Molecular Graphene Nanoribbon Junctions

One of the challenges for the realization of molecular electronics is the design of nanoscale molecular wires displaying long-range charge transport. Graphene nanoribbons are an attractive platform for the development of molecular wires with long-range conductance owing to their unique electrical properties. Despite their potential, the charge transport properties of single nanoribbons remain underexplored. Herein, we report a synthetic approach to prepare N-doped pyrene-pyrazinoquinoxaline molecular graphene nanoribbons terminated with diamino anchoring groups at each end. These terminal groups allow for the formation of stable molecular graphene nanoribbon junctions between two metal electrodes that were investigated by scanning tunneling microscope-based break-junction measurements. The experimental and computational results provide evidence of long-range tunneling charge transport in these systems characterized by a shallow conductance length dependence and electron tunneling through >6 nm molecular backbone.


■ INTRODUCTION
The development of molecular electronics requires materials that enable long-range charge transport.−6 These include oligoenes, 7,8 oligomethines, 9,10 oligoynes, 11,12 cumulenes, 13,14 p-phenylene radicals, 15 acenes, 16−19 and porphyrin tapes, 20,21 among others. 22,23Several studies have shown that coherent tunneling dominates over short distances, whereas incoherent hopping dominates over long distances.−33 Most importantly, some of these studies have correlated larger tunneling-to-hopping transition values for planarized conjugated backbones. 30,31,33−45 In fact, GNRs are an attractive platform for studying long-range electronic transport, given their broad structural and electronic diversity.For instance, the electronic properties of GNRs are determined by their edge structure, width, length, and heteroatom-doping.However, despite their potential and the impressive recent advances in the synthesis of monodisperse (or molecular) 36,37,39,46−68 and polydisperse GNRs, [34][35][36]38,[40][41][42][43][44]69 the charge transport properties of single GNRs remain underexplored 70−74 and, in turn, the theoretical transport models for GNRs remain unverified. Single GNR charge transport studies have been carried out almost exclusively on in situ-synthesized GNRs, 70−74 and only a few experimental studies have been reported on ex situ-synthesized GNRs if work on 2 nm long heteroacenes is considered. 16−19Therefore, expanding our abilities to investigate ex situ-synthesized GNRs will pave the way to directly investigating GNRs synthesized in solution and allow correlating results between complementary techniques.
Herein, we report the ex situ synthesis and single-molecule charge transport studies of a new series of pyrenepyrazinoquinoxaline molecular GNRs, namely, NR-6, NR-16, and NR-26 (Figure 1).These nitrogen-doped molecular GNRs show, respectively, 1.7, 4.1, and 6.5 nm long backbones displaying high stability, thanks to their hybrid zigzag-cape edge structure.The GNRs have been synthesized through a new protecting-group-free approach that simplifies their synthesis and furnishes them with diamino anchoring groups at each end (Scheme 1).Such terminal diamino groups enable the formation of stable molecular GNR junctions between two gold electrodes, which allows their single-molecule electrical investigation using the scanning tunneling microscope-based break-junction (STM-BJ) technique.Conductance measurements and theoretical calculations of this series of GNRs demonstrate long-range tunneling charge transport behavior along distances beyond 5.6 nm with a shallow length decay constant (0.068 Å −1 ) that is among the lowest for semiconducting GNRs.

■ RESULTS AND DISCUSSION
Synthesis and Structural Characterization.The synthesis of NR-6, NR-16, and NR-26 starts from 1,2,4,5tetraaminobenzene A 66,75 and pyrenetetraone B 76 (Scheme 1).These building blocks have been equipped with tri-isobutylsilyl and tert-butyl solubilizing groups, respectively, to ensure the solubility of the intermediates and the final GNRs.All of the details about the synthesis and characterization of the GNRs are given in the Supporting Information.To avoid the use of protective groups and reduce the number of synthetic steps, the cyclocondensation reactions were carried out with a 4-equivalent excess of either the tetraamine or the tetraone building block.By doing so, the formation of the double cyclocondensation adduct is favored versus the polymerization of the two components.The resulting dicyclocondensation adducts were isolated by column chromatography in all cases.NR-6 (20%) was obtained by a double cyclocondensation of tetraone B with a 4-equivalent excess of tetraamine A. Then, NR-16 (32%) was obtained by a cyclocondensation of B with a 4-equivalent excess of NR-6.Finally, NR-26 was obtained in two steps by an iterative cyclocondensation between B and NR-6.First, NR-6 underwent a double cyclocondensation with a 4-equivalent excess of tetraone B to yield tetraketoneterminated NR-12 (61%), which was subsequently cyclocondensed with a 4-equivalent excess of NR-6 to yield NR-26 (29%).
