Introduction

The promise of synthetic carbon allotropes (i.e., allotropes or derivatives not naturally occurring) lies in their lightweight, environmentally friendly, and electronically diverse properties1,2. As an example, graphene is a two-dimensional monolayer of graphite that consists of hexagon lattices of sp2-hybridized carbon atoms. Graphene and its derivatives have been termed “wonder materials” as a result of unique optical and electrical properties, many of which derived from the fact that it is a conductor, i.e., it lacks a band gap3. On the other hand, interest in graphene nanoribbons (GNRs) has recently exploded, in order to bridge the properties between 1D allotropes and 2D graphene. GNRs share many of the desirable properties of bulk 2D graphene, while quantum confinement in one dimension offers a defined band gap necessary for semi-conductor applications and devices4,5,6.

Graphynes are a class of synthetic carbon allotropes that are obtained by interspersing sp-hybridized carbon atoms into the sp2-carbon framework of graphene. Recent calculations have predicted that graphynes could be superior to graphene for electronic and optoelectronic applications7,8,9,10,11,12,13,14,15,16. Four graphyne derivatives, in particular, have attracted attention, namely α-, β-, γ-, and 6,6,12-graphyne, which differ in structure as well as the percentage of sp-hybridized carbon content versus sp2-carbon in their conjugated structure. For example, α-graphyne contains 75% sp-carbons, which is reduced to 55.56% for 6,6,12-graphyne. Theoretically, graphynes are predicted to exhibit Dirac cones, which means that the valence and conduction bands are shaped as upper and lower conical surfaces that meet at the so-called Dirac point (corresponding to a zero band gap)9. Experimentally, there are very few synthetic attempts reported for either α- or β-graphyne, although Li et al.17 recently found that a small fragment of α-graphyne, a so-called carbo-benzene18, exhibits a very high single-molecule conductance, and Maraval, Chauvin et al.19 have recently reported an elegant synthesis of carbo-naphthalene, a polycyclic carbo-benzenoid fragment of α-graphyne. Synthetically, most efforts have focused on γ-graphyne fragments comprised of tribenzocyclyne units, pioneered by Haley20,21 and Tobe22,23. From time-resolved microwave conductivity measurements, it was concluded that large γ-graphyne fragments could be applied as organic semiconductors23. Of the various graphynes that have been commonly considered to date, however, one stands out as unique, namely 6,6,12-graphyne (Fig. 1a). This material does not share the hexagonal symmetry common to graphene and other graphynes. Rather, 6,6,12-graphyne features rectangular symmetry, and, according to theoretical studies, it should feature two self-doped nonequivalent Dirac cones, i.e., electrons are present as charge carriers in one Dirac cone, while holes are present in the other8. For these reasons, direction-dependent electronic conduction is expected for this allotrope of graphyne. In addition, the carrier mobility of 6,6,12-graphyne is predicted to be even larger than in graphene10. Finally, segments of 6,6,12-graphyne, graphyne nanoribbons (GyNRs), are predicted to exhibit unique transport properties for electronic devices24.

Fig. 1
figure 1

Structure of 6,6,12-graphyne. a Schematic representation of 6,6,12-graphyne with fragments highlighted in colors. b Radiaannulene oligomer model systems of Type 1 and Type 2. tBu = tert-butyl, TIPS = triisopropylsilyl

Full size image

Despite the remarkable properties predicted by theoretical calculations, the synthesis of either 6,6,12-graphyne or significant segments of its structure (exhibiting quantum confinement) remains elusive. For this reason, we initiated synthetic attempts toward segments of this allotrope. Examining the structure of 6,6,12-graphyne, one of the core repeat units can be described as a hybrid between a dehydrobenzannulene and an expanded radialene, a structure known generally as a radiaannulene25. Based on this motif, we identify two series of oligomers to help explore the potential properties of 6,6,12-graphyne (Fig. 1b). Type 1 radiaanulene oligomers feature edges that are closed by a butadiyne segment (i.e., the exocyclic C = C is removed at each end of the oligomer, representing a potentially stable edge structure), and Type 2 radiaanulene oligomers contain edges that are open as functionalized tetraethynylethene (TEE) units (representing the basic repeat units). Herein, we present the synthesis and properties of oligomers of Type 1 and Type 2. In addition, macrocyclic subunits (a bis-dehydrobenzannulene—highlighted in green in Fig. 1a—and radiaannulenes) are identified and synthesized in order to support comparisons as a function of molecular size.

Results

Synthesis

The synthesis of bis-dehydroannulene 1 and radiaannulenes of Type 1 (compounds 2 and 3) is shown in Fig. 2, while the synthesis of radiaannulenes of Type 2 (compounds 4a–b, 5–6) is presented in Fig. 3. In general, these syntheses relied on an iterative series of Pd-catalyzed Sonogashira coupling reactions between terminal alkynes and a suitable vinyl- or arylhalide (Br, I) electrophile. Furthermore, our strategy exploits the ability to selectively remove trimethylsilyl (TMS) and triethylsilyl (TES) alkyne-protecting groups before the triisopropylsilyl (TIPS) group.

