Effect of Ring Strain on the Charge Transport of a Robust Norbornadiene–Quadricyclane-Based Molecular Photoswitch

Integrating functional molecules into single-molecule devices is a key step toward the realization of future computing machines based on the smallest possible components. In this context, photoswitching molecules that can make a transition between high and low conductivity in response to light are attractive candidates. Here we present the synthesis and conductance properties of a new type of robust molecular photothermal switch based on the norbornadiene (NB)–quadricyclane (QC) system. The transport through the molecule in the ON state is dominated by a pathway through the π-conjugated system, which is no longer available when the system is switched to the OFF state. Interestingly, in the OFF state we find that the same pathway contributes only 12% to the transport properties. We attribute this observation to the strained tetrahedral geometry of the QC. These results challenge the prevailing assumption that current will simply flow through the shortest through-bond path in a molecule.

Scheme S1: The synthesis route for the synthesis of NB-1 starting from 2,3dibromonorbornadiene a. 4-Iodophenyl thioacetate (1): In a 200 mL round-bottom flask provided with a stirring magnetic bar, 4iodobenzenesulfonylchloride (4 g, 13.2 mmol) was dissolved in acetonitrile (50 mL). Triphenylphosphine (10.4 g, 39.7 mmol, 3 eq.) was added in portions and the mixture was stirred at 60 o C for 20 min. The content was cooled to 50 o C prior adding water (0.4 g, 22.2 mmol, 1.5 eq). After stirring for 5 minutes at this temperature, K 2 CO 3 (2.1 g, 15.2 mmol1.15 eq) was added and the resulting mixture was stirred for five additional minutes. Subsequently, acetic anhydride (4.04 mL, 3 eq) was added and the content was stirred at room temperature S4 for 2 hours. The content was passed through short silica gel plug using DCM as eluent. The solvents were removed in vacuo. The white residue was purified by column chromatography using a gradient of DCM: petroleum ether 1:2 as eluent to collect 1 as a white solid. (yield 89 %). 1  In an oven-dried three necked flask provided with magnetic bar, potassium tert-butoxide (11.2 g, 0.1 mol) was dissolved in THF (200 mL). The solution was cooled to -84 o C using ethyl acetate/Liquid N 2 bath prior the addition of norbornadiene (12.2 mL, 0.12 mol) followed by dropwise addition of n-BuLi (2.5 M in hexanes, 40 mL, 0.1 mol) using dropping funnel. The mixture was stirred at this temperature for 15 minutes followed by 30 min at -41 o C (acetonitrile, Liquid N 2 bath). The solution was cooled back again to -84 o C and ptoluenesulfonyl bromide (11.7 g, 0.05 mol) was added portion wise at which time the yellow solution turned brown. The content was stirred for 15 min at -84 o C and 1h at -41 o C. The solution was cooled again to -84 o C and the remaining p-toluenesulfonyl bromide (11.7 g, 0.05 mol) was added. The mixture was then stirred at this temperature for 15 min and stirred at ambient temperature. The reaction was quenched by adding water (50 mL). The product was extracted with diethyl ether (3*30 mL). The solution was dried over MgSO 4 , filtered and the solvent was evaporated in vacuo to get a brown crude. The crude was re-dissolved in pet.
Ether, washed with DI water and brine, dried over MgSO 4 and filtered. The solvent was S5 removed in vacuo to get a red solution. The crude was purified using kugelrohr distillation to collect the mono norbornadiene fraction up to 80 o C and the 2,3-dibromonorbornadiene product (2) from 90-100 o C. ( In a 100 mL two-necked round bottom flask, 4-bromoiodobenzene (3g, 10.6 mmol), Pd(PPh 3 ) 2 Cl 2 (0.3 g, 4mol%), CuI (43 mg, 4 mol%) were added followed by diisopropyl amine (50 mL). Then triisopropylsilyl acetylene (TIPSA) (2.7 mL, 12 mmol) was added dropwise and stirred at room temperature for 24 h. Then trimethylsilyl acetylene (TMSA) (14 mmol) was added dropwise and stirred at 60 o C overnight. The content was cooled, and filtered through short silica gel column with DCM. The solution was washed with DI water.
The organic phase was collected and the solvent was removed in vacuo. The crude was then submitted to automated chromatography using hexane as eluent at a flowrate of 50 mL/min to collect 3 as a colorless oil, 2.68 g (71.3%).