The structures of NR-6, NR-16, and NR-26 have been unambiguously confirmed by nuclear magnetic resonance (NMR) and high-resolution mass spectrometry (HRMS).The 1 H NMR spectra of the different GNRs showed peaks and integrations that perfectly match with their structure, allowing us to establish unambiguously the length of the GNR (Figure 2A, the peak assignments correspond to the lettering shown in Scheme 1).For example, the signals of the terminal pyrene protons a and of the terminal tetramines b show a constant integration of 4 in all GNRs irrespectively of their length, whereas the integration sum of signals a′ and a′′ that correspond to the inner pyrene protons integrate 8 and 12 protons, respectively, for NR-16 and NR-26.Furthermore, matrix-assisted laser desorption/ionization time-of-flight highresolution mass spectrometry (MALDI-TOF HRMS) shows molecular ion peaks and isotopic distributions in agreement with the calculated ones (Figure 2B).
The optoelectronic and redox properties were investigated by UV−vis absorption and cyclic voltammetry, respectively.The UV−vis electronic absorption spectrum of NR-6 (Figure 3) shows the typical β and ρ bands�at wavelengths ca.−85 The spectra of NR-16 and NR-26 show a similar absorption pattern in which the bands are red-shifted as a result of the extended π-system and also show a new band (α band) that dominates the spectra at low energies.This absorption pattern is again consistent with Scheme 1. Synthesis of (A) NR-6, (B) NR-16, and (C) NR-26.

Journal of the American Chemical Society
those observed for pyrene-pyrazinoquinoxaline GNRs and nanographenes. 56,57,67,86The energy of the ρ and α bands of NR-16 and NR-26 is practically the same, whereas the β band is red-shifted for NR-26.Furthermore, the molar absorptivities (ε) increase with the length of the GNRs, as illustrated by the ε values of the ρ band of NR-6 (82,780 M −1 cm −1 ), NR-16 (105,766 M −1 cm −1 ), and NR-26 (210,266 M −1 cm −1 ).The optical highest occupied molecular orbital−least unoccupied molecular orbital (HOMO−LUMO) gap (E gap ) was estimated from the onset of the longest wavelength transition.The optical E gap values decrease from NR-6 to NR-16, and then it saturates (2.61, 1.97, and 1.96 eV, respectively, for NR-6, NR-16, and NR-26).This trend is consistent with that observed on other pyrene-pyrazoquinoxaline molecular GNRs and nanographenes 56,57,67,86 in which the absorption shifts rapidly, hence the bandgap, from 6 to 10 linearly fused rings and then for backbones >10 rings, the E gap values saturate and remain almost invariable.The cyclic voltammograms carried out in solutions of n-Bu 4 NPF 6 (0.1 M) in CH 2 Cl 2 show irreversible reduction and oxidation waves in all cases (Figure S1).The potentials of the reduction waves appear at more cathodic potentials than those reported for similar molecular pyrenepyrazinoquinoxaline GNRs, 56,67 which is consistent with the presence of the four electron-donating amines.Similarly, to other molecular GNRs, 56,67 the onset reduction potentials of the first wave appear at increasingly less cathodic potentials with increasing length (−1.33, − 1.01, and −0.98 V, respectively, for NR-6, NR-16, and NR-26).The presence of oxidation processes that had not been previously observed in similar pyrene-pyrazinoquinoxaline molecular GNRs 56,67 was also attributed to the presence of the four amine substituents.