Fig. 2
figure 2

Stepwise synthesis of 6,6,12-graphyne fragment 1 and Type 1 radiaannulenes 2 and 3. Reagents and conditions: a (i) K2CO3, THF/MeOH, (ii) 8a, Cat. I, Et3N/THF, ))), 40 °C. b 1,2-Diiodobenzene, K2CO3, Cat. I, THF/Et3N/MeOH, ))), 40 °C. c (i) TBAF, THF/H2O, (ii) Cat. I, THF/Et3N. d TMS-acetylene, Cat. I, THF/Et3N, ))), 35 °C. e (i) K2CO3, THF/MeOH, (ii) Cat II, O2, 4 Å MS, CH2Cl2. f (i) TBAF, THF/H2O, (ii) Cat II, O2, 4 Å MS, CH2Cl2. g (i) K2CO3, THF/MeOH, (ii) 11, Cat. I, THF/Et3N/MeOH, MW, 75 °C. h ClAuPPh3, NaOMe, THF/MeOH. (i) 11, Cat. I, THF/Et3N/MeOH, MW, 75 °C. j (i) TBAF, THF/H2O, (ii) Cat II, O2, 4 Å MS, CH2Cl2. Cat. I = Pd(PPh3)2Cl2/CuI, Cat. II = CuCl, TMEDA = N,N,N’,N’-tetramethylethylenediamine, tBu = tert-butyl, Ph = phenyl, TMS = trimethylsilyl, TIPS = triisopropylsilyl, TBAF = tetrabutylammonium fluoride, MS = molecular sieves, MW = microwave, ))) = sonication (ultrasound)

Full size image
Fig. 3
figure 3

Stepwise synthesis of 6,6,12-graphyne fragments 4a, b and Type 2 radiaannulenes 5 and 6. Reagents and conditions: a R = TES: TIPS-acetylene, Cat. III, Et3N, 50 °C. b R = TIPS: phenylacetylene, Cat. III, THF/(iPr)2NH, 60 °C. c R = TES, R’ = TIPS: (i) tBuOK, THF/MeOH, 40 °C, (ii) 8b, Cat. III, THF/(iPr)2NH, 60 °C. d R = TIPS, R’ = Ph: (i) TBAF, THF, 0 °C, (ii) 8c, Cat. III, PhMe/(iPr)2NH, 50 °C. e R = TES, R’ = TIPS: (i) KOH, THF/MeOH, (ii) 15a, Cat. III, Et3N, 50 °C. f (i) KOH, THF/MeOH, (ii) 8b, Cat. III, Et3N, 50 °C. Cat. III = Pd(PPh3)4/CuI, TBAF = tetrabutylammonium fluoride, Ph = phenyl, TES = triethylsilyl, TIPS = triisopropylsilyl

Full size image

The first target, bis-dehydroannulene 1, was synthesized starting from the silyl-protected diethynylbenzene derivative 7, which was selectively desilylated and treated with the vinyl dibromide 8a26,27 under Sonogashira coupling conditions, to provide product 9. This compound (9) was then elaborated to 11 via desilylation followed by a Sonogashira coupling with 1,2-diiodobenzene. With 11 in hand, a subsequent sequence of desilylation followed by intramolecular Sonogashira coupling gave annulene 1. Compound 11 represented a common precursor for the synthesis of GyNR 2. Cross-coupling of 11 with TMS-acetylene gave 12, and desilylation followed by intramolecular Hay homocoupling gave 13. A second sequence of desilylation and homocoupling then gave the dimer 2. Noteworthy, the oxidative Hay couplings to give the butadiyne edges of the target molecules 2 and 13 were accomplished in the presence of molecular sieves (4 Å) to remove water as previously described to be beneficial for such couplings28.

Compound 14, the precursor for the trimer 3, was first prepared from 9 via desilylation and subsequent intermolecular Sonogashira coupling with 11, but yields were low (Fig. 2; conditions g). Alternatively, triphenylphosphine Au(I) end-capped intermediate 10 could be formed from 9 and coupled with 11 under Sonogashira-like conditions. In spite of previous reports that suggested Au(I) oligoynyl complexes can be excellent substrates in cases in which the unprotected terminal alkynes exhibit limited stability29, an improved yield of 14 was not realized by this protocol (Fig. 2; conditions h). In a final step, trimer 3 was completed using the standard desilylation and homocoupling sequence, representing a significant segment of 6,6,12-graphyne (the magenta-colored segment in Fig. 1).

