SUPPLEMENTARY INFORMATION

Transcription

1 SUPPLEMENTARY INFORMATION DOI: /NNANO Graphene nanoribbon heterojunctions Jinming Cai, Carlo A. Pignedoli, Leopold Talirz, Pascal Ruffieux, Hajo Söde, Liangbo Liang, Vincent Meunier, Reinhard Berger, Rongjin Li, Xinliang Feng, Klaus Müllen, and Roman Fasel Supplementary Information S1. Synthesis of the precursor monomers S2. Single crystal X-ray structure analysis S3. Structure of the N-GNR precursor polymer S4. STM di/dv mapping of p-gnrs and N-GNRs S5. STM di/dv mapping of p-n-gnr heterostructures in the valence band region S6. Summary of valence and conduction band positions derived from di/dv maps S7. DOS and electrical conductance of heterostructures S8. DFT simulation of di/dv plots of a p-n-gnr heterojunction S9. Concept of GNR doping via monomer chemical substitution S10. GNR fabrication on Au/mica substrates and transfer to arbitrary substrates S11. Supplementary References NATURE NANOTECHNOLOGY 1

2 S1. Synthesis of the precursor monomers Supplementary Figure 1 Synthesis of the precursor monomers. a, One-step synthesis of 1,2-di(pyrimidin-5-yl)ethyne 5. b, Synthetic route to 5,5'-(6,11-dibromo-1,4- diphenyltriphenylene-2,3-diyl)dipyrimidine 2. c, Chemical structure of the pristine monomer 6,11-dibromo-1,2,3,4-tetraphenyltriphenylene 1. The compounds 5-bromopyrimidine (3) and bis(tributylstannyl)acetylene (4) are commercially available from Sigma-Aldrich. All other starting materials were purchased from Fluka, Alfa Aesar and Strem Chemicals and were used as received without further purification. The synthesis of 5,10-dibromo-1,3-diphenyl-2Hcyclopenta[l]phenanthren-2-one 6, and 6,11-dibromo-1,2,3,4-tetraphenyltriphenylene 1 is described elsewhere R1. Synthesis of 1,2-di(pyrimidin-5-yl)ethyne (5). Figure S1a shows the one-step synthesis of compound 5. In a dry and inert schlenk flask, a mixture of 3 equivalents of 5-bromopyridine 3 (11.05 g, mmol), 5-mol% of Pd(PPh 3 ) 2 Cl 2 (0.81 g, 1.15 mmol), 10-mol% of Cu(I) (0.44 g, 2.31 mmol), 10-mol-% of PPh 3 (0.60 g, 2.31 mmol) and 2 equivalents of CsF (7.04 g, mmol) were dissolved in DMF (140 ml, anhydrous, argon bubbled). One equivalent of Bis(tributylstannyl)acetylene 4 (14.00 g, 12 ml (d: 1.147), mmol) was added dropwise via a syringe and warmed to 45 C in a preheated oil bath. The reaction was stirred for two days, quenched with 80 ml water and 200 ml dichloromethane (DCM). The mixture was Supplementary Information - 2 -