Single-Molecule Charge Transport Studies.Singlemolecule charge transport characterization of each GNR was carried out in single-molecule junctions using the STM-BJ technique. 87,88In the STM-BJ approach, an Au STM tip is driven in and out of contact to/from an Au(111) substrate in a solution of the GNRs in 1,3,5-trimethylbenzene at very low concentrations (<10 −9 M) to ensure an efficient anchoring of individual GNRs between the electrodes and to avoid molecular aggregation in solution.During the gap-opening process, individual GNRs can spontaneously bridge between both biased (a few mV) electrodes through the terminal odiamine groups (Figure 1).Thousands of current versus gap separation curves (pulling traces) are recorded, from which 20−30% of those display current plateau features (Figure 4A− C) below the quantum conductance (G 0 = 2e 2 h −1 , where e is the electron charge and h is Planck's constant).These plateau features indicate the successful formation of a molecular GNR junction and are accumulated to build conductance histograms whose maxima illustrate the most probable molecular junction (MJ) configuration (Figure 4D).Between 800 and 1000 pulling traces displaying plateaus (illustrative ones in Figure 4A−C) are selected using the same sorting algorithms and accumulated into 1D conductance histograms (Figure 4D),  yielding average conductance values of 1.7 × 10 −4 G 0 for NR-6, 4.6 × 10 −5 G 0 for NR-16, and 6.5 × 10 −6 G 0 for NR-26.To corroborate the formation of GNRs junctions, the same pulling traces displaying plateaus are accumulated in two-dimensional (2D) histograms, which visualize the conductance evolution as a function of the pulling process (Figure 4E−G).Average pulling lengths can be evaluated from the 2D histograms yielding 1.5, 3.2, and 5 nm for NR-6, NR-16, and NR-26, respectively, with the longest plateaus (see illustrative ones overlaid in Figure 4E−G) spanning nearly the full molecular length for each of the studied GNRs. Figure 4H shows an excellent correlation between the molecular length and the recorded longest plateau lengths in the break-junction experiment, adding the 0.5 nm of the Au snapback occurring during the Au−Au opening process. 17In the NR-6 case, two different molecular configurations yielding two different conductance sets are observed (Figure S2), with the high conductance one corresponding to a molecular junction configuration likely probing a shorter electron pathway, as demonstrated by the much shorter plateau length associated with the high conductance feature (see illustrative overlaid trace in Figure S2).Both NR-16 and NR-26 display a single conductance feature with a number of traces spanning the full molecular length (Figure S2).The natural logarithm representation of the average conductance values associated with the molecule-long plateau features versus molecular length (Figure 4I) yields a linear behavior with a shallow slope, resulting in a length decay constant (β) of 0.068 Å −1 .This β value is comparable to values reported for armchair GNRs. 70,73o deepen the charge transport mechanism for these systems, temperature dependence conductance measurements were carried out on the longest NR-26.−32 Theoretical Calculations.Computational calculations were carried out to support the mechanistic picture drawn above the GNR charge transport.Density functional theory (DFT) calculations (B3LYP-6-31G(d,p)) were used to get insight into the molecular structure of the GNRs.The models of NR-6, NR-16, and NR-26 illustrate a virtually flat backbone with lengths of 1.7, 4.1, and 6.5 nm, respectively.
In the case of NR-6, all quasidegenerate occupied and unoccupied orbitals spread longitudinally across the backbone, illustrating how the effect of the localization of Clar sextets on the off-linear pyrene rings in the linear conjugation is small (Figure S3).This is consistent with the high conductance values observed for NR-6.
In the case of NR-16, the electronic density in the quasidegenerate HOMO/HOMO−1 and LUMO/LUMO+1 couples is localized along nine central rings between the two terminal pyrenes, whereas in HOMO−2, HOMO−3, HOMO−4, and HOMO−5, the electronic density is localized at both ends of NR-16 (Figure S4).Similarly, in the case of NR-26, the LUMO and LUMO+1 spread along nine central rings contained across the three central pyrenes, the HOMO− 3 and HOMO−4 are localized on the central pyrazinoquinoxalines, the LUMO+2/LUMO+3 and HOMO/ HOMO−1 couples are located in-between the two terminal pyrenes, and the HOMO−4 and HOMO−5 are localized at opposite ends of NR-26 (Figure S5).
