Spiers Memorial Lecture Carbon nanostructures by macromolecular design – from branched polyphenylenes to nanographenes and graphene nanoribbons

Nanographenes (NGs) and graphene nanoribbons (GNRs) are unique connectors between the domains of 1D-conjugated polymers and 2D-graphenes. They can be synthesized with high precision by oxidative ﬂ attening processes from dendritic or branched 3D-polyphenylene precursors. Their size, shape and edge type enable not only accurate control of classical (opto)electronic properties, but also access to unprecedented high-spin structures and exotic quantum states. NGs and GNRs serve as active components of devices such as ﬁ eld-e ﬀ ect transistors and as ideal objects for nanoscience. This ﬁ eld of research includes their synthesis after the deposition of suitable monomers on surfaces. An additional advantage of this novel concept is in situ monitoring of the reactions by scanning tunnelling microscopy and electronic characterization of the products by scanning tunnelling spectroscopy.


Introduction
The elucidation of the benzene structure and the discussion of the aromaticity concept are classic topics in organic chemistry. [1][2][3] Benzene is a fascinating species, not only as a model for spectroscopic and theoretical studies, but also as a starting compound for chemical functionalization and the subsequent synthesis of natural products and as a versatile building block for conjugated psystems. One of the oldest challenges of organic chemistry, the synthesis of colourants, would not have been possible without benzene chemistry. These areas might appear to be distinct at rst glance, but a very good case for their association can be made by relating two major breakthroughs in materials science, conjugated "conducting" polymers as synthetic metals [4][5][6] and graphenes as new carbon allotropes, 7,8 to structural and functional concepts derived from the study of benzene compounds. Going from benzene to biphenyl, terphenyl and higher oligophenyls builds up a "box" of delocalized, polarizable electrons, and this property enables the chain molecules to interact with light and serve as chromophores or to undergo electron transfer and serve as "electrophores". One can even argue that key processes in modern energy technology, such as the transformation of light energy into electrical energy in solar cells [9][10][11][12] or the storage of charge in batteries, [13][14][15] would have remained impossible without the concepts developed during the study of benzene and conjugated p-systems.
While oligophenyls are chain structures, the benzene p-system can be extended in another way, by fusing the hexagons to naphthalene, anthracene and higher acenes, or by creating other topologies such as those in phenanthrene, triphenylene and higher polycyclic aromatic hydrocarbons (PAHs). The description of the resulting rod-and disc-type p-systems based on molecular orbital theory and spectroscopy has been instrumental for modern organic chemistry. Notably, however, a comprehensive discussion of such conjugated systems cannot be restricted to molecular features but must also consider supramolecular chemistry. 2,16,17 By attaching different aliphatic chains at the peripheries of a PAH, various packing modes and phases can be realized in the solid state, depending on the arrangement of chains and discs, which determines the kinetics of intermolecular processes such as exciton or electron transfer. Equally important are interfacial structures, such as the edge-on or face-on arrangement of conjugated molecules on surfaces.
Two characteristics are crucial for the molecular and supramolecular description of p-systems and provide powerful design concepts: size and dimensionality. Building homologous series by the stepwise addition of another repeating unit is one of the key principles when trying to "synthesize" desired properties. 2,18,19 Additionally, plotting a measurable quantity, such as the wavelength of an absorption band or the redox potential, against the number n of repeat units and extrapolating to innite n offers deep insight into the nature of the corresponding polymers. 20 Rigid-rod-type poly-para-phenylene (PPP) with an "innite" chain can be regarded as a prototype of a conjugated polymer (CP). A similar argument holds for 2D-polymer graphene, which is thus the innite form of a PAH.
The organic chemistry of small benzene-based unsaturated molecules and the materials science of materials such as PPP and graphene have largely advanced as separate disciplines, oen kept further apart by different nomenclatures. Two main criteria distinguish an organic material from an organic compound. For a material, (i) one examines a complex function as part of a complex system, e.g. a device, which cannot be achieved by just recording a spectrum, and (ii) one must create a dened macroscopic state of matter such as a bre or a thin lm. One readily concludes that organic chemistry would have much synthetic expertise to contribute if it were not restricted to focusing on dilute solutions.
Chemistry has become more interdisciplinary over recent years, and while classical topics such as synthetic method development will always be important, the challenges dened by materials science have opened new horizons. A similar stimulus came from another emerging disciplinenanoscience. Two achievements have helped to bring about nanoscience, namely the ability in physics to "see" small single molecules or particles, and the ability in chemistry to fabricate molecules or particles extending over different length scales. The example of graphene is instructive. Graphene can be fabricated by exfoliating a single layer from bulk graphite. 21,22 The critical features of such a "top-down" protocol are the size of the resulting sheet and the avoidance of multilayers. In an alternative approach, described as "bottom-up", one could consider synthesizing increasingly larger PAHs, and ultralarge PAHs have been coined nanographenes (NGs). [23][24][25] Compared with the hot eld of graphene, the study of CPs has already seen several ups and downs. Following research on the science and technology of synthetic metals, based on the occurrence of electrical conductivity due to partial electron transfer and lost momentum, the light emission from CPs in electroluminescent devices attracted attention for use in displays and lighting systems, [26][27][28] followed by the application of CPs in solar cells as light absorbers and electron donors. [9][10][11][12] One should be careful not to prematurely describe a research eld as mature. CPs offer nearly unlimited opportunities for structural variations and thus synthesis-driven property control. However, problems of a fundamental and technological nature remain. One drawback relates to the stability of devices, which may depend not only on the prevailing structure but also on the occurrence of defects and impurities. Another drawback relates to the special situation in a device such as a eld-effect transistor (FET). The charge-carrier transport, an important performance criterion, is dependent on the supramolecular order between the molecules and their arrangement with respect to the substrate. The resulting morphologies can thus be inuenced by the molecular structures and by the applied processing techniques, and the fabrication of thin-lm devices by unavoidable trial and error may lead to undesirable packing modes. 29,30 This uncertainty has strengthened the desire for electronic materials with either high intrinsic charge-carrier mobility or robust, controllable packing motifs. Such materials can potentially be realized with graphene and NG molecules. Graphene was shown in the pioneering work of Geim and Novoselov to possess an ultrahigh charge-carrier mobility, which qualies it for use as an active material in FETs. [31][32][33] However, the problem is that graphene also features a vanishing electronic bandgap, which would never allow an off state of the current owing in an FET. A switch that is never off is not practical, so the bandgap must be opened. As suggested by theory, geometric connement of the graphene layer, such as that in a graphene nanoribbon (GNR), is the method of choice to open the bandgap. 34,35 A GNR is a quasi-1D p-system, as opposed to 2D-graphene, and thus bears a close resemblance to the classical 1D-CPs described above (Scheme 1). Doublestranded CPs (ribbon polymers) have already played an important role in polymer science and can pave the way to GNRs. [36][37][38][39] From a synthetic point of view, the Scheme 1 From branched polyphenylenes to GNRs and infinite graphene sheets. transition from CPs to GNRs is far from trivial, and the underlying protocols are one of the key considerations of this text.