3 filtered through Celite (800 ml, DCM:EtOAc (ethyl acetate) = 1:1). The organic layer was separated and dried over MgSO 4. Evaporation of solvents and chromatography on silica gel with hexane: EtOAc (1:3) afforded compound 2 (4.0 g, 95 %) as colorless crystalline material. Synthesis 5,5'-(6,11-dibromo-1,4-diphenyltriphenylene-2,3-diyl)dipyrimidine (2). Figure S1b shows the synthetic route of compound 2. A solution of 5,10-dibromo-1,3- diphenyl-2h-cyclopenta[l]phenanthren-2-one 6, recrystallized out of pyridine, (1 g, 1,9 mmol) and 1 equivalent of 1,2-di(pyrimidin-5-yl)ethyne 5 (0.37 g, 1.9 mmol) in diphenylether (8 ml) was placed in a microwave tube and purged with argon before sealing. The reaction was carried out in a CEM Discover microwave oven at 300 W in safe temperature mode, allowing to reach 300 C two times. Chromatography on silica gel with hexane: DCM (1:1) first and pure EtOAc afterwards gave compound 2 (0.33 g, 25 %). S2. Single crystal X-ray structure analysis BD001(RB336) Identification code CCDC Formula weight (g/mol) Temperature (K) 193 Wavelength (Å) Space group Monoclinic, P2 1 /n Cell dimensions a, b, c (Å) (10), (6), (2) ( ) 90, (6), 90 Volume (Å 3 ) (4) Z, calculated density (g/cm 3 ) 4, Absorption coefficient (mm -1 ) 2.70 F(000) 1392 Crystal size (mm 3 ) 0.15x0.35x0.60 Theta range for data collection ( ) 2.8 to 28.4 Limiting indices -16 h 17, -15 k 15, -26 l 26 Reflections collected/unique 22858/7324 [R int = ] Completeness to theta = % Absorption correction Integration Max. and min. transmission and Final R indices R1 = , wr2 = R indices (all data) R1 = Supplementary Table 1 Data collection and refinement statistics of single crystal X-ray structure analysis of precursor monomer 2 Supplementary Information - 3 -

4 CCDC contains the detailed crystallographic data. These data can be obtained free of charge via or by ing or by contacting The Cambridge Crystallographic Data Centre, 12, Union Road, Cambridge CB2 1EZ, UK; fax: Crystal of precursor monomer 2 was grown by slow evaporation of a concentrated solution in methylene chloride layered with ethanol. Crystal diffraction data were recorded using a Stoe IPDS 2T diffractometer with Mo-Kα radiation. The crystal structure was solved by direct methods (SIR-97). Refinement was done by leastsquares calculations on F with anisotropic temperature factors for all non-hydrogen atoms. The H-atoms were included in the refinement with isotropic temperature factors in the riding mode. S3. Structure of the N-GNR precursor polymer Supplementary Figure 2 STM images of N-GNRs before cyclodehydrogenation. a, precursor polymer of the chevron-type N-GNRs after surface-assisted C-C coupling of monomer 2 (T=35 K, U=1.0 V, I=0.05 na). b, Small-scale STM image of the polymer chains (T=35 K, U=-1.0 V, I=0.07 na) with partly overlaid chemical model. Supplementary Fig. 2a shows a STM image (recorded at 35 K) of N-GNRs on Au (111) before the cyclodehydrogenation step. The polymer chains preferably grow Supplementary Information - 4 -

5 along the step edges of the surface. On the flat terraces, however, they agglomerate into compact islands. Supplementary Fig. 2b presents a high resolution STM image of the polymer chains on Au (111) (recorded at 5 K), with a structural model of the polymer overlaid. As can be discerned from the model, nitrogen-hydrogen interactions between neighboring polymer chains stabilize the axially shifted, parallel packing into islands. S4. STM di/dv mapping of p-gnrs and N-GNRs Supplementary Figure 3 di/dv maps of p-gnrs and N-GNRs at various bias voltages and corresponding STM images. a & g, STM images of p-gnrs (left, gray) and N-GNRs (right, blue) scanned at 1.65V and -1.65V (T=5K, I=0.15nA). b-f & h-l, Corresponding di/dv maps at different bias voltages (T=5K, =860 Hz, U mod =20mV). The white dashed line indicates the boundary between p-gnrs and N-GNRs. According to the di/dv maps, no electronic states of p-gnrs appear between 1.45 V and V, whereas N-GNRs exhibit none between 0.35V and -1.65V. The band gaps of both GNRs are thus equal (~2.0eV), but both the valence band and the conduction band of the N-GNR are shifted down in energy by ~1.1eV. Both p-gnrs and N-GNRs were prepared on the same Au (111) substrate one after the other. Supplementary Fig. 3a & g show STM images (T=5K, U=1.65V, -1.65V, I=0.15nA) of an area covered with p-gnrs on the left side (highlighted in gray) and N-GNRs on the right side (highlighted in blue). The GNRs can be distinguished by their specific packing on Au (111). STM di/dv maps were performed at different bias voltages from 1.65 V to -1.65V. Upon reducing the bias voltage, the electronic states of p-gnrs disappear at 1.45V and appear again at -0.55V, indicating a band gap of ~2.0 ev for p-gnrs on Au(111). Meanwhile, the states of N-GNRs persist down to significantly lower bias voltages, disappear at 0.35V and reappear again at -1.65V. The band gap is thus the same as the one of the p-gnrs (~2.0 ev), but the band edges Supplementary Information - 5 -