The localization of Clar sextets in the off-linear pyrene rings has a larger impact on the longitudinal conjugation of NR-16 and NR-26 that results in the compartmentalization of the orbitals in different segments along the GNR backbone, similar to what we have previously described in other pyrenepyrazinoquinoxaline molecular GNRs and nanographenes. 56,57,67,86This is consistent with the fact that the dominant resonance structures of the GNRs are best represented by Clar's sextet rule with biphenyl and anthracene groups on the pyrene and pyrazinoquinoxaline residues, respectively.This increasing level of compartmentalization of the electronic density with increasing lengths accounts for the decreasing conductance values.Remarkably, such compartmentalization does not interrupt the transport along the longer GNRs, as shown by the conductance values measured, respectively, for NR-6 (1.7 × 10 −4 G 0 ), NR-16 (4.6 × 10 −5 G 0 ), and NR-26 (6.5 × 10 −6 G 0 ), which are in the same order of magnitude typically observed in π-conjugated molecular wires.
To understand how charge transport takes place across the GNR backbones, electron transport calculations in the frame of the time-energy uncertainty relation approach to molecular conductance were performed. 89Additionally, an MO analysis of the electron transport was carried out with the help of the electron deformation orbitals (EDOs). 90To get additional insight into the role played by the metal−molecule contacts in the electron transport efficiency, electron transport calculations were first performed for the individual GNRs (see below) and then for the molecular junction models, where the GNRs are complexed between metal contacts (Figure 5, where MJ-6, MJ-16, and MJ-26 refer, respectively, to the molecular junctions (MJ) of NR-6, NR-16, and NR-26).Twenty gold atom pyramids represent the metal contacts.Despite the fact that there are some combinations and rearrangements of the GNRs and gold pyramids' MOs, most of the π-MOs at the frontier region can be identified in the MO diagram of the molecular junction (Figures S6−S8) slightly displaced in position with respect to isolated GNRs (Figures S3−S5).The calculated conductance values decrease with the GNR length (Figure S9a) and fit an exponential decay, in agreement with the experimental trend.The theoretical β value (0.035 Å −1 ) of the molecular junction models is slightly lower but consistent with that obtained experimentally (Figure S9b).This is also consistent with the theoretical β value of the isolated molecular GNRs (0.047 Å −1 ), which indicates that the binding mode has a moderate influence on electron transport (Figure S9b).
As mentioned above, EDOs enable us to apply the MO theory to understand electronic transport at the molecular scale.Introducing the EDOs in the time-energy uncertainty relation approach to molecular conductance allows the construction of individual conducting channels built from mixing occupied and virtual orbital spaces induced by the electric perturbation.From the contributions of these channels to the total conductance, it is possible to identify the relevant orbitals that are involved in the electronic transport, giving rise also to a description of the process in terms of electron and hole density regions.In Figure 5, the sum of the individual channels that contribute more than 75% to the total conductance trace is represented.This is the distribution of electrons and holes (purple and blue, respectively), which describes the decrease of electron density (hole function) and the increase of electron density (electron function) within the MJ upon the application of the electric perturbation.In all GNRs, there is a large intercalation of electron and hole regions that increases with the length of the GNR.This distribution of the electron and hole functions bears some resemblance to the charge polarization in a dielectric material.It should be noted that the electron−hole boundaries are mainly located at the pyrene and pyrazine fragments.According to these representations, if the electron transport mechanism seems to be analogous for all the systems, then a length-dependent transition between mechanisms cannot be inferred.This is in accordance with the experimental conductance measurements at variable temperatures of NR-26.