The reader might thus consider GNRs as bridges connecting the graphene and CP elds, and this view appears to have some merits. Nevertheless, the present text is not restricted to simply making CPs slightly better. Instead of only looking at classical cases of charge-carrier transport or light emission, the focus will be on where GNRs and their NG model systems open new avenues for fundamental research and technology. We anticipate that the bandgaps of GNRs can be tailored by the width of the polymerand by the nature of the periphery. 24,40-42 A zigzag periphery is particularly relevant since it holds promise for creating edgelocalized electronic states. These, in turn, are the origin of exotic quantum states giving rise to high-spin systems and topological insulators. 43,44 Stimulated by pioneering theoretical work, the importance of GNRs has been recognized by physicists and materials scientists. However, the methods of GNR fabrication, such as graphene lithography or slicing open carbon nanotubes, suffer from a lack of control over the edge structures, [45][46][47][48][49][50] which is necessary to truly enter the unknown territory just mentioned. Herein, we therefore dwell on the power of precision polymer synthesis, which is initially a challenge for solution chemistry. An additional, closely connected concept is GNR synthesis aer the immobilization of monomeric building blocks on metal surfaces. In both cases, the expertise in synthesizing unique polymers and uncovering their unconventional electronic properties must be closely connected. Understandably, this text is a tribute to both the synthetic chemists from our group and the physicists from our partner groups.

Solution synthesis
A pioneer of PAH chemistry is Erich Clar, who has contributed not only to the synthesis of PAHs but also to their electronic and spectroscopic characterization. 51 Aer a long time during which PAHs were mainly considered from a toxicological perspective, their chemistry has experienced a remarkable renaissance since the mid-1990s. This has been further stimulated by the rise of graphene and the fact that both the chemical and physical properties of PAHs are of strong signicance for the graphene eld. In this spirit, we described the emergence of "PAHs in the Graphene Era" in a recent review that the reader may want to consult. 52 Herein, ultralarge PAHs, which we name NGs, serve as model cases for the synthesis and structural elucidation of GNRs.
Extending the size of PAHs to the nanoscale becomes possible by "dimerization" via the fusion of biaryl precursors under oxidative conditions in the Scholl reaction. A known case is the transition from 1,1 0 -binaphthyl to perylene, while the fusion of perylene to quaterrylene can be regarded as an intermolecular version. More recently, we used bichrysenyl 1 and tetrachrysenyl 3 as starting compounds to achieve fusion to realize short segments of the cove-edge GNRs 2 and 4, respectively (Scheme 2A). 53,54 A more elegant concept that allows a much steeper increase in size comprises the synthesis of a polyphenylene precursor whose topology allows full attening of the benzene rings into the PAH plane by cyclodehydrogenation, namely the Scholl reaction. The synthesis of hexa-perihexabenzocoronene (HBC) 5 from hexaphenylbenzene is the most prominent example. HBC, with a large hexagonal aromatic structure, can be well described as a superbenzene. Interestingly, supernaphthalene and other higher homologues of HBC with varying symmetries, including C72 6, C96 7, and C138 8, have become accessible starting from appropriately designed polyphenylene precursors (Scheme 2B). 55,56 Astonishingly, polyphenylene 9 can be transformed into C 222hexagon 10 (Scheme 2C). 57 The crystal structure of the soluble precursor 9 nicely exhibits the presence of twisted, tightly packed and interlocked benzene rings. Molecule 9 is just a dendrimer made only from benzene rings with intra-and intermolecular voids ready for the uptake of guest molecules. In a separate strand of our research, increasingly larger polyphenylene dendrimers with molecular weights up to 1.9 MDa could be synthesized by a divergent route and characterized as monodisperse, structurally perfect 3D polyphenylenes. What makes these dendrimers different from all other dendrimers is that they are semi-rigid, since their arms do not undergo back-bending. Therefore, the incorporation of functional groups such as chromophores at the core, in the scaffold or on the rim of the dendrimers leads to perfect nanosite denition and provides, for example, unique light harvesting systems with controlled transfer of the excitation energy. Another issue is important in the present context: multiply branched or dendritic polyphenylenes with suitable structures are carbon reservoirs for the formation of graphenic molecules by oxidative 3D-to 2D-transformations.