6 are 1.1 ev lower in energy for both the conduction and the valence band. S5. STM di/dv mapping of p-n-gnr heterostructures in the valence band region Supplementary Figure 4 di/dv maps of p-n-gnr heterostructures around the valence band edges, and corresponding STM images. Heterostructures in all images are outlined by white dashed lines. a, STM image (T=5K,U=-0.7V, I=0.1nA) of p-n-gnr heterostructures. Pristine monomers 1 and N-substituted monomers 2 are indicated by gray and blue dots, respectively, in (b) according to the corresponding di/dv maps (c-f, T=5K, =860 Hz, U mod =20mV) which reveal a ~0.5 ev valence band offset across the heterojunctions. The red symbols correspond to the symbols in Fig. 3 and relate the di/dv maps to the energy band lineup. Heterostructures were synthesized by alternating deposition of the pristine monomers 1 and the nitrogen substituted monomers 2 onto a Au (111) held at 200 C, followed by annealing to 420 C to induce cyclodehydrogenation. Due to rapidly alternating precursor deposition, heterostructures with many short segments containing few monomers can easily be identified. STM di/dv maps were recorded at energies in the conduction band (Fig. 4) and valence band regions (Supplementary Fig. 4). From Fig. 4, we derived a conduction band offset of ~ 0.5 ev across the heterojunctions. Supplementary Fig. 4 shows di/dv maps of p-n-gnr heterostructures in the valence band region, and the corresponding STM images. Upon going to more negative bias voltages, electronic states first appear on p-gnr segments (-0.85 ev) and then also on Supplementary Information - 6 -

7 N-GNR segments (-1.3 ev). We derive a valence band shift of about 0.5 ev across the heterojunctions, which is in good agreement with the calculations shown in Fig. 3. S6. Summary of valence and conduction band positions derived from di/dv maps Supplementary Figure 5 Band lineup determined from the experimental results shown in Supplementary Fig. 3 (p-gnrs, left part; N-GNRs, right part), Fig. 4 and Supplementary Fig. 4 (heterostructures, middle part). Band gaps for individual p-gnrs and N-GNRs are 2.0 ev, but both the conduction and the valence band of the N-GNR are shifted down in energy by 1.1 ev. Band gaps for p-gnr and N-GNR segments in p-n-gnr heterostructures are 2.0 ev as well, but with a band offset of 0.5 ev across the heterojunction. Supplementary Figure 5 gives a summary of the VBM and CBM positions determined from the di/dv mapping results shown in Supplementary Fig. 3, Fig. 4 and Supplementary Information - 7 -

8 Supplementary Fig. 4. The vertical axis gives the bias values of the di/dv maps. The letters in the left and right red rectangle refer to the corresponding panels in Supplementary Fig. 3, those in the central red rectangle to Fig. 4 (upper half, conduction band region) and to Supplementary Fig. 4 (lower half, valence band region). They indicate the bias voltages at which the corresponding di/dv maps were taken. S7. DOS and electrical conductance of heterostructures Supplementary Figure 6 DOS and electrical conductance of heterostructures. a, Transport device built from the heterostructure of p-gnr and N-GNR. The left (right) electrode is the semi-infinite p- (N-)GNR in the left (right) direction. b, Densities of states of the heterostructure (red solid line), p-gnr (gray dashed line) and N-GNR (blue dashed line). c, Electrical conductance of the heterostructure (red solid line), p-gnr (gray dashed line) and N-GNR (blue dashed line). Fermi level of the heterostructure is set at 0 ev. d, Transport device with the N-GNR as the central scattering region and p-gnrs as both electrodes. e, Densities of states of the heterostructure (red solid line) and p-gnr (gray dashed line). f, Electrical conductance of the heterostructure (red solid line) and p-gnr (gray dashed line). The Fermi level of the heterostructure is again set at 0 ev. Supplementary Information - 8 -

