We thank reviewer for the critical assessment of the bond distances and the text. We have deleted the text quoting "Please note that these are the best experimental structural models achieved so far from any type of bent molecular crystals". We mentioned this as this is the first case with anisotropic refinement.
Solely from X-ray diffraction point of view, one may argue that the changes in bond distances are not dependable considering the three times standard uncertainties (that are normally applied for standard ambient structures). For this reason, we used the complimentary data from the spatially resolved μ-IR and μ-Raman experiments for supporting the trends from our analysis of the results from X-ray crystal structures. This we presented in page 7 paragraph 2, as "The μ-IR and μ-Raman spectroscopy experiments have been performed to independently validate these results".
Later on in pages 10 and 11, we have discussed our μ-IR results and shown in Fig. 4(b1). We In page 9, we have revised the text and included "It is noteworthy that the changes in bond distances are within three times their standard uncertainties." Hence, we feel that the obvious concerns arising from the X-ray analysis at the extreme (bent) conditions have been addressed by IR and Raman studies.
Reviewer #3 comment 7: * The lack of critical assessment of bond distances in Table S4 made   me look more carefully at Table S4. Here there are many apparently quite large changes in covalent bond lengths, and these appear to be in the authors' mind when noting on p. 11 that there is "weakening of these covalent bonds due to overall lattice perturbations arising from anisotropic mechanical stress". But there is a great deal known about changes in covalent (intramolecular) and noncovalent (intermolecular) separations from high pressure crystallography, where it has become very clear that even at pressures up to 7 GPa the molecules re-orient to reduce so-called void space as much as possible (as this requires much less energy).
Only at extreme pressures do covalent bonds change dramatically. Certainly, bending of crystals is in many ways analogous to subjecting them to anisotropic pressures, but they seem unlikely to cause the changes in covalent bonds that are suggested by Table S4.

Response:
We have written "weakening of these covalent bonds due to overall lattice perturbations arising from anisotropic mechanical stress" to stress on the fact that the vibration bands of the M.m to M.i regions (in the IR spectra) are red-shifted with respect to the ambient crystal. Hence, although the bond strengths become enhanced from M.o to M.i region, they appear to be weaker (due to mechanical perturbation) as compared to ambient crystal.
We agree that it is reported for some structures at high pressures where molecules re-orient to reduce void-space. However, it is also possible that covalent bond distances change even as Also please see response to comment 6 regarding IR and Raman spectra.
Reviewer #3 comment 8: * The underlying reason for the often large changes to covalent bond distances in Table S4 can be identified in Figure S8: the refinement of anisotropic displacement parameters for heavy atoms in these refinements. The thermal ellipsoids in these figures hint to serious problems such as correlation between ADPs and atomic coordinates (and there appears to be a non-positive definite ADP for one of the caffeine C atoms in the M.i diagram, Figure S8 We have discussed isotropic refinement in detail and the problems in supporting information [Please see section S2, structure solution and refinement, Table S3, S4 and S5].
Reviewer #3 comment 9: * p. 12 - Figure 4 (5?) contains energy estimates from CrystalExplorer for key intermolecular pairs in the ambient and three regions of a bent crystal. These energies are reported to 0.1 kJ/mol, which is unrealistic; I would never claim them to be meaningful to this degree (at best at least 1 or 2 kJ/mol). But worryingly, these energies are based on the experimental geometries derived from the refinements discussed above. As such, they will incorporate significant uncertainties due to the relatively poorly determined molecular geometries. I doubt that much meaning can be attached to the changes inferred from any apparent trend in these energies. (A further point: CrystalExplorer energies rely on a model of molecular geometries that standardizes bonds to H atoms to typical neutron values. The authors have not commented -in the paper of the supporting information -on whether this was performed).

Response:
The figure number is corrected to Figure 5.
We thank reviewer for critically reviewing the energy framework values. We agree that the energies obtained from CrystalExplorer software are meaningful to 1 KJ/mol and not to 0.1 KJ/mol. We have now corrected the text by providing energies in whole numbers upto 1 KJ/mol (also n Figure 5).
It is to be noted that the structures obtained here are from non-ambient conditions and the level of accuracy can't be expected to match the ideal cases. In the given conditions, we did our best to provide the trends (supported by the μ-IR and μ-Raman) and I am sure readers are aware that these values can't be as accurate as for the ambient cases. In this situation, the anisotropic refinement yielded the best results in our case as far as bond distances and statistical parameters for refinement are concerned. Moreover, the π-stacking energies show average ~10% increase as compared to the ambient crystal which is a possible reason the elastic modulus E increases by ~10% in the bent crystal as compared to ambient (note that indent is done along c which is along the π-stacks. Hence stronger the π-stacking, more resistance it offers to elastic deformation upon application of external load). The trends from SCXRD are not surprising considering the large changes observed in the nanoindentation data (experimental) on ambient and bent elastic crystals.
We thank reviewer for pointing out the H atom treatment to typical neutron values. Although we did not explicitly mention in the supporting information, this is embedded in the software suite and we have performed our calculations standardizing bonds to hydrogen atoms to typical neutron values. In page 23, Materials and Methods, Energy frameworks calculations the text is re-written as follows, "using experimental crystal geometries, with X-H bond lengths are normalized to standard neutron diffraction values".