The synthesis and structural characterization of PAHs and their larger NG congeners is severely hampered by their limited solubility in organic solvents, which arises from the pronounced tendency of the disc-type molecules to aggregate. A surprisingly simple way of solubilizing rigid discs is exhaustive edge chlorination, by treating suspensions of the hydrocarbons with iodine chloride (Fig. 1A). 58 The sterically encumbered chloro substituents cause strong ring puckering; thus, sufficient solubility for crystallization is induced ( Fig. 1B and C). Structural analysis gives information of crucial importance to describe the bonding situation in NGs, such as bond length alternation and packing in the solid. Attaching long alkyl chains and thus affording a higher solvation entropy is another way of dealing with the solubility issue. This structural variation, except when using very bulky substituents, does not avoid aggregate formation in solution and even promotes the formation of discotic mesophases with a columnar liquid-crystalline order in the melt. [59][60][61] Notably, this phase formation under stacking of the discs does not necessarily result from attractive p-p interactions, but rather from nanophase separation between the "hard" aromatic cores and the "so" aliphatic shells. For a long time, hexa-alkoxy triphenylenes have been regarded as the main model systems for discotic mesophase formation. [62][63][64] The HBC congeners have much more stable phases, which allow systematic studies of charge-carrier transport along the columnar superstructures. Furthermore, mesophases can persist at temperatures at which thermal carbon-carbon cleavage within the alkyl chains and subsequent crosslinking occur. This is reminiscent of carbomesophase formation during the graphitization of pitch. Similar "connections" are possible, such as in the fabrication of carbon nanobres by pyrolysis of either polyacrylonitrile or pitch. 25,[65][66][67][68][69] Again, complex crosslinking reactions occur in a preorganized phase, but the detailed chemical processes are oen described in rather speculative terms. Model reactions under controlled conditions are urgently needed. A helpful case is the use of poly(methyl vinyl ketone) as the starting compound because acene formation can proceed by an aldol condensation rather than thermolytically. [70][71][72] Revitalized PAH chemistry can branch out into many other areas since PAHs are much-needed model compounds, for example, in astrophysics and the analysis of asphaltenes. 73,74 Importantly, PAHs play a key role as part of the emerging nanoscience eld. PAHs could be deposited into ordered monolayers on conducting surfaces such as metal or highly oriented pyrolytic graphite (HOPG), and the resulting patterns could be visualized by scanning tunnelling microscopy (STM), oen with atomic resolution. Ground-breaking work in this direction was performed by Roman Fasel and his group. While, for example, HBC and p-extended HBC with two additional rings, or K-regions, could be sublimed under ultrahigh vacuum (UHV) conditions, higher NGs escaped similar treatment due to the higher sublimation point and thermal decomposition. 75 Additionally, having soluble, alkylsubstituted NGs available allowed their deposition from solution. The pioneering experiments of Jürgen Rabe and Paolo Samorí demonstrated the formation of an ordered monolayer by physisorption of alkyl-substituted HBC 14, visualizing single molecules with the STM tip immersed in the supernatant solution on the substrate ( Fig. 2A and B). 76,77 While the crystallization of large NGs is hardly possible and their NMR spectra are severely broadened due to the strong aggregation tendency, STM provided an additional possibility for their characterization, visibly revealing their structures. One can thus conclude that the synthesis of large NGs and their visualization by STM were closely connected. Even more important from an electronic point of view was a subsequent scanning tunnelling spectroscopy (STS) experiment that allowed one to record separate currentpotential curves for the aromatic and aliphatic domains of individual molecules, thus disclosing their electronic differences. Later, even more sophisticated electronic functions of a single molecule could be resolved by the Rabe group for dyad molecule 15 (Fig. 2C). 78,79 Aer physisorption, HBC cores can be identied at the corners as bright circular features with central spots corresponding to high tunnelling currents, together with four anthraquinone (AQ) moieties in between (Fig. 2D). The AQ moiety was subjected to donor-acceptor formation with an electron-rich dimethoxyanthracene (DMA) added to the solution ( Fig. 2E and F). The dipole arising within the donor-acceptor complex acted as a kind of gate to modulate the current through the HBC. This design could therefore be regarded as a single-molecule FET with source and drain electrodes formed by the STM tip and the conductive substrate.
Since its beginning, the application of the Scholl reaction for HBC synthesis has offered many surprises. 80 The stepwise planarization could be stopped by isolating the dibenzo[fg,ij]phenanthro[9,10,1,2,3-pqrst]pentaphene with a single twisted phenyl substituent. 81 Additionally, planarization of the hexaphenylbenzene precursors worked even in the case of pronounced steric hindrance and thus nonplanarity of the new p-system. Thus, dodecamethyl hexaphenylbenzene 16 gave HBC 17, while octadecamethoxy derivative 18 yielded 19, which, due to its doubly concave structure, could bind C 60 fullerene, yielding the crystalline columnar structure displayed in Fig. 3. 82,83 The electronic structure of HBC can be described as consisting of only "Clar sextets" and thus as belonging to a family of fully benzenoid PAHs with large highest occupied molecular orbital (HOMO)-lowest unoccupied molecular orbital (LUMO) gaps and pronounced chemical stability. [84][85][86][87] This situation changes when additional rings are formed at the periphery. To understand the rich opportunities of this extension, introducing the different PAH peripheries of graphenic sheets is important (see Scheme 3). While hexa-peri-hexabenzocoronene (yellow) features armchair edges, its analogue hexa-cata-hexabenzocoronene (violet) possesses coves at the edges. The "upper" part of perylene (sky blue) has a zigzag edge, whereas its bay area represents an armchair. From a chemical point of view, it is relevant that the bay region can act as a diene for Diels-Alder cycloaddition with reactive dienophiles to furnish a new six-membered ring, affording ovalene 21 from bisanthene 20, as displayed in Scheme 4A. 88 This scheme also presents an example of PAH modication using the ring closure of an ethynyl unit attached to a bay area (Scheme 4B). 89 Depending on the topology of the twisted oligophenylenes and oligoarylenes, six-membered ring formation has also been accomplished by Friedel-Cras reactions, as exemplied by the synthesis of 25 (Scheme 4C). 90 In the meantime, HBC and its oligomers have found widespread use as building blocks to synthesize various NGs, and this manifold of hydrocarbons includes nonplanar and chiral structures. [91][92][93][94] However, a full account of this rich chemistry is beyond the scope of this text.