Several approaches were explored for assembly of Type 2 radiaanulene oligomers. Initially, a one-pot approach was tested, which had previously been used effectively for the synthesis of radiaannulenes30. While successful for the assembly of, for example, 4a and 5, these tactics consistently gave low yields (see Supplementary Methods, p. 24–28, for details) and were eventually abandoned in favor of a stepwise approach. The advantage of the revised method derived from a common building block to be used in the synthesis of each oligomer, namely enediyne 15a. Sonogashira coupling of 15a with TIPS-acetylene gave 16a, which could be converted to radiaannulene monomer 4a via desilylation and Sonogashira coupling with dibromoolefin 8b31. An analogous sequence of reactions, starting from 15b and phenylacetylene, gave 16b, which underwent ring closure using dibromoolefin 8c to give radiaannulene 4b. Alternatively, starting with 16a, a sequence of desilylation and cross-coupling with 15a gave the expanded system 17, which was then readily converted into dimer 5 via desilylation and Sonogashira coupling with 8b. Finally, the entire expansion process was repeated, starting with 17, to give 18 and then 6, completing the blue-colored segment of 6,6,12-graphyne highlighted in Fig. 1. The structural assignment of 6 rests on high-resolution mass spectrometry and UV–vis spectroscopy from a small amount (μgs) of sample purified by preparative thin-layer chromatography, and the limited amount of sample precluded further characterization. The majority of the product isolated from the reaction mixture, calculated as 5% yield (3.5 mg) based on the assumption of a pure compound, decomposed upon exposure to CDCl3 during sample preparation for NMR analyses (see Supplementary Methods, p. 20, for full synthetic details).

The isolated Type 1 and Type 2 molecules showed reasonable solubility in common organic solvents as a result of the t-butyl and triisopropylsilyl groups incorporated at the periphery of the molecules. All compounds exhibited good stability, except for the trimer 6 that appeared to rapidly (and surprisingly) decompose when dissolved in CDCl3, which ultimately limited characterization. When macroscopic quantities were available for analyses, differential scanning calorimetry confirmed excellent thermal stability with decomposition temperatures often above 250 °C.

X-ray crystallography

An interesting structural aspect of the Type 1 radiaannulenes is the ring strain introduced through the closing of the butadiyne segment. Unfortunately, the presence of tert-butyl substituents, introduced to maintain solubility, also appear to thwart the formation of X-ray quality single crystals, with only one exception. An intermediate radiaannulene was successfully analyzed by X-ray crystallography (see Supplementary Fig. 77 for details), showing that the interior angles of the butadiyne segment are only moderately strained with acetylenic bond angles from 177 to 167°.

Single crystals of the series of Type 2 GyNRs were obtained, including radiaannulenes 4a and 4b, as well as dimer 5 (Fig. 4a–c). A comparison of bond lengths in the series of structures shows few discernible trends. Overall, the endocyclic alkyne bonds of all three structures are close to linear, with the only notable of exception of angle C(14)–C(3)–C(4) 171.3(2)° (5). The exocyclic alkylidene angles in 4a and 5 (C(31)–C(2)–C(41) and C(5)–C(15)–C(7), respectively), at 116.0 and 115.7°, are distorted from the ideal of 120°. The corresponding endocyclic alkylidene bonds of 4a and 5 are 117–118°, the analogous value found for both the exo- and endocyclic alkylidene angles of arylated 4b (117.4(3) and 118.2(3), respectively). In total, these observations suggest that structure of large GyNRs might show slight distortions from ideal bond angle values at the alkylidene centers, balanced by slight bending of the alkyne units.

Fig. 4
figure 4

Molecular structures obtained by X-ray crystallography. ORTEP structure (top) and solid-state packing (bottom) of (a) radiaannulene 4a, (b) radiaannulene 4b, and (c) radiaannulene dimer 5; selected bond lengths [Å] and angles [°] 4a: C(1)–C(2) 1.378(2), C(11)–C(12) 1.201(3), C(21)–C(22) 1.199(3), C(31)–C(32) 1.206(3), C(41)–C(42) 1.202(3); C(31)–C(2)–C(41) 116.02(15), C(11)–C(1)–C(21) 117.45(15), C(1)–C(11)–C(12) 174.6(2), C(11)–C(12)–C(13) 177.8(2), C(1)–C(21)–C(22) 178.6(2), C(21)–C(22)–C(23) 176.56(19); 4b: C(5)–C(6) 1.374(5), C(7)–C(8) 1.188(4), C(9)–C(10) 1.197(4), C(1)–C(2) 1.182(4), C(3)–C(4) 1.181(4); C(3)–C(2)–C(1) 117.4(3), C(7)–C(6)–C(9) 118.2(3), C(6)–C(7)–C(8) 178.9(4), C(7)–C(8)–C(31) 177.8(3), C(6)–C(9)–C(10) 179.3(4), C(9)–C(10)–C(32) 179.4(4); 5: C(13)–C(13’) 1.384(3), C(1)–C(2) 1.198(3), C(11)–C(12) 1.202(3), C(14)–C(15) 1.379(3), C(3)–C(4) 1.172(3), C(9)–C(10) 1.193(3), C(5)–C(6) 1.210(3), C(7)–C(8) 1.211(3); C(5)–C(15)–C(7) 115.74(16), C(13)–C(1)–C(2) 179.30(19), C(1)–C(2)–C(21) 175.8(2), C(13)–C(12)–C(11) 176.52(18), C(12)–C(11)–C(36) 177.9(2), C(1)–C(13)–C(12) 117.91(15), C(14)–C(3)–C(4) 171.3(2), C(3)–C(4)–C(22) 177.5(2), C(14)–C(9)–C(10) 178.8(2), C(9)–C(10)–C(31) 174.2(2), C(3)–C(14)–C(9) 117.59(15); ORTEP ellipsoids at the 30% probability level, hydrogen atoms omitted for clarity. CCDCs 1894937 (4a), 1894939 (4b), 1894938 (5)

