Electronic components embedded in a single graphene nanoribbon

The use of graphene in electronic devices requires a band gap, which can be achieved by creating nanostructures such as graphene nanoribbons. A wide variety of atomically precise graphene nanoribbons can be prepared through on-surface synthesis, bringing the concept of graphene nanoribbon electronics closer to reality. For future applications it is beneficial to integrate contacts and more functionality directly into single ribbons by using heterostructures. Here, we use the on-surface synthesis approach to fabricate a metal-semiconductor junction and a tunnel barrier in a single graphene nanoribbon consisting of 5- and 7-atom wide segments. We characterize the atomic scale geometry and electronic structure by combined atomic force microscopy, scanning tunneling microscopy, and conductance measurements complemented by density functional theory and transport calculations. These junctions are relevant for developing contacts in all-graphene nanoribbon devices and creating diodes and transistors, and act as a first step toward complete electronic devices built into a single graphene nanoribbon.

the double heterojunctions is in nice agreement with the calculations. However, while the selected spectra support that description, it is not visible from the experimental conductance maps. C onductance profiles across the double heterojunction should be added to the figure to make the claimed effects visible.
- Figure 4a: Experimental and predicted β fit almost perfectly until Δz ≈ 1. From then on, the predicted slope changes and differs from the experimental one. Do the authors have a hypothesis for that mismatch? - Fig.4b page 9: experimental data (for 5-GNR) show initial increase in the current. Authors should propose an interpretation for this phenomenology.
-Figure5a: when the single junction is lifted from the 5-GNR segment (green line) one doesn´t see a slope β≈1.2mm like in figure 4b and 5b. This should be mentioned in the text and, if possible, propose and explanation for it.
- Figure 5b and 5c: a description of the current decay variations for the do uble junctions would be appreciated. Only the central region of the long heterojunction is addressed in the text, but the rest is not described. Also an explanation on the asymmetry in 5b depending on which side is lifted would be greatly appreciated.
-Methods: Authors claim that C O terminated tips were prepared as described previously and refer to C u substrate. Did the Authors pick up C O molecule directly from Au surface? If yes, they should describe the methodology.
In addition, please correct the following typos. -In the same paragraph (page 8). Sentence: "This is also seen in the dI/dV maps: in the short double junction ( Fig.3e middle row)(…). Should be ( Fig.3d middle row). Authors should check the text carefully and fix all the editorial bugs -Page 8: "For both ribbons, the first resonance at negative bias is localized on the 7 -GNR segment (bottom rows in Fig.3e (should be 3d) and 3f)". For the short ribbon, experimental and simulated data are consistent but for the long ribbon it's not so obvious.
In conclusion Authors present state-of-art results proving that obtained GNR heterostructures can be considered as a potential building-blocks of novel nanoelectronic devices. I thus suggest acceptance of this manuscript in Nature C ommunication once these points have been addressed.
Reviewer #3 (Remarks to the Author): This manuscript reports a detailed experimental investigation of the assembly and characterization of GNR junctions. The work is adequately complemented by a dedicated theoretical analysis based on density functional calculations. In addition, the reported study includes the measurement of electronic transport properties using a "lift-off" technique based on a specific use of STM.
The field of "bottom-up" GNR synthesis is an extremely active one, following the pioneering paper from C ai et al in 2010. Since then, a number of groups have worked to reproduc e the C ai's results and have moved beyond single-GNRs, including junctions. This work is the latest addition to the existing literature. However, this work is not a mere incremental development on existing work. The originality of the work consists in: 1. Well defined junctions between metallic and semiconducting GNRs (most of the prior works have been limited to junctions between semiconducting GNRs). 2. The study of "sandwiched" systems, akin to tunnel junctions. 3. The experimental determination of transport properties of the junctions.
The paper is particularly well and logically written and I have not identified any specific and critical weakness. In other words, it is recommended that this paper be published. However, the authors should consider the following points: 1. To my knowledge this work is the first to identify the presence of pentagonal rings at the interface between GNRs (in contrast, pentagons present on the edges have been reported repeatedly). I would think that the presence of the pentagon at the junction should have important impact since its effect is to break the electron-hole symmetry, and therefore significantly affect the transport properties. Maybe the authors should discuss this some more?

Reviewer #1
We thank the referee for reviewing our manuscript, affirming its clarity and comprehensibility, and for her/his suggestions on how to improve the manuscript.
Comment: I'm wondering what is the main topic of this paper: analyzing the electrical structure of GNR or the demonstration of the GNR heterojunction, since half of the main content is on analyzing the electrical structure of GNR rather than GNR electronic components.