S6
In a 200 mL round bottom flask provided with magnetic bar, 3 (1.76 g, 5 mmol) was dissolved DCM (60 mL) and methanol (60 mL) and purged with nitrogen. Then potassium carbonate (1.5 g, 2.2 eq) was added and stirred under nitrogen for 2 h at room temperature.
The solid residue was filtered out and the solvent was removed in vacuo. Then the crude was washed with DI water and brine, extracted with DCM, dried over anhydrous MgSO 4 filtered and the solvent was removed in vacuo to get a colourless oil, 4, which was further purified using automated chromatography with hexane as eluent at a flow rate of 50 mL/ min. (  In 100 mL two-necked round bottom flask, Pd(PPh 3 ) 4 (0.27 g, 5 mol%), CuI (45 mg, 5 mol %) were added and evacuated and refilled with nitrogen three times. Then, toluene (15 mL) were added and stirred. 2 (0.59 g, 2.4 mmol) in 5 mL toluene was transferred by syringe followed by 4 which was also dissolved in toluene (5 mL). Triethylamine (3 mL) which was purged under nitrogen was added and stirred at 30 o C overnight. The solution was yellow which changed to reddish brown overnight. The solvents were removed using rotary S7 evaporator. The crude was washed with water and extracted with DCM, dried over anhydrous MgSO 4 , filtered and the solvent was removed in vacuo. The crude which is reddish brown was submitted to automated chromatography with gradient hexane (100%), to DCM (10%) and hexane (90%) with flow rate 50 mL/min to get the desired product (5) Figure S1). 13 C NMR ( Figure S2).

II. Electrochemistry
The energy gap between the HOMO and LUMO levels of NB-1 is estimated from the UV-vis spectrum taking the onset value 461 nm as well as cyclic voltammetry (CV). The experimental value from CV measurement is in agreement with the UV-vis (taking 461 nm as the onset) estimation. This is also in agreement with the DFTB+ values. These results are summarized in Table S1.
The electrochemical study was conducted using a CH-Instruments 650A Electrochemical Workstation with a three electrode setup: platinum wires both as working and counter electrode, and silver/silver ion as reference electrode. A solution of tetrabutylammonium hexafluorophosphate (Bu 4 NPF 6 ) (0.1 M) in acetonitrile was used as supporting electrolyte.
Prior the measurement the electrolyte solution was purged with nitrogren and during the measurement a positive pressure of nitrogen was maintained over the solution. The sample was applied on the working electrode from concentrated chloroform solution of NB-1. The S13 potentials were measured against Ag/Ag + reference electrode and the measurements were calibrated with ferrocene/ferrocenium redox couple.

III. STM-BJ conductance measurement Back switching from NB-1 to QC-2 on gold surface
Having demonstrated that the molecule would undergo a relaxation from QC-2 to NB-1 state accompanying with a conductance change, we further investigate whether it is possible to use UV light to switch the molecule from NB-1 state to QC-2 state on gold surface. To ensure that there is no solution phase molecules involved, we prepared a gold surface with a selfassembled monolayers (SAMs) of NB-1 molecules and carried out the STM-BJ experiment in vacuum (~ 10 -5 Torr). STM-BJ measurement was performed initially and after irradiation with UV-LED (68 mW) onto the surface for 1 hr. As it has been shown in Figure S4

Estimate of the switching ratio
The conductance of a molecular junction at low bias follows an exponential decay as a function of the molecular length, specifically: ‫ܣ‬ • ݁ ିఉ . Here ‫ܣ‬ is a constant, ߚ is a bond-type dependent decay constant taken from the literature and ‫ܮ‬ is the bond length. We can use this relationship to make a back-of-the-envelope estimate of the conductance ratio between NB-1 and QC-2. When this formula is applied, all the terms corresponding to the photo-inactive bonds vanish (and here we also assume that the prefactor vanishes). Furthermore, by assuming the current only flows through the shortest through-bond path the ratio is only dependent on the change between the two molecular structures (C-C double bond becomes C-C single bond), thus the conductance ratio is simply given by:

S16
The β values and bond lengths were taken from Reference 8 . This calculation yields a modest ON/OFF ratio about 2. A posteriori estimate, considering now the longer 3-sigma-bonds path would yield an ON/OFF ratio of 28, considerably much larger than the experimental value of 6.6.
We note here that this estimate is necessarily sensitive to the choice of β values and bond lengths. While there have been numerous measurements of β values from both electron transfer and transport measurements, the values measured depend on the choice of donor and acceptor or binding group and electrode materials. In order to take an unbiased value for our estimation we use values from complex band structure calculations (essentially infinite molecules) to avoid any influence of the binding group or electrode.

