Single-Molecule Junction Origami

Stimuli-responsive molecular junctions, where the conductance can be altered by an external perturbation, are an important class of nanoelectronic devices. These have recently attracted interest as large effects can be introduced through exploitation of quantum phenomena. We show here that significant changes in conductance can be attained as a molecule is repeatedly compressed and relaxed, resulting in molecular folding along a flexible fragment and cycling between an anti and a syn conformation. Power spectral density analysis and DFT transport calculations show that through-space tunnelling between two phenyl fragments is responsible for the conductance increase as the molecule is mechanically folded to the syn conformation. This phenomenon represents a novel class of mechanoresistive molecular devices, where the functional moiety is embedded in the conductive backbone and exploits intramolecular nonbonding interactions, in contrast to most studies where mechanoresistivity arises from changes in the molecule-electrode interface. Abstract Stimuli-responsive molecular junctions, where the conductance can be altered by an external perturbation, are an important class of nanoelectronic devices. These have recently attracted interest as large effects can be introduced through exploitation of quantum phenomena. We show here that significant changes in conductance can be attained as a molecule is repeatedly compressed and relaxed, resulting in molecular folding along a flexible fragment and cycling between an anti and a syn conformation. Power spectral density analysis and DFT transport calculations show that through-space tunnelling between two phenyl fragments is responsible for the conductance increase as the molecule is mechanically folded to the syn conformation. This phenomenon represents a novel class of mechanoresistive molecular devices, where the functional moiety is embedded in the conductive backbone and exploits intramolecular nonbonding interactions, in contrast to most studies where mechanoresistivity arises from changes in the molecule-electrode interface. functional theory calculations demonstrated that mechanosensitivity arises from an anti ⇌ syn conformation switch, as the molecule is folded along the flexible bond following junction compression. Non-bonding interactions are increased in the folded syn state, allowing through-space charge tunnelling that “shortcuts” a poorly-conductive fragment of the molecule. The switching is reversible and robust, and fabricated junctions can be reliably cycled between high and low conductance states. Our results show the promise of deliberately introducing flexible fragments in otherwise rigid molecular wires, and the fine degree of control on molecular conformation (and hence, charge transport properties) that can be attained in single-molecule junctions. Fold me up: A single-molecule junction that can be mechanically folded around a dicarbonyl is presented. In the folded conformation, molecular conductance increases as direct tunnelling between two phenyl fragments of the molecule becomes efficient. The junction can be “unfolded” to its original, thermodynamically stable state simply by pulling the electrodes apart.

conductance can be attained as a molecule is repeatedly compressed and relaxed, resulting in molecular folding along a flexible fragment and cycling between an anti and a syn conformation. Power spectral density analysis and DFT transport calculations show that through-space tunnelling between two phenyl fragments is responsible for the conductance increase as the molecule is mechanically folded to the syn conformation. This phenomenon represents a novel class of mechanoresistive molecular devices, where the functional moiety is embedded in the conductive backbone and exploits intramolecular nonbonding interactions, in contrast to most studies where mechanoresistivity arises from changes in the molecule-electrode interface.