The stepwise extension of HBCs with extra rings, or K-regions, has been accomplished within the series of 26-29, where the successive lowering of the HOMO-LUMO gap was reected by the bathochromic shi of the electron absorption bands (Fig. 4A). [95][96][97] A recent case is 29, where HBC was extended by  While such open-shell NGs are typically unstable, dibenzo[hi,st]ovalene (DBOV) 35, which also has zigzag edges, displays closed-shell properties with very high stability (Fig. 5A). [106][107][108] DBOV was revealed to have sharp absorption and emission peaks with a small Stokes shi of 12 nm (Fig. 5B). DBOV exhibited strong red emission with a high photoluminescence quantum yield of $80%, which could later be improved up to 97% by substituting the bay regions with aryl groups. 109 Although aggregation-caused uorescence quenching, a typical problem for large aromatic compounds, was observed, solid-state emission and   emission were achieved by using a polymer matrix. Moreover, ultrafast transient absorption measurements by the group of Francesco Scotognella elucidated stimulated emission from DBOV at 695 nm, indicating the potential of DBOV as an optical gain material. Amplied spontaneous emission could be further demonstrated, rendering such NGs highly promising not only for organic light emitting diodes but also organic laser applications. Wu, Casado, Díaz-García and their co-workers later demonstrated the fabrication of lasers using NGs, providing a proof of concept. 110 Moreover, DBOV and other NGs, including HBC, were revealed by the group of Mischa Bonn to show uorescence blinking on timescales useful for modern super-resolution microscopy, including singlemolecule localization microscopy and minimal emission ux microscopy. 111 Remarkably, DBOV continued to blink in various environments, including air, and maintained blinking behaviour aer 125 days of storage, in stark contrast to standard organic dyes such as Alexa 647. The latter requires a special buffer and stops blinking aer a few hours. NGs thus also have high potential for imaging applications with super-resolution microscopy, both in bio-related studies and materials science. Additionally, we have also regarded DBOV as a new building block for constructing novel and even larger NGs. Thus, p-extension at the two bay regions of DBOV by annulation of diarylacetylene led to circumpyrene 37 with multiple zigzag edges ( Fig. 5C and D), which represents an exciting NG structure studied theoretically but never previously achieved. 112

On-surface synthesis
As early as 1999, the group of Christof Wöll deposited a hexaphenylbenzene derivative with alkoxy groups on copper surfaces and heated the resulting layers to temperatures above 710 K. 113 The oxygen atoms were thought to serve as anchors to prevent the organic compound from desorption under UHV and upon heating. X-ray photoelectron spectroscopy (XPS) and near-edge X-ray absorption ne structure analysis of the product proved the successful planarization of the HBC species. The formation of PAHs by thermally induced dehydrogenation of oligophenylene precursors on a metal substrate, the on-surface variant of the above-described Scholl reaction in solution, has achieved great importance, leading to a variety of NGs and GNRs with fascinating properties in the last decade. While the resolution of STM images was limited in early studies, Roman Fasel demonstrated monitoring of the ring-closing processes of cyclophane 38 to triangular PAH 39 by high-resolution STM, providing unambiguous proof of the structure with atomic resolution (Fig. 6A and B). 114 The detailed mechanism of the reaction is still unknown, although its core process can be formally described as an electrocyclic ring closure of an o-terphenyl to a triphenylene moiety accompanied by the loss of hydrogen, which is also reminiscent of the stilbene-(dihydro) phenanthrene transformation. 115,116 Photochemical cyclization of 40 to coronene 41 in solution is formerly the same as the conversion of 38 to 39 on surface. 117 These PAH syntheses are intramolecular processes. Later, we will consider diradicals as reactive intermediates for polymerization on surfaces, leading to GNRs (see Section 3.2). A kind of preparatory case is the reaction of cation 42 (Scheme 5A). 118 In solution, if subjected to deprotonation upon treatment with amine, this cation gives rise to polycyclic azomethine ylide 43. This species can be described as either a 1,3-dipole or a diradical and undergoes a typical 3+2cycloaddition with olens as dipolarophiles, providing a protocol to construct nitrogen-containing PAHs (Scheme 5B). 119 Furthermore, Ito, Nozaki and their coworkers demonstrated the annulation of 43 with fullerene and even corannulene, 120,121 highlighting the unique reactivity of 43. This is a recent example of the famous Huisgen chemistry. Moreover, as demonstrated by Carlos-Andres Palma, 43 can also undergo dimerization and oxidation in solution to realize pyrazine-incorporated hexabenzoperylene (HBP) 47 (Scheme 5C), or dimerization on a metal surface followed by cyclodehydrogenation to yield 48 (Scheme 5D and E). 122 Notably, both 47 and 48 are antiaromatic diaza-hexacycles.
One might conclude that thermal on-surface reactions basically replicate the solution reactions, but this is not true. Oxidative cyclodehydrogenation in solution is a complex process that depends on the conformation of the polyphenylene precursor and the prevailing spin and charge density distributions within the intermediate (radical) cations. Thus, tetranaphthyl-p-terphenyl 49 affords double [7]helicenes with racemic (50) and meso (51) forms, which can be separated by crystallization utilizing their distinct solubilities (Fig. 7A). 123 The enantiomeric forms of chiral isomer 50 can be separated by chiral high-performance liquid chromatography and do not reveal any racemization at room temperature due to the very high isomerization barrier, calculated to be 46 kcal mol À1 by DFT. Assuming a radical cation species as an intermediate for ring closure, the spin density at the a-position of its naphthyl group is much higher than that at the bposition, so carbon-carbon bond formation occurs for the former. In contrast, as proven by Lifeng Chi, an on-surface reaction of the same precursor 49, when thermally induced aer adsorption on a metal surface, yields the "fully collapsed" benzo-fused perihexacene 52 (Fig. 7B). 124 The interaction between the p-system and the metal appears to favour planarization.
One can thus change the course of a reaction in solution by immobilizing a precursor on a surface; more importantly, one can obtain access to molecular structures that cannot be made or are too unstable in solution, such as extended acenes, 125,126 peripentacene, 127,128 and triangulenes. [129][130][131] NGs thus appear to be ideal model cases for the new chemistry of GNRs. This holds for the (i) synthesis, (ii) spectroscopic and microscopic methods of structure proof, and (iii) understanding of their electronic structures.

