Single-Molecule Junctions with Epitaxial. Graphene Nanoelectrodes. (Supporting Information)

Transcription

1 Single-Molecule Junctions with Epitaxial Graphene Nanoelectrodes (Supporting Information) Konrad Ullmann, Pedro B. Coto, Susanne Leitherer, Agustín Molina-Ontoria,, Nazario Martín, Michael Thoss, and Heiko B. Weber*, Lehrstuhl für Angewandte Physik und Interdisziplinäres Zentrum für Molekulare Materialien, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Staudtstr. 7/A3, D Erlangen, Germany Institut für Theoretische Physik und Interdisziplinäres Zentrum für Molekulare Materialien, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Staudtstr. 7/B2, D Erlangen, Germany Departamento de Química Orgánica, Facultad de Química, Universidad Complutense, E Madrid IMDEA-Nanoscience, Campus de Cantoblanco, E Madrid, Spain

2 Detailed protocol of graphene nanojunction formation In order to form gaps in the desired nanometer range in epitaxial graphene, an electroburning process similar to the one in ref 1 is used. We start with a graphene stripe which is typically 6 µm x 3 µm and has a constriction of ~1 µm in the center. Metallic (Ti/Au) pads are attached to the graphene stripe by standard e-beam lithography to contact each microstripe individually. To form a gap in the constricted part of the stripe, we apply a voltage ramp of ~0.5 V/s in ambient air. With an initial resistance of ~3 kω, we have to increase the voltage up to ~30 V to reach the critical current density of ~5 ma/µm at which the oxidation process starts (see Fig. 1d-e of the manuscript). To avoid uncontrolled oxidation of the graphene a feedback loop is used (20 ms sampling time), which is triggered by a drop in the conductance (5 %) within the last 200 mv. As soon as such a drop is detected, the voltage is immediately reset to zero and a new ramp is started, now with an increased device resistance. Fig. 1d-e show the individual IVs, obtained during a burning process. The increase of resistance between each voltage ramp varies from gap to gap, as the oxidation process is not perfectly controllable. This protocol is repeated until the gap shows a resistance in the 10 GΩ range. The yield of junctions with the desired 10 GΩ resistance is %. For the remaining junctions no tunneling current can be observed, probably because of uncontrolled oxidation. Total number of devices In total, we fabricated 244 graphene nanoelectrodes with the desired GΩ resistance using the protocol mentioned above. From these, only 5 % showed a nonlinear increase of conductance after application of molecules and 2 % showed step-like features in the IVs at low temperatures. These 2 % equals 5 junctions which are all shown in the manuscript and the SI.

3 Time and temperature stability of graphene tunnel junctions To investigate time-dependent stability of bare graphene tunnel junctions, we measured 19 tunneling IV which were stored for 8 days at ambient conditions. After these 8 days, 11 of the 19 junctions showed no change in the tunneling behavior (representative example in Fig. S1a). The remaining 8 of 19 junctions showed a decreased conductance (one order of magnitude maximum, representative example in Fig. S1b). Figure S1: Representative examples of bare graphene tunneling junctions after 8 days in ambient conditions. (a) 11 of 19 junctions remained unchanged. (b) 8 of 19 junctions showed a decreased conductance after 8 days. In order to investigate the temperature dependence of the bare graphene tunneling junctions, we compared IVs taken at room temperature to measurements carried out at 15 K. All of the 20 junctions under investigation showed a slightly reduced tunneling current (by a factor 2 to 5) after cooling down the sample, but remained stable in shape. Fig. S2 depicts a representative example. With this finding we can exclude current pathways through the SiC substrate which would freeze out at low temperature. The residual temperature dependence may stem from adsorbates or changes in the dielectric environment near the gap.

4 Figure S2: Representative examples of a bare graphene tunneling junction after cooling down to 15 K. The current is slightly reduced, but the IVs remain stable in shape. Influence of the gaseous environment on the gap formation The electroburning process has been carried out under various conditions in order to investigate the influence of the gaseous environment during the procedure. We used vacuum (10-7 mbar and 1 mbar), helium, carbon dioxide and ambient air (humid) as testing conditions. When no oxygen is available while driving a current through the microstripe, the graphene can withstand much higher current densities without the typical drop in conductance. Devices tested under high vacuum (10-7 mbar) and under helium atmosphere are able to sustain current densities up to 8 ma/µm, without breaking. Subsequently recorded SEM images show changes in the graphene layer, which however do not decrease the conductance of the device (see Fig. S3a, d). These devices reach temperatures high enough to emit white light, visible to the naked eye. When the applied voltage is further increased, a collapse of the structure is observed, with increased conductance afterwards, see Fig. S3b (further cf. ref. 2 ).