Main Text
Apart from fundamental studies focussing on archetypal saturated compounds such as ,alkanedithiols [1][2][3] and , -alkanediamines, [4][5][6] the majority of molecular wires employed to fabricate molecular junctions are rod-like, π-conjugated structures. They have found such a widespread use because their extensive conjugation results in high conductance, and the conformationally rigid π-system avoids complications which could arise, for instance, from gauche defects in saturated carbon chains. [7,8] Controlled conformational flexibility, however, can be used to impart functionality. Franco et al. proposed a theoretical exploitation of flexible fragments in single-molecule junctions to develop force-sensitive single-molecule devices, [9] based on π-stacking perylene units linked by a saturated propyl chain, that would "unstack" as the junction is stretched. Stacks of phenyl rings and other simple heteroaromatics are able to act as efficient conductors in molecular junctions, [10,11] with charge transported through ππ interactions. [12][13][14][15][16][17] As the stacking configuration is unfolded, conductance is predicted to drop by orders of magnitude, as tunnelling through a saturated propyl chain is very inefficient. Such a device would represent a new class of molecular electronic devices responsive to mechanical stimuli, complementing the existing range that exploits changes in the electrodemolecule interface, [18][19][20][21][22][23][24] stereoelectronic configuration switching, [25] and stretching-dependent quantum interference effects [12,26] as the molecular junction is compressed and relaxed. Intrigued by these phenomena, we designed molecule 1 to have two phenyl rings spaced by a diketone chain. The molecule has a conformationally flexible bond between the two sp 2hybridised carbonyls (orange in Figure 1a). While the ground state structure is a quasi-anti conformation, similar to benzil (an analogue of 1 without the thiomethyl termini), [27] the central C-C bond in such compounds has a dihedral torsional barrier of only 15 -45 kJ/mol (0.15 -0.5 eV) for interconversion between the syn and anti conformations. [28,29] We performed simple molecular mechanics calculations (MM2 Force Field) on 1 to obtain the ground state energy as a function of the (C=O)-(C=O) dihedral angle . We found that the energy profile has two minima, with approximately 130° of difference between them ( Figure 1) and shallow energy barriers. Our results suggest that a syn ( < 90°) ⇌ anti ( > 90°) conversion should be readily attained by compressing a metal-molecule-metal junction made with 1 and mechanically folding the molecule along the central C-C bond. π-π interactions between the two phenyl rings in the syn conformer would then contribute significantly to charge transport (in a way akin to π-stacking), allowing us to detect the interconversion by measuring the junction conductance.
We therefore synthesized 1 and used the scanning tunnelling microscope -break junction technique (STM-BJ) [1]  We performed the process in a 1 mM mesitylene solution of 1, and the results can be observed in Figure 2. The conductance histogram ( Figure 2b) shows two main contributions: a highconductance (~10 -3 G0) feature, at small junction size (~0.8 nm, accounting for the electrodes snapback [18] ), and a low conductance feature (~10 -4 G0) at larger junction size (~1.2 nm including snapback). The results therefore suggest the presence of two possible conformers of 1 in the junction, with conductance difference of more than one order of magnitude and a difference in size of 0.4 nm. We then performed piezo-modulation experiments on the fabricated junctions. In these experiments, nanogaps able to accommodate the extended anti conformation (1.2 nm) are initially fabricated, and their size is then modulated by applying a square wave signal to the piezo voltage. The molecule is therefore allowed to assemble in the gap in its thermodynamically favoured anti state, and it is then mechanically folded into the syn conformation by compressing the junction. More information on the piezo-modulation experiments can be found in our previous publications on the subject [19,20] and details are provided in the SI. Under a modulation amplitude of 0.4 nm the junctions could be cycled reliably from the high-to the low-conformation, with excellent repeatability (Figure 2d). We can rule out any contributions from variations in the molecule-electrode interface as a mechanism for the observed switching phenomena, as we have already demonstrated that thioanisole contacts do not change binding configuration upon junction compression. [20] Similarly, we can discount an interpretation of our results based on the formation of shorter junctions through Au-carbonyl contacts, as no interactions between a (di)ketone and gold electrodes (e.g. in measurements of molecular wires containing fluorenones, [30] anthraquinones [31] or diketopyrrolopyrroles [32] ) have been reported. To verify that a syn ⇌ anti conformation change is responsible for the observed mechanoresistive effects, we performed power spectral density (PSD) analysis [33] on the junctions in their relaxed and compressed state. PSD has been used in the literature to characterize through-space coupling in molecular junctions where charge transport does not follow the chemical bond organization, but travels through eigenchannels opened by non-bonding interactions. [13,14,34,35] When charge transport is purely through-bond, the (the integral of the power spectral density) scales approximately with , while the scaling increases when charge transport has a through-space character, to approach the value found for pure through-space tunnelling of . When is normalized by the average conductance ( / ) it is therefore generally found to be insensitive to the junction conductance for through-bond coupling. On the other hand, a strong correlation between / and is observed when through-space coupling significantly contributes to charge transport. [33] In order to estimate  Junctions in their relaxed, low-state showed insensitive to conductance, thus confirming a dominant through-bond mechanism of charge transport (Figure 3b). In the compressed, high-state, however, strongly correlates with the average conductance of the junction (Figure 3c), indicating that through-space tunnelling phenomena are now contributing to the overall charge transport. This further reinforces our proposed interpretation of a mechanoresistive behaviour arising from conformational change, where a non-bonding transport channel is opened upon folding the junction to the syn conformation.