9 For a better understanding of the nitrogen doping effects on GNRs, localized-orbital density functional theory (DFT) calculations have been performed with SIESTA R2 on the heterostructures shown in Supplementary Fig. 6a & d, within the generalized gradient approximation (GGA) using the Perdew-Burke-Ernzerhof (PBE) exchangecorrelation functional R3. A real-space mesh equivalent to a plane-wave cutoff energy of 250 Ry and the Γ-point sampling in the Brillouin zone R4 were used. The size of the unit cell along the periodic direction was sufficiently large (in excess of 6 nm) for the Γ-point sampling to yield converged results. The basis set was composed of doublezeta with single polarization (DZP) orbitals, with an energy shift of 50 mev R5. All atoms were relaxed until the residue forces were all below 0.04 ev/å. After reaching the equilibrium, the converged Hamiltonians and overlap matrices were cast in a format suitable for electronic transport and density of states calculations within the non-equilibrium Green's function (NEGF) formalism framework R6,R7. As shown by the blue dashed line of Supplementary Fig. 6b, the density of states (DOS) of the N-GNR (Supplementary Fig. 6a) indicates that it has an electronic band gap of 1.50 ev in the energy range (-0.96, 0.54) ev while the DOS of the p-gnr shows an electronic band gap of 1.56 ev in the energy range (-0.51, 1.05) ev (gray dashed line Supplementary Fig. 6b). They have similar electronic band gaps, but the gap energy range of the N-GNR is shifted by approximately 0.48 ev to lower energies compared to the gap range of the undoped one, due to the presence of nitrogen. Consequently, the heterostructure made of p-gnr and N-GNR segments presents a band offset, leading to a reduced electronic band gap of 1.05 ev (red solid line in Supplementary Fig. 6b). To further study the band alignment effects and determine the transport gap, electronic transport calculations have been carried out, as shown in Supplementary Fig. 6c. For an electron with a given energy to transfer through the heterostructure (i.e. non-zero electrical conductance), both p- and N-GNRs are required to have non-zero DOS at such energy. Hence, the band alignment induced by nitrogen doping increases the transport gap of the heterostructure to 2.02 ev (red solid line in Supplementary Fig. 6c). Compared to the electrical conductance of both p- and N-GNRs (gray and blue dashed lines in Supplementary Fig. 6c respectively), the heterostructure has an overall decreased electrical conductance (due to nitrogen scattering) and larger transport gap. We have also studied the other heterostructure, the N-GNR segment sandwiched by p-gnr segments (Supplementary Fig. 6d), and similar conclusions can be made, as shown in Supplementary Figs. 6e & f. Supplementary Information - 9 -

10 S8. DFT simulation of di/dv plots of a p-n-gnr heterojunction To rationalize experimental observations of di/dv maps of the heterojunctions at various bias (Fig. 4, Supplementary Figs. 3 & 4), we also used SIESTA to obtain partial charge plots of the heterojunction at different bias to visualize the band alignment, as shown in Supplementary Fig. 7. At a bias of ±0.35 V, neither p-gnr nor N- GNR can be seen since no states fall in that energy region. At a bias of 0.55 V, the N-GNR can be observed while the undoped one cannot be. In contrast, at the bias of V, the p-gnr can be seen while the doped one is invisible. Therefore, the electronic band gap of the heterojunction is ~1.1 ev. Supplementary Figure 7 Partial charge plots of the heterostructure at various bias. The Fermi level of the system is set at 0 ev. Supplementary Information