Even if the mechanism seems to remain invariable, the length-dependent conductance decays can be understood in terms of the number of electron transport channels and of the frontier orbitals involved in the process and their distribution within the molecule.For instance, there are two main channels in MJ-6 (Figure S10); the most important involves the HOMO, where the electron density is spread over the NR-6 structure, and a set of quasidegenerate unoccupied orbitals mainly located at the gold tips and the outer rings of the GNR.This leads to a channel that spreads over the whole molecular junction.The second channel, however, arises from the mixing of both occupied and unoccupied orbitals located at the outer rings and the gold tips, leading to an electron transport channel without the contribution of the central pyrene ring.In MJ-16 (Figure S11), the number of relevant channels increases to three, where the first two channels build up more than 60% of the total conductance.There are, however, significant differences between these channels and those from MJ-6.Thus, they are split and localized at different regions along the MJ (similarly to MOs), which, in combination, allow for an effective edge-to-edge charge transfer.The first channel, which is mainly distributed along the GNR edges, involves only the HOMO−2 and HOMO−6 orbitals, where the electronic density is concentrated at the edges and center of the GNR, respectively, and unoccupied orbitals are mainly located at outer rings and the gold tips.The second channel involves the HOMO−3 orbital located at the edges and gold tips and the LUMO+4 orbital mainly distributed in the central region of the GNR.Except for the central pyrene ring, this channel covers most of the backbone of the GNR.The remaining channel is mainly formed by HOMO−6 and LUMO orbitals, located mostly in the GNR's center, leading to a channel that runs between the furthest pyrene rings.In MJ-26 (Figure S12), the number of relevant channels increases to four, where the first three channels build up more than 60% of the total conductance (26, 23, and 18%, respectively).These channels involve occupied orbitals of the NR-26 structure and a large set of unoccupied orbitals located at the tips and edges.Due to the different weights of these orbitals in each transport channel, there are noticeable differences between them.The first channel covers most of the backbone of the GNR, with the exception of the central pyrene and outer pyrazine fragments.The second channel is mainly located at the GNR's edges, while the third channel again involves most of the MJ.The fourth channel, however, is the one that covers the largest surface of the GNR.These results are also consistent with those obtained for the isolated GNRs (Figures S13−S16).

■ CONCLUSIONS
We synthesized a series of diamino-terminated molecular GNRs that reach 6.5 nm in length.The terminal diamines have proven to be optimal anchoring groups to prepare singlemolecule Au-GNR-Au junctions by using an STM-BJ setup.This family of GNRs shows a mixed zigzag-cape edge structure that is known to compartmentalize the π-system, providing them with high stability.Conductance measurements show that this compartmentalization does not interrupt the longitudinal transport along the GNRs, which deliver conductance values in the same order of magnitude as those typically observed in π-conjugated molecular wires, giving rise to a low length decay constant (0.068 Å −1 ) that is among the lowest measured for semiconducting graphene GNRs. 70,73ariable temperature conductance measurements, together with computational calculations, suggest that charge transport is dominated by electron tunneling within the studied length scales.The conductance decay with length can be understood with regard to the number of conducting channels involved in the process and its distribution within the molecule.Overall, we reveal a comprehensive picture of the charge transport phenomena that arise in long Au-GNR-Au junctions.The study of electron transport through such junctions shines light on the fundamental charge transport properties of GNRs and opens up prospects for the study of nanoscale charge transport for other types of ex situ-synthesized GNRs.

Figure 1 .
Figure 1.Molecular GNR junctions studied in this work (the molecular formula indicates only the atoms in the aromatic core, highlighted in gray).

Figure 4 .
Figure 4. Charge transport results by STM break-junction.(A−C) Illustrative individual pulling traces for NR-6, NR-16, and NR-26, respectively.Panel (D) shows 1D conductance histograms built accumulating between 800 and 1000 traces displaying conductance plateau features.Peak maxima are extracted via a Gaussian fit (from the linear scale), representing the average conductance for each GNR.(E−G) 2D histograms accumulating hundreds of traces displaying current plateaus.Overlaid individual traces illustrate examples of long plateau lengths spanning the molecular length.(H) Correlation plot representing plateau length (including Au snapback) versus molecular length.(I) Semilog graph of GNR conductance versus molecular length.(J) Arrhenius representation of molecular conductance against 1/T.Applied bias voltage was 30 mV for NR-6, 200 mV for NR-16 and 300 mV for NR-26.

Figure 5 .
Figure 5. Sum of the individual channels, which contribute more than 75% to the total conductance trace of the molecular junctions MJ-6, MJ-16, and MJ-26 and the gold tips under the bias voltage of 0.1 V calculated at the B3LYP-6-31G(d,p)/lanl2dz level of theory (e-function (purple) and hole function (blue)).