As a further control experiment, we studied the (E)-stilbene analogue of 1 ((E)-1,2-bis(4-(methylthio)phenyl)ethene, see SI), which the central C=C bond renders conformationally locked. We found no evidence of mechanically-controlled conductance switching in the STM-BJ and piezo-modulation experiments, showing that (i) a conformationally flexible molecule is needed to attain the switching behaviour and (ii) direct electrode-electrode tunnelling does not contribute significantly to the overall charge transport with the piezo ramps used in this study.
Furthermore, PSD analysis suggests a purely through-bond charge transport mechanism ( / insensitive to ). Details on these control experiments can be found in the SI. Our results may also provide a novel explanation for high-conductance features at short junction extensions, which have been seen in similar foldable molecules incorporating two phenyl rings such as MeS-C6H4-Q-Q-Q-C6H4-SMe (Q = CH2, SiMe2, GeMe2). [36,37] It is worth noting that the prominence of the high conductance feature is significantly greater in our data compared to these systems, where it only appears as a weak peak in the histograms.
We attribute this increased prominence to the sp 2 hybridization of the diketone bridge of 1 (in contrast with the -Q-Q-Q-sp 3 hybridization), that results in a preferential rotation around the (O=C)-(C=O) axis and a more efficient folding of the junction, which is also energetically favoured by the relatively long C-C bond (1.54 Å in the solid-state crystal structure of 1).
We then performed density functional theory (DFT) quantum transport calculations on the junction in the relaxed and extended states, by imposing constraints on the distance between the two metallic electrodes and letting the molecular portion of the junction relax to its energy minimum. Our calculations show that the molecule is indeed likely to adopt a syn conformation upon junction compression (Figure 4a). DFT modelling predicts a total energy difference between the anti and the syn conformation of 0.42 eV. This is less than the binding energy between the molecule and the two Au electrodes (0.54 eV) and we therefore expect the molecule to fold around the conformationally flexible C-C bond as the junction is compressed.
The Gollum code [38] was then used to calculate the transmission coefficient ( ) for electrons of energy passing from the source to the drain electrode of the junction, through molecule 1. The electrical conductance then can be obtained using the Landauer formula, and we analysed the whole behaviour of ( ) within the HOMO-LUMO gap (see methods in the SI).
Transport calculations on the two states ( Figure 4b) show that the value of ( ) is indeed strongly dependent on electrode compression. Our calculations show that for the whole energy range between the HOMO and LUMO, T(E) is lower for the anti conformation compared to the  Analysis of the LUMO isosurface of the two conformers (Figure 4c and 4d) shows that two lobes of opposite sign are brought within short distance as the junction is compressed, as the bonds on which they are located are only separated by ~0.3 nm. The π-π interaction between these two lobes opens a new transport channel that allows efficient and short-range throughspace tunnelling responsible for the observed boost in molecular conductance. To further verify the proposed mechanism, we re-calculated ( ) for the syn conformation setting the relevant through-space coupling parameters to zero in its mean-field DFT Hamiltonian (see SI, Figure S7). We found that ( ) for the compressed junction drops by one order of magnitude when only through-bond transmission is allowed, approaching the values found for the relaxed junction. This is further confirmed by a tight-binding model (Figure 4e). Here, a scattering region is weakly coupled to two one-dimensional leads and ( ) increases only as the parameters for through-space coupling are set to values > 0.
In conclusion, we present here a novel way to impart mechanosensitivity to molecular junctions, by exploiting a flexible diketone moiety. We found a large increase in conductance upon compression of the junction, and power spectral density analysis combined with density functional theory calculations demonstrated that mechanosensitivity arises from an anti ⇌ syn conformation switch, as the molecule is folded along the flexible bond following junction compression. Non-bonding interactions are increased in the folded syn state, allowing through-space charge tunnelling that "shortcuts" a poorly-conductive fragment of the molecule. The switching is reversible and robust, and fabricated junctions can be reliably cycled between high and low conductance states. Our results show the promise of deliberately introducing flexible fragments in otherwise rigid molecular wires, and the fine degree of control on molecular conformation (and hence, charge transport properties) that can be attained in single-molecule junctions.
STM-BJ Measurements. Junctions were fabricated and characterised using the STM-BJ technique, [1] using a modified Keysight 5500 STM. Further details on the equipment used, the data acquisition process and its analysis are available in our previous publications on the subject [19,20] and in the SI.
DFT and Transport Calculations: The optimised geometry, with ground-state Hamiltonian and overlap matrix elements were obtained using DFT and the SIESTA [41] code. These results were then combined with the Gollum [38] implementation of the non-equilibrium Green's function method [42] to calculate the phase-coherent, elastic-scattering properties of the system, consisting of two gold electrodes and the molecule as scattering region. Further details are available in the SI.