Response:
The main topic of the paper is the synthesis and characterization of metal-semiconductor junctions embedded in graphene nanoribbons. This involves a discussion on the synthesis through copolymerization of the precursor molecules, as well as an electronic characterization of the as-formed GNR through scanning tunneling spectroscopy and conductance experiments, and corrobated by theoretical calculations. We think our manuscript presents a complete picture of the fabrication and characterization of GNR heterostructures.  Commun. 6, 10177 (2015). The main findings and methods used in this manuscript are similar to, especially the third Nature Commun. paper. The novelty is not very obvious. It should be considered as incremental work.
Response: The first two papers mentioned by the referee report on the synthesis of semiconductorsemiconductor nanoribbon heterojunctions and have been cited in our manuscript. The third paper describes the synthesis of pure 5-acGNR. That article partly forms the basis for our work: we use the GNRs described by Kimouche et al and combine them with the well-known 7-acGNRs to generate metalsemiconductor junctions. These junctions are fundamentally different from semiconductorsemiconductor junctions. Furthermore, we report a detailed investigation on the atomic structure of the junctions and perform, for the first time, lifting experiments on nanoribbon heterojunctions. In particular the lifting experiments on graphene nanoribbons incorporating heterojunctions are unprecedented in literature. Hence, the new elements of our work are (i) preparation of metal-semiconductor heterojunctions embedded in graphene nanoribbons (ii) characterization of their geometric and electronic structure (iii) the fact that we have performed two-terminal transport experiments on graphene nanoribbon heterojunctions. These points are also recognized by the other two referees.

Response:
We agree with the referee insofar that we realize that the density variations in the dI/dV maps are quite subtle, and the orbital pattern can be hard to recognize. Several factors can contribute, i.e. the finite size of the tip, as well as "non-resonant" (background) tunneling. Finally, in the experiment, the contours of the local density of states are probed at a (much) lower contour value than the simulated maps shown in the original version of the manuscript (i.e. larger tip-sample distance). Nevertheless, the intensity variations are in fact visible from the maps. In particular, the -1.0V map in Figure 3d shows a significantly increased orbital density in the bottom of the 7-GNR segment as compared to the map at 0.1V, whereas the 1.5V map shows a clear decrease in density in the same area. The maps at 0.1V and 1.1V in Figure 3f show an increased density on the 5-GNR leads, whereas the map at -0.7V has its largest density on the 7-GNR segment.
We find that adding conductance profiles to the maps do not improve the readability of the figure. In contrast, we feel that adding more curves makes the figure more confusing, without significantly improving the clarity of the spatial electron density variations. We think the most clear way of interpreting the differential conductance maps is still to look at the density variations, and to compare them with maps recorded at different bias voltages and simulated maps. Even though the variations are not very pronounced at first sight, a closer look should still be enough to convince the reader that the maps are in agreement with spectroscopy experiments and theory.
Action: In order to convey the message that the orbital density only shows up in a subtle fashion in the differential conductance maps due to aforementioned effects, we have added the following text: The molecular orbitals are not reproduced as such in the dI/dV maps due to the finite size of the tip and the presence of background non-resonant tunneling. Nevertheless, the individual states can still be clearly recognized in the density variations over the ribbons, and are in good agreement with the results from theory. Figure 4a: Experimental and predicted β fit almost perfectly until Δz ≈ 1. From then on, the predicted slope changes and differs from the experimental one. Do the authors have a hypothesis for that mismatch?

Comment:
Response: Indeed, there is a mismatch between theory and experiment. In a more general sense, we see that the qualitative features of the I(z) curves are nicely reproduced by the calculations, but the quantitative values of the current decay are often underestimated. Any discrepancies are most likely due to the limitations of the tight binding model in describing the complicated electronic structure of the junction during the lifting experiments (bending of the ribbon, modifications of the contact with the substrate). In order to better convey this point, we have added a brief discussion to the SI.

Action:
We have added the following paragraph to the SI, explaining what we believe to be the limiting factors in the transport calculations: We note that in the calculations, the current decay parameter is often underestimated. We believe this to be due to the absence of explicit electron-electron interactions in the tight binding model, which results in an underestimation of the band gap. In the case of non-resonant transport, this error translates into an underestimation of β, since the energetic distance between the Fermi level of the tip and the HOMO and LUMO is reduced. An additional effect is that the end-localized state may contribute to the transport, but due to its confined nature, its contribution may quickly decay upon lifting. We believe this to be the reason why in the calculation of the conductance in Figure 4a, β is higher at the beginning of the I(z) curve, where the end state still plays a role. After lifting to Δz > 1 nm, the current decay converges to a monoexponential decay, the β value of which underestimates the real value due to the underestimated band gap. The fact that the same end state feature cannot be seen in the experiments could be due to the ribbon already being slightly lifted from the substrate upon being contacted, meaning that the zero height is shifted with respect to the zero in the calculations. Unfortunately, we cannot measure the absolute value of the tip height in STM.

Comment:
Methods: Authors claim that CO terminated tips were prepared as described previously and refer to Cu substrate. Did the Authors pick up CO molecule directly from Au surface? If yes, they should describe the methodology.
Response: Yes, the CO molecule was picked up directly from the Au(111) surface. The method that was developed for Cu(111) also works very well for Au(111).

Action:
In the methods section, the relevant sentence now reads: CO terminated tips were prepared by picking up a CO molecule from the Au(111) surface using a method developed for Cu (111). 34,35 Comment: In addition, please correct the following typos. -In the same paragraph (page 8). Sentence: "This is also seen in the dI/dV maps: in the short double junction (Fig.3e middle row)(…). Should be ( Fig.3d middle row). Authors should check the text carefully and fix all the editorial bugs -Page 8: "For both ribbons, the first resonance at negative bias is localized on the 7-GNR segment (bottom rows in Fig.3e (should be 3d) and 3f)". For the short ribbon, experimental and simulated data are consistent but for the long ribbon it's not so obvious.