Solution synthesis
The synthesis of a CP, which has been amply described in previous dedicated reviews, 4,10,132-134 requires a suitably functionalized building block from which two new bonds, usually carbon-carbon bonds, are formed for incorporation into a chain. Oen, transition metal catalysis is involved, such as in the polymerization of acetylenes to realize polyacetylenes or the polycondensation of functionalized arenes to obtain polyarylenes. In these cases, the connection of the repeating units is accomplished via single bonds. Cases can also be found, for example, in the synthesis of poly(phenylene vinylene)s or their nitrogenincorporating analogues, where the coupling process directly furnishes double bonds. 135,136 The synthesis of GNRs is more complex, both conceptually and experimentally, since the growth of the polymer requires the formation of multiple bonds in sequential processes. Referring again to NG synthesis as a test case, one would have to rst synthesize polyarylenes (e.g., with naphthalene or anthracene units) or polyphenylenes with additional oligophenylene side arms as precursor polymers. These would then be subjected to a planarization step, possibly mild graphitization, via additional carbon-carbon bond formation by either dehydrogenation or the loss of hydrogen halides or halogens. The prime target is then the synthesis of precursors, such as branched polymers like 55 and 58. Two critical issues thus arise, namely the synthesis of the precursor polymers, for example, 55 and 58 by Suzuki and Yamamoto-type coupling, respectively, and the subsequent dehydrogenation (Scheme 6A and B). [137][138][139] Regarding the rst step, the molecular weights are known to be limited by side reactions, especially the loss of functional groups. The possibility of synthesizing multiply branched polyphenylene 61 by repetitive Diels-Alder cycloadditions of AB-type building block 60 has therefore been a major breakthrough (Scheme 6C). 140 This approach has two main advantages: the formation of each new benzene ring is "driven" by the extrusion of carbon monoxide, and no side reaction occurs as in the above protocols. Remarkably, polymer 61 could be obtained as a soluble material that, according to analysis by size-exclusion chromatography and laser light scattering, had an average molecular weight of $600 kDa. Size, as shown for NGs, is of course a critical issue, not only for electronic reasons but also for when one wants to bridge the channel of an FET with GNRs as semiconductors. Notably, the concept of repetitive Diels-Alder cycloadditions has also been extended to the synthesis of polyphenylene dendrimers. [141][142][143][144] The diene and dienophile functions have been combined with the AB 2 moiety, which serves as a branching unit. A second critical issue in GNR synthesis is cyclodehydrogenation by the Scholl reaction.
The fact that giant NGs can be made from polyphenylene precursors by the loss of more than a hundred hydrogen atoms through the formation of multiple new carbon-carbon bonds is amazing. In the case of GNRs, this fusion has even been extended to polymers with ultrahigh molecular weights.
Does this synthetic protocol leading to GNRs meet the strict requirements of CP synthesis in terms of structural perfection, processability and scalability? The question regarding failure of the attening process cannot be set aside. First, electron absorption spectroscopy results reect the formation of extended psystems without being able to rigorously exclude residual 3D units. Raman and IR spectroscopy serve as additional tools to detect possible defects resulting from incomplete dehydrogenation. STM visualization, aer the deposition of polymers from solution, provides additional microscopic evidencehowever, without being able to exclude the possibility that defective molecules have gone undetected. An advantage of precursor protocols for polymer synthesis is oen that the precursors are more soluble and can be processed from solution, followed by transformation into the target polymer. 145 In the case of GNR 62, surprisingly, solution liquid-phase processing is still possible. As described for NGs in Section 2.1, this is due to the attached alkyl chains. In both cases, however, one expects pronounced aggregation. The synthesis of GNRs can be modied for the incorporation of functional substituents to build in chromophores or stable free radicals, which will be useful later. [146][147][148][149][150] Additionally, GNRs can be subjected to further transformation by employing chlorination 58 and potentially hydrogenation, 151 as already established for NGs. Regarding scalability, the above-described synthesis has been performed on a gram scale and can certainly be developed for even larger quantities. Furthermore, an advantage of this repetitive cycloaddition is that the AB-type moieties can be modied to make the ribbons increasingly widerwithout ever giving up the strict requirement of structural perfection. One approach uses the somewhat larger tetraaryl cyclopentadienone 63, according to Scheme 6D, 152 and another employs Diels-Alder additions not only for "longitudinal" polymer growth but also for "lateral" broadening of the ribbons. 153 The key trick in the latter method is to avoid "spatial overlap" of benzene rings, since they must all t into one plane upon cyclodehydrogenation.
Structural perfection and processability are mandatory requirements for studies on nanoscience and device fabrication. Solution deposition of GNR 62 allowed the fabrication of FETs. [154][155][156] Therefore, the active components were present as a network of ribbons whose contact resistance was the limiting factor for the chargecarrier mobility. The fabrication of single-GNR FETs will be described below. A preliminary experiment aimed at the electronic characterization of single GNRs was performed that consisted of depositing 62 onto a channel formed from metal electrodes and recording a current-potential curve ( Fig. 8A and B). 157 GNRs provide new prospects in many other areas of functional studies. 158,159 The exciton dynamics of GNR 62 were investigated by ultrafast transient absorption spectroscopy (Fig. 8C). 160 Stimulated emission could be detected, which could be assigned to biexciton formation, which possessed a surprisingly high binding energy (Fig. 8D). We will further exemplify the power of solution synthesis of GNRs when we discuss even more complex chemical modications in Section 4, but introducing the complementary role of on-surface techniques as a new route to NG chemistry is helpful at this point.