11 These results are consistent with those shown in Supplementary Fig. 6b. When the bias increases, the interface between p- and N-GNRs begins to be visible at ±0.75 V. With the further increase of the bias magnitude, the N- (p-) GNR segment begins to be seen at (1.15) V, the onset bias where the whole system begins to have non-zero DOS and non-zero electrical conductance, in excellent agreement with Supplementary Fig. 6c. More importantly, the transition of partial charge plots with respect to the bias in Supplementary Fig. 7 well matches the experimental di/dv mappings of the heterostructure in Fig. 4 & Supplementary Fig. 4. Note that DFT is well known to considerably underestimate the band gaps of lowdimensional nanostructures R8, and hence the onset bias in the DFT partial charge plots are lower than experimental ones. The first-principles many-body Green s function approach within the GW approximation is known to yield accurate quasiparticle band gaps R11. In our case, however, the band gaps are also influenced by the presence of the Au(111) substrate. It has been found that band gaps of GNRs are reduced due to the substrate screening effect R11. For both free-standing p-gnrs and N-GNRs, the DFT band gaps are around ev. From GW calculationserror! Reference source not found., Error! Reference source not found.,r12, we estimate the band gaps to be enhanced to ~3.60 ev. The band gap reductions due to the metal substrate are estimated between 0.7 and 1.4 ev using the semi-empirical image charge model R13,R14. Therefore, the band gaps of p-gnrs and N-GNRs on the gold substrate are estimated in the range of 2.2 to 2.9 ev, in good agreement with experimental values (Supplementary Fig. 5). S9. Concept of GNR doping via monomer chemical substitution Besides the nitrogen-rich monomer 1 discussed in the main part of the manuscript (Figure 1), other oligophenylene precursors containing nitrogen (or sulfur) atoms can of course be used to fabricate GNRs with different concentrations of heteroatoms per repeating unit. In Supplementary Fig. 8, the systematic variation of doping level by nitrogen substitution is illustrated, resulting in GNRs with widely tunable n-type character. DFT calculations for GNRs with nitrogen substitution levels of 0, 1, 2,, 8 show that both the VBM and the CBM shift to lower energies with increasing nitrogen content (Supplementary Fig. 8 d). A linear fit of the energy level positions vs. N Supplementary Information

12 substitution levels yields a band shift of 0.13 ev per nitrogen atom. Supplementary Figure 8 GNR electronic level control via monomer chemical substitution. a-c, Illustration of different molecular precursors towards n-type Nitrogen substituted GNRs. Doping levels from 1 to 8 are readily conceivable. d, Energy of the valence band maximum (VBM) and conduction band minimum (CBM) for GNRs with different levels of nitrogen substitution. Energies are given with respect to the vacuum level and are computed within Supplementary Information

13 DFT for infinitely long GNRs of the indicated N substitution level. S10. GNR fabrication on Au/mica substrates and transfer to arbitrary substrates In view of technological applications of GNRs, obvious requirements are the scalability of the GNR fabrication process and the use of non-conductive substrates. To this end, we have developed a procedure to grow GNRs on gold films on mica substrates, and a convenient and clean transfer procedure for bringing the ribbons onto arbitrary target substrates. In brief, the GNRs are grown under vacuum conditions on epitaxial Au(111) thin films ( nm) grown on freshly cleaved mica substrates (PHASIS, Switzerland). The GNR/Au/mica samples are then floated on hydrofluoric acid (HF) (40 wt.%) solution, which results in detachment of the mica substrate and leaves the GNR/Au film floating on the surface of the liquid (Supplementary Fig. 9, left) which is successively replaced by ultrapure water. In a next step, the GNR/Au film is adhered to the target substrate and removed from the liquid. The Au/GNR/substrate sample (Supplementary Fig. 9, middle) is then dried and annealed at 100 C for 10 minutes to improve adhesion. Finally, the Au film is dissolved by KI/I 2 gold etchant, and the resulting GNR/substrate sample (Supplementary Fig. 9, right) is rinsed several times with ultrapure water and acetone. Supplementary Figure 9 Transfer of GNRs grown on Au/Mica substrates to nonconductive target substrates. It should be emphasized that this transfer procedure can be adapted to various substrates, such as Al 2 O 3, CaF 2, SiO 2, etc. Most importantly, it does not involve the use of an intermediate transfer membrane, such as polymethylmethacrylate, which needs to be removed afterward, and works well for large area samples. Using the prototypical 7-AGNRs that exhibit a width-specific radial-breathing-like Raman mode Supplementary Information