Associated Content
The supporting information is available on the publisher website.
 Synthetic procedures and characterization

Synthetic Details
All reactions were carried out under an inert N2 atmosphere, THF and toluene were dried using an Innovative technology PS-400-6-MD solvent purification system. All other reagents and solvents were purchased from Sigma Aldrich, Acros Organics or Fischer and used as received. Column chromatography was performed on silica gel, monitored by thin layer chromatography (TLC) using Merck silica gel 60 F254 plates and visualised under UV light. NMR spectra were recorded on a Bruker Avance III HD and chemical shifts referenced to residual solvent or tetramethylsilane. Mass spectra were recorded by the University of Liverpool Analytical Services on an Agilent QTOF7200 or Agilent QTOF6540 mass spectrometer. Elemental analysis was carried out by The University of Liverpool Analytical Services on an Elementar Vario Micro Cube. Fourier transform infrared (FT-IR) spectra were recorded with a PerkinElmer Spectrum 100 FT-IR spectrometer.

Linear Piezo Ramp Experiments
All traces collected with a linear piezo ramp (20 / in this study) in the main text and later in this document were used to compile histograms and 2D plots with no further processing or selection.

Piezo Modulation Experiments
When a non-linear ramp was applied to the piezo to verify the mechanoresistive behavior, a modified protocol was applied. The raw traces, after conversion of current to conductance by using Ohm's law ( = / ) were sliced between abrupt stretches by analysing the second derivative of the signal applied to the piezo transducer and cutting when its value is above a threshold.
The slices were then fed into a sorting algorithm, that takes the average of the first and last modulation and checks that both are below 0.1 and above the noise level of the preamplifier (10 -5.5 at the bias voltage used in this study). This process filters out slices where the tip and the substrate are in contact, those where there is no molecular bridge present and those where the molecular junction did not survive the whole modulation process, leaving only the slices relative to stable molecular junctions. All the slices selected by the sorting algorithm were then used to compile 2D density maps (conductance vs time). More details on the process are available in our previous publication on the subject. [4] 2

.3 PSD Analysis
In order to perform PSD analysis on the junction in both its compressed and relaxed state, we performed a measurement with a single modulation cycle, while monitoring current at 100 kSa/s. The FFT algorithm outputs the across the whole frequency spectrum. Numerical integration between 100 Hz and 1 kHz yielded the in this region, that was plotted vs (average of the conductance of the slice of interest) to give the plots shown in the main paper. The 1 kHz and 100 Hz cut-offs were chosen as they have been demonstrated in the literature to be effective in limiting the contributions to arising from thermal noise (> 1 kHz) and mechanical vibrations (< 100 Hz). [5] 3

. Control Experiments
As discussed in the main text, we used (E)-1,2-bis(4-(methylthio)phenyl)ethene to demonstrate that the switching phenomena do not occur in conformationally-locked compounds.
The synthesis of the compound is described earlier in this document. We started by measuring conductance using a linear ramp, in a regular STM-BJ experiment. No evidence of switching was found, as the results only show a single, sharp and well-resolved peak in the conductance histogram.  We observed a similar behaviour in other, thioanisole-terminated compounds [4] and we attributed this to an increase in "lateral coupling" -weak interactions between the aromatic -system and the metallic electrodes already observed in other molecular wires [6] and clearly not a change in the molecule-electrode binding mode. Finally, we performed PSD analysis on the traces, to verify the transport pathway in 1,2-bis(4-(methylthio)phenyl)ethene. No correlation between / and was found, thus confirming the molecule has a pure through-bond charge transport mechanism. Figure S6: / vs heatmap for 1,2-bis(4-(methylthio)phenyl)ethene. The dashed lines are the contours at 25%, 50% and 75% height of the 2D gaussian fit of the data. 8775 traces.

Details on DFT and Transport Calculations
The optimized geometry, ground state Hamiltonian and overlap matrix element of each structure was selfconsistently obtained using the SIESTA [7] implementation of density functional theory (DFT). SIESTA employs norm-conserving pseudo-potentials to account for the core electrons, and linear combinations of atomic orbitals to construct the valence states. The generalized gradient approximation (GGA) of the exchange and correlation functional is used with the Perdew-Burke-Ernzerhof parameterization (PBE) a double-ζ polarized (DZP) basis set, a real-space grid defined with an equivalent energy cut-off of 250 Ry.
The geometry optimization for each structure is performed to achieve forces smaller than 10 meV / Å.
The mean-field Hamiltonian obtained from the converged DFT calculation was combined with Gollum [8] implementation of the non-equilibrium Green's function method [9] to calculate the phase-coherent, elastic

Transport Calculations in the absence of π-π interactions
As discussed in the main paper, we performed DFT calculations setting through space coupling parameters to zero, to evaluate the effect of π-π interactions on junction conductance. Results can be found in Figure   S7.  Figure S8) and net charge transfer (inset of Figure S8) as a function of the distance between the thiomethyl terminus of 1 and the electrode. As can be seen, up to 0.3 electrons are transferred at the optimal S-Au distance of 2.6 Å, and the LUMO is displaced by up to 20 meV towards the electrodes . Figure S8: LUMO displacement and net electron transfer (inset) as a function of the S-Au distance d for molecule 1.