On-surface synthesis
On-surface reactions of reactive monomers and polymerizations of their reactive intermediates were studied by Bent in 1993, 161 followed by many ingenious attempts at obtaining, for example, polyuorenes aer thermally induced carbon-halogen cleavage of uorene species and subsequent polymerization. 162 The reaction could be monitored by STM. We could utilize the concept proposed by Roman Fasel and his group for the on-surface synthesis of polyanthrylene 66 at 200 C, followed by thermal dehydrogenation to realize the target GNR 68 at 400 C (Fig. 9A). 163 Visualization of the new GNR and the perfection of its edges by STM is impressive (Fig. 9D). This ribbon, along with its width, is denoted 7-AGNR, where "A" indicates the armchair periphery and 7 is the number of carbons counted across the ribbon. As we have described in detail elsewhere, the width of the ribbon allows synthetic control of the nite bandgap in an atomically precise way. 24,52 To make GNRs wider, one can just use a different monomer design. As can be seen from Fig. 9B and E, dibromo-o-terphenyl 69 serves as a precursor to fabricate a 9-AGNR. 164 The synthesis of this monomer might appear straightforward at rst glance. However, its purication by repeated crystallization is crucial, since even traces of the monobromo compound would be preferentially sublimed and act as endcappers, thus limiting the achievable molecular weight. The narrower counterpart, 5-AGNR 74, predicted to have a particularly low bandgap and thus be favourable for achieving high chargecarrier mobility, can be obtained from functionalized naphthalene or perylene precursors. Thus, as has been shown by the group of Lifeng Chi, the commercially available tetrabromonaphthalene 72 gives rise to 5-AGNR 74 without requiring dehydrogenation (Fig. 9C and F). 165 The same product can be obtained from, for example, a dibromoperylene, following thermal dehydrogenation of the twisted precursor polymer. 166 Access to 5-AGNRs becomes possible by onsurface and solution polymer chemistry. 167 The tetrabromoperylene is obtained by Hunsdiecker decarboxylation of the perylene tetracaboxiimide. 168 This opens an interesting pathway from colourant to graphene chemistry. We also developed the perylene-based chromophore into a homologous series of dyes with absorption bands far into the near-infrared (NIR) range. [169][170][171] These structures comprise 5-AGNRs. Other sources of diradical formation can be utilized in addition to the most commonly employed carbon-halogen cleavages, such as the Bergman cyclization of diethinyl arene species. [172][173][174] In an on-surface process, this ring closure is followed by polymerization. Remarkably, to be practically useful in solution, this ring closure is performed in the presence of a cyclohexadiene as a hydrogen source to avoid coupling between radicals. Currently, the ever-growing number of studies on on-surface syntheses are tackling an increasing number of chemical issues, such as phthalocyanine formation from phthalodinitrile, head-to-tail polymerization starting from nonsymmetric monomers, or copolymerization of different monomers, with many more to come. 172,[175][176][177][178] The above cases, while extremely useful, rely on carbon-carbon bond formation via aryl-aryl coupling. Under the same conditions, if the precursor polymer contains a methyl group in the spatial neighbourhood of another aryl unit, then a kind of CH-bond activation occurs, leading to bond closure between the aryl and methyl groups. This affords ve-membered rings to yield uorene-type structures, such as 77 in Fig. 10 (ref. 179 and 180), or new six-membered rings, as in the case of 80 in Fig. 11. 181 This case deserves special attention because it gives rise to a GNR with zigzag edges. Polymer 80 is unique since, unlike other GNRs, its synthesis has no equivalent in solution synthesis. More importantly, DFT calculations predict extremely low bandgaps and the presence of edge-localized electronic states. This was proven by Roman Fasel et al. via STS. For this purpose, the polymer had to be lied from the metal surface by shiing it with the tip onto a sodium chloride monolayer, thus weakening the electronic interaction between the metal and adsorbate. This interaction is one of the problematic aspects of the metal surface, although it is highly advantageous for the synthesis (see below). While the on-surface synthesis of NGs and GNRs appears to be a fascinating branch of current nanoscience, its scope and limitations deserve further discussion. First, scanning probe methods, preferably STM and nc-AFM due to the accessible atomic resolution, allow one to precisely analyse the resulting structures and follow the mechanistic details of product formation. Under the prevailing UHV conditions, (di)radical intermediates can be made to couple without side reactions involving solvent molecules, which could quench the radical in solution chemistry. Second, apart from the beautiful visualization of the target polymers as part of the proof of the structure, these methods allow detailed electronic characterizations (see below). Third, NGs and GNRs can be synthesized that are not accessible by solution chemistry or that escape further characterization due to the limited solubility and/or stability. Fourth, feeding different  monomers into the UHV chamber provides access to more complex GNR structures, such as by the site-selective incorporation of heteroatoms and the formation of molecularly dened p-n-junctions. 182 However, characteristic conceptual and experimental drawbacks exist. First, the reactions require the assistance of a metal to stabilize either the radical intermediates or the target structures, as well as to catalyse the cyclodehydrogenation to form C-C bonds. The p-systems thus obtained may well owe their stability to the electronic interactions with the metal. This can obscure the intrinsic properties of the adsorbate, such as the bandgaps. Second, the subsequent utilization of thus-formed GNRs in FET devices requires transfer from the metal to insulating substrates. This step is certainly an additional complication, but suitable li-off techniques are available. [183][184][185][186] Even if the visualization becomes more demanding, it might be desirable to perform similar GNR syntheses on insulating surfaces with the use of metal atoms. Third, deposition of the precursor molecules by sublimation in UHV must be possible, which, as already mentioned, imposes restrictions on the molecular size. Alternative deposition techniques are currently being studied, such as electrospray ionization deposition. 156 Fourth, upscaling is certainly an issue in terms of the small amounts of material obtained. One might respond, however, that making a structurally precise monolayer with designed electronic characteristics or with a precise pore size is an extreme form of atom efficiency. Working FET devices can be fabricated by transferring a monolayer lm of GNRs synthesized on a surface. Furthermore, chemical vapour deposition (CVD) allows not only graphene fabrication from carbon precursors such as methane, 187-189 but also GNR synthesis using the same dihalogenated precursors as in the on-surface synthesis described above, even under atmospheric pressure (Fig. 12A). 