Electron transport theory
The electric properties were simulated using the non-equilibrium Green's function (NEGF) method. The NEGF formalism consists of dividing the system in three parts: the device or scattering region and the two electrodes, source and drain. The device region was defined as the molecular junction sandwiched between 4 and 3 layers of gold (111) of surface size 3x6.
Source and drain were identical slabs of gold (111) of size 3x6x6 gold atoms with a Au-Au distance of 2.885 Å (lattice constant 4.08 Å).
We assumed that the electron transport is coherent. The current, I, was calculated at low bias (0.1 eV) with the Landauer formula (eq. 1).
Here, e is the electronic charge, h is the Planck constant, ݂ L(R) is the Fermi function of the left (right) electrode which depends on E the electron energy and µ L(R) the electrochemical potentials of the left (right) electrode, and T is the transmission. The transmission is calculated as a function of energy and is given by eq. 2.

ܶሺ‫ܧ‬ሻ = ܶ‫ݎ‬ሾડ ோ ડ ோ ሿ ሺ2ሻ
Here Tr is the trace, Γ L(R) is the imaginary part of the left (right) electrode self-energy matrix and G R(A) is the retarded (advanced) Green's function of the scattering region (device). Both Γ and G are energy dependent matrices.
The current through a molecular junction can be divided into inter-atomic contributions. We call these local currents and they make possible to attribute electrical properties to functional groups in a molecule 9 .

Gas-phase structures optimizations
The molecular structures used in the theoretical simulations were relaxed with Density Functional Theory (DFT) using the ASE/GPAW 10,11 packages and the PBE exchange

Molecular junction structure optimizations: scattering region
The molecular junctions were generated by manually chemisorbing the relaxed molecular structures between two flat gold (111) electrodes with a unit cell size of 3x6x4 and 3x6x3 (to avoid lattice mismatch over the boundary conditions). Two small gold pyramids were placed between molecule and surface to form the contact. The energy of the system was initially scanned through single point energy calculations to find a value close to the optimal S-Au distance, which was found to be 2.3 Å for both junctions. Finally, the molecule and the gold tips were relaxed while the gold slabs atom positions were kept fixed. The calculations were S18 carried out using the LCAO method, which uses a basis set of atomic-like orbital functions. A polarized double-zeta basis set for C and H, and a polarized-diffuse double-zeta basis-set for Au and S were used with k-points sampling (4,2,1). The structures of the molecular junctions are shown in the main text Figure 4.

Transmission calculation with DFT
The transmission calculations were carried out using the LCAO method with a polarized double-zeta basis set for C and H, polarized-diffuse double-zeta basis-set for Au and S and kpoints (4,2,1). The Hamiltonians and overlap-matrices were calculated from the optimized molecular junctions and the coherent electron transmission was calculated with the ASE transport calculator. The DFT-calculated transmission ratio at the Fermi energy was 13.

Transmission calculation with DFTB+NEGF
The coherent electron transmission was also calculated with the SCC-DFTB method 13 using the DFTB+NEGF 14,15 package and the auorg set of Slater-Koster parameters 16 using the DFT relaxed structure without further optimization. A symmetric bias of 0.1 eV was applied on the system with periodic boundary conditions on all directions and the temperature was set to 0 S19 K. The local currents were calculated with the same system without periodic boundary conditions.
In addition the highest occupied and lowest unoccupied molecular orbital (HOMO-LUMO) energy gap was determined from the isolated molecule relaxations with DFT-PBE and DFTB+ using the mio-1-1 set of Slater-Koster parameters 17,18 . The HOMO-LUMO gap energy obtained was 2.20 eV for NB-1 and 2.66 eV for QC-2.
We used a patched version of the DFTB+NEGF software supplied by Gabriele Penazzi, from the University of Bremen, Germany.

Ring current and energy dependence of the local currents
We analysed the local currents at different energies between the HOMO-LUMO gap in the transmission spectrum in order to understand whether quantum interference effects are in play. Fig. S6 shows that the main features of the inter-atomic transport remain unchanged as the energy moves along the HOMO-LUMO gap.
The only difference visible is seen in the local current of the short path in the QC unit. Its magnitude decreases from energy 0.0 to 0.5 eV, then the direction changes between 0.5 and 1.0 eV, and the magnitude increases as the energy increases. This change in direction suggests that a quantum interference might be present in the sigma system 19 . We therefore further analyzed the LCs magnitudes at 0.0 eV and 1.8 eV (see Fig S7). From this analysis we find the LCs magnitudes (E = 0.0 eV) of two bonds are smaller when compared to the LCs at E = 1.8 eV. This difference is very close to the LC magnitude of the short path which might indicate the presence of a ring current that suppress the overall current at the Fermi energy.