14 Intensity (a.u.) Intensity (a.u.) Intensity (a.u.) at ~397 cm -1 R15, we have verified that the transferred GNRs are unaltered and that the transfer results in a homogeneous coverage on the target substrates (Supplementary Fig. 10). GNR 117 on CaF 2 GNR 120 on Al 2 O 3 GNR 15 on SiO 2 GNR 117 on Au/mica GNR 120 on Au/mica GNR 15 on Au/mica CaF 2 Al 2 O 3 SiO Raman shift (cm -1 ) Raman shift (cm -1 ) Raman shift (cm -1 ) Supplementary Figure 10 Raman characterization of 7-AGNRs transferred onto CaF 2, Al 2 O 3 and SiO 2. Raman spectra (532 nm, 20 mw) from the GNRs on their Au/mica growth substrate (red curves), the clean target substrates (black curves), and the transferred GNRs (blue curves) are shown. In addition to the prominent G and D peaks and several smaller peaks due to the ultranarrow width and low symmetry R15, the GNRs also exhibit a width-specific radialbreathing-like mode at ~397 cm 1. The Raman spectra before and after GNR transfer are virtually identical, demonstrating the intact nature of the GNRs on the target substrates. S11. Supplementary References R1. Saleh, M., Baumgarten, M., Mavrinskiy, A., Schäfer, T., & Müllen, K. Triphenylenebased polymers for blue polymeric light emitting diodes. Macromol. 43, (2010). R2. Soler, J. M. et al. The SIESTA method for ab initio order-n materials simulation. J. Phys.: Condens. Matter 14, 2745 (2002). R3. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation made simple. Phys. Rev. Lett. 77, (1996). R4. Owens, J. R., Cruz-Silva, E. & Meunier, V. Electronic structure and transport properties of N2AA-doped armchair and zigzag graphene nanoribbons. Nanotechnology 24, (2013). R5. Junquera, J., Paz, Ó., Sánchez-Portal, D. & Artacho, E. Numerical atomic orbitals for linear-scaling calculations. Phys. Rev. B 64, (2001). R6. Meunier, V., Nardelli, M. B., Bernholc, J., Zacharia, T. & Charlier, J.-C. Intrinsic electron transport properties of carbon nanotube Y-junctions. Appl. Phys. Lett. 81, 5234 (2002). R7. Cruz-Silva, E. et al. Electronic transport and mechanical properties of phosphorus- and Supplementary Information

15 phosphorus-nitrogen-doped carbon nanotubes. Acs Nano 3, (2009). R8. Yang, L. et al. Quasiparticle energies and band gaps in graphene nanoribbons. Phys. Rev. Lett. 99, (2007). R9. Prezzi, D., et al., Optical properties of graphene nanoribbons: The role of many-body effects. Phys. Rev. B 77, (2008). R10. Wang, S. and Wang J. Quasiparticle energies and optical excitations in chevron-type graphene nanoribbon, J. Phys. Chem. C 116, (2012). R11. See e.g. Aulbur, W. G., Jonsson, L., and Wilkins, J. W., Quasiparticle Calculations in Solids, Solid State Phys. 54, 1 (2000). R12. Wang, S., Prezzi, D., Ferretti, A., Ruini, A., and Molinari, E., private communication. R13. Ruffieux, P., et al., Electronic structure of atomically precise graphene nanoribbons. ACS Nano 6, 6930 (2012). R14. Liang, L. & Meunier, V. Electronic structure of assembled graphene nanoribbons: Substrate and many-body effects. Phys. Rev. B 86, (2012). R15. Cai, J. et al. Atomically precise bottom-up fabrication of graphene nanoribbons. Nature 466, (2010). Supplementary Information