190,191 The structure of CVDgrown GNR has been characterized by Raman spectroscopy and XPS, the results of which are both in excellent agreement with those obtained by the UHV method ( Fig. 12B and C). As is the case for all new chemical procedures, many challenges remain for mechanistic studies and structural modications. Regarding the former aspect, the coupling of dihaloarene molecules on metal surfaces has been described as an Ullmann-type reaction, which is known in organic chemistry as the biaryl formation of haloarenes upon treatment with copper powder. 192,193 This reaction implies electron transfer, such that the halogen is removed as a halide anion, whereas in the above case, carbon-halogen cleavage occurs homolytically. Bromine or iodine atoms can be detected by Fig. 12 (A) Schematic illustration of GNR synthesis by CVD. (B) Raman and (C) XPS spectra of the UHV-grown and CVD-grown GNRs for comparison. In the inset of (C), the C 1s core-level spectra are presented. 190 Copyright 2016. Adapted with permission from the American Chemical Society. STM but may readily evaporate upon heating. This swi removal does not prevent halogen from reacting with growing chain ends and acting as an endcapper. Further chain growth may also be inhibited by the premature removal of hydrogen from the branched or dendritic precursors forming C-C bonds. 194 Reactions under UHV conditions will thus not mean that no side reaction occurs. Nevertheless, the use of diiodinated monomers was found to reproducibly provide longer polymers compared with their dibrominated counterparts, which could be explained by the lower temperature of the homolytic cleavage of the carbon-halogen bonds, suppressing the undesired release of hydrogen. 195 Another essential feature is connected to the concept of "immobilization". One may expect some stabilization of diradicals by interactions with the metal or by the intermediate formation of covalently bound carbon-metal species. Dimerization or polymerization, however, can only occur if the intermediate radical species can still diffuse on the surface. Here, a brief look at Carothers' equation is instructive; it describes the molecular weight achievable by step-growth polycondensation as a function of the degree of conversion. 196 High molecular weights can only be obtained at high degrees of conversion, that is, when higher oligomers begin to condense. In on-surface synthesis, however, larger oligomers may no longer be able to migrate, so growth is expected to mainly arise from the capture of fresh monomers.
The main motivation for GNR synthesis was to obtain FETs with pronounced on-off ratios, while maintaining the high charge-carrier mobility of graphene. The groups of Jeff Bokor and Roman Fasel could li-off 9-AGNR 71 from the metal aer on-surface synthesis, achieve deposition on a substrate with a hafnium oxide layer as the gate dielectric, and fabricate single-ribbon FETs with a remarkable on-off ratio of 10 5 and a high on-current of >1 mA at a drain voltage of À1 V (Fig. 13). 197 However, the challenge of contact resistance with the metal electrodes hindered demonstration of the expected high mobilities of the GNRs. Close collaboration between chemists, physicists and electrical engineers is essential for further improvement of FET devices using synthesized GNRs. This might be achieved through further optimization of (1) GNR synthesis, in terms of a lower bandgap and a larger length, (2) the GNR transfer process, and (3) device fabrication, including the choice of electrode and gate dielectric materials.

Synthesis and opportunities in physics
Does a chemist target small or polymeric molecules only because the underlying structure could be imagined, and is a successful synthesis sufficient justication for this research? GNRs, as a new generation of CPs, are certainly a challenge for precision polymer synthesis, but they have been fabricated for their extremely promising electronic, optical and even magnetic functions. This is why the above text, while focusing on synthetic concepts, also includes brief excursions into  physics that are hopefully not too supercial. In this section, we focus on future opportunities in materials science, even if a very large gap remains between the initial synthesis and robust technologies. We combine organic and macromolecular, as well as solution and on-surface chemistry. The role of surface physics in achieving the nal step towards realizing zigzag GNR (ZGNR) 80 is not obscured in noting the basic chemical issues. The nature of the periphery can be controlled by the design of the monomeric building blocks, but the monomer must rst be synthesised. Without going into detail, the 12-step synthesis of the monomer 78 needed for the achievement of the ZGNR 80 can be said to be far from trivial PAH chemistry. A closely connected eld of research in which all the above synthetic aspects must be combined is the formation of highspin structures. The creation, manipulation and transport of spins dene special challenges in graphenic structures due to their importance for future spintronics and quantum technologies. [198][199][200][201] Defects in semiconductor materials can give rise to high-spin structures, but the challenge is to form these defects in a controllable fashion. [202][203][204] This so-called defect engineering, a well-established protocol for inorganic materials such as steel and ceramics, has also recently been applied to organic materials. 205 How can one achieve deviations from the regular "honeycomb" lattice of graphene in a rational way? Extending PAHs by the formation of additional rings, as described above, can produce not only six-membered, but also veand seven-membered rings. A remarkable case is compound 85 (Fig. 14A), which has been made available by Xinliang Feng and Junzhi Liu. 206 Five-membered rings are rst formed aer 1,2-phenyl migration upon oxidative treatment of 81; then, seven-membered rings result from Friedel-Cras ring closures. Dehydrogenation of 84 towards the realization of fully unsaturated NG 85 incorporating two azulene units is accompanied by a drastic bathochromic shi of the absorption band of 435 nm. The reason for this shi is the occurrence of biradical structure 86, which can be corroborated by electron paramagnetic resonance (EPR) spectroscopy and superconducting quantum interference device (SQUID) studies ( Fig. 14B and C). A related case is that of diindeno-fused bichrysenyl 87, which gives rise to diradical 88 ( Fig. 14D and E). 207 Pulsed EPR spectroscopy of 88, measured by the group of Lapo Bogani, provides the spin lattice and coherence times as characteristic features of the high-spin structure. 208 The signicance of high coherence times will be reconsidered for spin-bearing GNRs below.