5 In a next step, we vented the chamber to 1 mbar with ambient air. Under these conditions drops in the conductance occur when ramping up the voltage, but SEM imaging shows a not well defined crack with wrinkled edges (Fig. S3c). It is also possible to form gaps that show tunnel characteristics using a carbon dioxide environment. However, these junctions show black residuals near the gap region, which is presumably carbon rich (see Fig. S3e). We also observed an influence of the humidity of ambient air. It turned out that a high content of water (depending on weather conditions) leads to a pale edge modification visible in SEM (see arrow, Fig. S3f). We assign this finding to an oxygen intercalation of the graphene sheet in the gap region, due to elevated temperatures during the burning process. 3

6 Figure S3: SEM micrographs of graphene junctions formed under different atmospheric conditions: (a) under high vacuum conditions, the graphene can sustain high current densities without breaking. (b) When the applied voltage is further increased, a collapse of the structure is observed. (c) At 1 mbar the formation of cracks can be seen together with damages in the graphene sheet. (d) Using a helium environment yields results similar to the high vacuum experiments, presumably because of the lack of oxygen. (e) Carbon dioxide produces some black residuals near the gap, probably excessive carbon. (f) With a high content of water in the ambient air, we observe a pale modification of the edge (see arrow), which we assign to oxygen intercalation.

7 Discrete multistabilities in molecular junctions After depositing molecules to a specific graphene nanojunction GNJg (not shown in the main manuscript) and cooling down the sample to 15 K, we see sudden and stochastic transitions between three branches in the IV characteristics, which are expected from small geometric multistabilities of a single-molecule junction, but are unlikely for larger ensembles. Figure S4: Stochastic switching between three branches of the IV, which would be unlikely for a large ensemble of molecules in the junction. Work functions of epitaxial graphene and gold The work functions of epitaxial graphene and gold are significantly different (graphene on SiC (0001): = ev, gold: = -5.0 to -5.3 ev), as is the character of the moleculeelectrode binding. It should be considered, however, that the formation of the graphene electrode pair at elevated temperature (a few hundred degrees Celsius are conceivable) in the presence of air modifies the outmost electrode region: First, we expect local intercalation of oxygen 3 close to the edge (cd. Fig. 2c in the manuscript), which would lower the local work function in the junction region by ~0.5 ev. The local blurring observed in SEM imaging

8 supports this assumption. The expected modification in the junction area is illustrated in Fig. 2c of the manuscript where the outmost structure (on blue background) on the left and the right represent the unperturbed monolayer graphene areas, which are negatively charged. Closer to the junction, we expect oxygen intercalated bilayer graphene (positively charged, red background), to which the fullerene endcaps of the molecules attach. In the middle, the graphene is removed and the gap is bridged by the molecule. We have no direct information, whether additional SiO 2 grows there, which would consume Si from the substrate. A significant protrusion of SiO 2 in the gap would, however, cause a distortion of the molecular bridge, which can be excluded because of the close similarity of the IV curves for gold and graphene electrodes. Additionally, open graphene edges readily are substituted when exposed to air, for example with carboxylic groups, 4 which further withdraws electrons form the graphene edge. A simple estimate of the length scales involved (fullerene-fullerene distance 2.4 nm) and the electrode spacing (~1-2 nm) yields that the fullerene is placed only few A ngström away from the edge, which is certainly a distance at which electronic depletion resulting from polar edge functionalization is not equilibrated. Altogether, a considerably lowered local work function very similar to that of gold is not surprising. Correlation between electric field and bias voltage In figure 5c,d of the manuscript, the dependence of the HOMO and the HOMO-1 on the external electric field along the molecule is shown. Hence, the polarizing effect of the applied bias can be included, which is important for the creation of asymmetries in the transport properties. The built in asymmetric response to external field qualitatively explains the observed asymmetry in our IVs, as the levels are significantly more separated in one bias direction. The conversion between bias voltage into an electric field used the fullerenes center

9 of mass distance of 2.4 nm (electric field = bias voltage / 2.4 nm), as a rough estimate. Figure S5 shows the dependence of the HOMO and the HOMO-1 on the applied bias voltage for conformer A and B. Figure S5: Dependence of the HOMO and the HOMO-1 on the applied bias voltage for conformer A and B. 1. Prins, F.; Barreiro, A.; Ruitenberg, J. W.; Seldenthuis, J. S.; Aliaga-Alcalde, N.; Vandersypen, L. M. K.; van der Zant, H. S. J. Nano Letters 2011, 11 (11), Hertel, S.; Kisslinger, F.; Jobst, J.; Waldmann, D.; Krieger, M.; Weber, H. B. Applied Physics Letters 2011, 98 (21), Oliveira, M. H.; Schumann, T.; Fromm, F.; Koch, R.; Ostler, M.; Ramsteiner, M.; Seyller, T.; Lopes, J. M. J.; Riechert, H. Carbon 2013, 52 (0), Hirsch, A. Angewandte Chemie International Edition 2002, 41 (11),