Response:
We thank reviewer for the critical assessment of the figures.
The strength of our article lies in the technical data and we feel that the figures are the best way to present it. One may shift most of these to supporting information, but that becomes even more tedious for cross-referencing. However, we have removed some panels from In Fig. 1, we present the nanoindentation load displacement curves, the corresponding 2D and 3D surfaces and their profiles obtained on ambient and bent crystal which is crucial for analysis of the mechanical properties.
In Fig 2, we present the structural architectures along different views to correlate the structural mechanical relations. The overlays help to describe the different kind of resistance to elastic deformations upon application of external load.
In Fig. 3, we have reduced the microscope images from 9 to 3. The variations in the unit cell parameters of the outer, middle and inner regions of center and shoulder of the bent crystal respectively including sections of diffraction pattern to emphasize the trends of the lattice parameters as function of regions and the differences are important.
In Fig. 4, we shown histogram plots of bond distances and how the different degrees of their variations are also seen in the IR spectra. Raman spectra are displayed where the differences between the center and shoulder of the bent crystal is apparent.
In Based on these arguments, we believe that these figures are relevant to the discussions in the main manuscript and to convey the relevant information to readers.
Reviewer #3 comment 11: * Based on the above considerations, I cannot agree with the authors that "Molecules in elastically bent crystals are not only related by rotation along the arc but also show significant differences in their internal geometry and intermolecular interactions", p.14.
Their own data does not demonstrate what I would call 'significant differences'.  Table S2. This information certainly supports our claims.
In page 9 paragraph 1 we have written "With respect to the volume of ambient crystal (V amb = 7664 Å 3 ), the average volume at M showed a contraction (-3.29%), but surprisingly a slight expansion was observed at L (1.55%) and R (1.60%)." In Fig. 4(b3) Raman spectra measured from shoulders and center of the crystal also show differences. For example, in the bent crystal, the band corresponding to C9-O4 at 1612 cm -1 (at ambient) is blue-shifted compared to that at the M region, while it is blue shifted with a shoulder at the shoulders with respect to the ambient. This indicates that while the C9-O4 bond strength is stronger at M it is intermediate at the shoulders L and R with respect to ambient. This is why we mentioned that the internal geometry of the molecule changes from center to shoulder. With respect to the intermolecular interactions, it is observed that the symmetric bands corresponding to aromatic C-H groups 3089 and 3065 cm -1 (in ambient) is slightly red-shifted at the shoulders while maximum for center, M. This indicates that the intermolecular π-stacking interactions vary from the ambient to the shoulders to the center of the bent crystal loop.
The above results and observations corroborate towards differences of internal geometry and intermolecular interactions at the shoulders and the center of the bent crystal. Reviewer #3 comment 13: -I would very strongly recommend that this paper be sent back to the authors to address the points above, and reviewed again after they have revised their discussion based on a re-refinement of their microfocus SCXRD data, using an isotropic thermal motion model for all atoms.

Response:
We thank the reviewer for recommending us to revise based on a re-refinement of our microfocus data using isotropic displacement parameters for all atoms. We have updated Supporting information Section S2, structure solution and refinement where we have included details about the isotropic refinement and the problems as mentioned in response to comment 8.
In the manuscript, we have updated section Atomic scale evidence of bond modulations in bent crystal from μ-SCXRD, μ-IR and μ-Raman spectroscopy and inserted the text in page 9, paragraph 2 "To provide a complete crystal chemical analysis of strained regions, M.o, M.m and M.i we refined the structures with atomic displacement parameters (ADPs) using both isotropic and anisotropic description, for all non-hydrogen atoms, but the refinements for the former yielded poor fit to the diffraction data and chemically non-meaningful distances, hence only the anisotropic descriptions are considered for further structural analysis (See supporting   information, section S2, Tables S3 and S5)." Hence, we feel that we have addressed all the comments satisfactorily and hope this would be now acceptable for all the three reviewers.