An example of a diradical structure that results from topological frustration rather than from veor seven-membered defects is the case of 90, the so-called Clar goblet with a bowtie shape, which was introduced by Roman Fasel and Xinliang Feng (Fig. 15A). 209 It can be fabricated by on-surface thermal treatment of 89 and was studied by STM and spin excitation spectroscopy (Fig. 15B-E). The essential outcome is a robust antiferromagnetic order with an exchange coupling strength of 23 meV, which is much higher than the Landauer limit of minimum dissipation at room temperature. 210 When discussing edge-localized and spin-bearing structures, ZGNR 80 appears to be an ideal model (Fig. 11A). 181 The trouble is its high reactivity. When exposed to an oxygen-containing atmosphere, it readily undergoes reactions, presumably under the formation of carbonyl functions, among others, and thus escapes detailed studies. A surprisingly simple way of producing spin density on a GNR is to use the bromo-functionalized GNR 92, prepared in solution as mentioned above, and subject it to palladium-catalysed coupling reactions with the gold complex of the nitronyl nitroxide radical (NIT), affording the connection of stable free radicals at the GNR edges of 94 (Fig. 16A). 146 The attached peripheral radicals inject part of their spin density into the p-system of the GNR, which remains chemically stable and becomes susceptible to electron spin resonance (ESR) experiments (Fig. 16B). According to Lapo Bogani, NIT-GNR 94 demonstrates an extremely high spin coherence time of 0.5 ms at room temperature. While this is a fundamental study, a stable phase relation is mandatory before qubits can enter superposition within quantum computing.
Equally important for the future of computing is access to topological insulators that have been obtained in 2D-and 3D-inorganic structures. During syntheses of the zigzag-extended 7-AGNRs 95 and 96, utilizing the cyclization involving methyl groups in the synthesis of ZGNR 80 described above, Oliver Gröning and Roman Fasel theoretically revealed that such zigzag edge extension of AGNRs can create topological insulators in quasi-1D-GNRs (Fig. 17A). 43 The concept is based on the electronic characterization of a cis-polyacetylene by the Su-Schrieffer-Heeger theory, which uses two coupling parameters to describe the interactions across the double and single bonds. 211 This concept is extended to zigzag-extended AGNRs such as 95 and 96 by depicting the intramolecular electronic coupling between the electronic states at the zigzag edges on opposite peripheries. In this theoretical model, GNR 95 was predicted to be a trivial insulator with a reduced bandgap compared to that of the pristine 7-AGNR. STS studies revealed a signicantly lower bandgap of 0.65 AE 0.1 eV for   212 Nevertheless, the fabrication of heterojunctions of GNR 96 and pristine 7-AGNR allowed unambiguous experimental elucidation of the topological end states at the junctions, which fully agreed with the theoretical prediction ( Fig. 17C and D).

Conclusion
NGs continue to present great challenges for synthesis, theory and spectroscopy and will serve as increasingly relevant active components of (opto)electronic devices. Promising examples include applications in photodetection, sensing and organic lasing. Moreover, special importance might also be acquired by chiral NGs in view of the chiral-induced spin-selectivity effect. [213][214][215] This promises control over long-range electron transfer processes with many applications, for example in biorecognition and spintronics. Obtaining enantiomerically pure NGs not only by chromatographic separation but also by enantioselective synthesis would be helpful. Here, synthetic organic chemistry will lead the way. Additionally, NGs have proven to be ideal test cases for GNRsin terms of synthesis and physical characterization. Above all, GNRs will dene further exciting targets for chemistry, including reactions at their surfaces. Edge chlorination and hydrogenation have already been mentioned in the case of NGs, but uorination or the attachment of auxochromic groups is equally challenging. In another direction, we have recently synthesized the 2:1-conjugated molecule 98 consisting of one porphyrin and two NG subunits, which demonstrated electronic absorption up to $1500 nm (Fig. 18). 216 Establishing a zigzag edge at the periphery  of NG precursor 97 for oxidative fusion with the porphyrin edge was mandatory. From a technical viewpoint, rigorous optimization of the cyclodehydrogenation process was necessary to avoid unwanted chlorination. This is a critical issue in many other cyclodehydrogenation processes and requires careful adjustment of the oxidants and their stoichiometry. Much can be gained, however, since the strong bathochromic shi seen for 98 with absorption far into the NIR domain suggests that despite the synthetic demand, researchers should aim for the related polymer. Special opportunities could arise from the incorporation of metals with a high-spin conguration or binding of axial ligands. Many other hybrid structures await realization. Site-selective heteroatom incorporation, which has already been mentioned, would be among the subjects for further study. 151,217 Replacing the carbon centres of a graphenic lattice by boron or nitrogen centres bears a fascinating similarity to the inclusion of main group III or V elements, respectively, in a silicon semiconductor. Unlike CVD techniques (see above), the bottom-up synthesis from monomers immobilized on surfaces holds promise for site-selective doping. In general, GNRs occupy a core position at the crossroads of materials science since they connect the domains of graphene, CPs and NGs. GNRs will play an important role as semiconductors in ensembles and in the quantum-transport regime for single-molecule devices. Additionally, they open avenues towards "synthesizing" exotic quantum states and thus will become instrumental for future computing technologies. The early history of CP research is rich in polymers with poorly dened structures and undisclosed structural defects, and it has taken the joint effort of the dedicated community to advance precision synthesis together with reliable structure-property relationships. The widely studied graphene oxide (GO) has, in a sense, played a similar role: it comprises a dispersible graphenic structure, can be processed, and offers many opportunities for chemical functionalization by covalent or noncovalent binding of partner molecules. This has stimulated many materials-related studies using commercial samples whose identity and reproducible properties oen remain unclear. GO has also served as a precursor for the fabrication of graphene by chemical or thermal reduction, but the quality has been low. GNRs can be obtained by "top-down" or the much more demanding "bottom-up" protocols, and the above text has emphasized the search for unconventional electronic structures and unprecedented functions. What will be the future role of GNRs in the broad context of graphene? While graphene and related 2D materials are nearly omnipresent in the literature, GNRs, by comparison, have attracted much less attention. This is partly due to their limited availability since there is still a high chemical barrier. It falls to synthetic chemists not only to devise new complex structures but also to remove this blockade.

Conflicts of interest
There are no conicts to declare.