Full size image

The benzo-fused cyclic cores of all three molecules, 4a–b and 5 demonstrate a nearly planar structure, in spite of small deviations from ideal bond angles (Fig. 4a–c). The molecules pack in a slip-stacked fashion (see Supplementary Figs. 78–80), and intermolecular packing is dramatically affected by molecular size and terminal substitution. With intermolecular separation of 3.60 Å, radiaannulene 4a shows the largest interplanar distance, presumably due to the large ratio of sterically demanding TIPS groups relative to the conjugated carbon structure. Increasing the size of the molecular skeleton decreases the interplanar distance to 3.38 Å in 5. Finally, formal replacement of the TIPS groups with phenyl rings, as in 4b, approximates the most realistic packing for 6,6,12-graphyne with intermolecular stacking of 3.27 Å, which is slightly less to that found in graphite (3.35 Å)32.

Electronic absorption properties

The electronic absorption spectra of Type 1 and Type 2 molecules are shown in Fig. 5a (absorption maxima are presented in Supplementary Table 1). The absorption characteristics of bis-dehydrobenzannulene 1 and Type 1 molecules 2–3 are analyzed on the basis of both the longest-wavelength absorption maximum (λmax) and the HOMO-LUMO gap (i.e., λonset, which is estimated as the intercept of the tangent line to the lowest-energy absorption with the x-axis). Both λmax and λonset show sequentially red-shifted values: λmax, λonset = 480 nm, 496 nm for 2; 490 nm, 521 nm for 1; 530 nm, 552 nm for 3 (λonset = 2.50, 2.38, and 2.25 eV, respectively, for 1–3). The bis-dehydrobenzannulene 1 and dimer 2 have similar HOMO-LUMO gaps, suggesting that structural expansion in either the vertical or horizontal direction (based on Fig. 1) has approximately the same effect, if extrapolated to small GyNRs; λmax values of 1 and 2 are also similar to that of radiaannulene 4b (λmax = 483 nm). We note that neither 1 nor 2 exhibit any apparent fluorescence. The effect of the conjugation pathway is evident in the comparison of 1 and 2. Dimer 1 (linearly conjugated) has four fewer sp-hybridized carbon atoms than dimer 2 (cross-conjugated), and yet λmax-values are comparable, while that of 1 actually appears at lower energy. DFT calculations reveal that the HOMO and LUMO of 1 (with tBu groups removed for computational convenience) span the entire π-system. On the other hand, while the LUMO encompasses the complete π-system of dimer 2, the HOMO does not include the butadiyne bridges (see Supplementary Fig. 85). Trimer 3 has the lowest HOMO-LUMO gap in this series, and, thus, expansion from 2 to 3 leads to a reduction of the HOMO-LUMO gap by 0.25 eV. Based on these three molecules, it is projected that the effective conjugation length, i.e., saturation of the electronic absorption trend, has not yet been reached for Type 1 radiaannnulene oligomers with butadiyne edges.

Fig. 5
figure 5

Properties of the compounds. a Electronic absorption spectra of 1 (green line), 2 (orange line), and 3 (pink line) in CH2Cl2 and of 4a (black line), 5 (red line), and 6 (blue line) in THF. The spectra were normalized based on the longest-wavelength absorption maximum (note, the broadening and tailing of the low energy absorption of 6 could arise from impurities, see the text for additional discussion). b Cyclic voltammograms of 1 (green line), 2 (orange line), and 3 (pink line) in CH2Cl2 (+0.1 M Bu4NPF6); scan rate 0.1 V s−1. Fc = ferrocene. c Nucleus independent chemical shift, NICS(1) πzz, values calculated for different charge states (q) at 1 Å above the center of the rings marked by a solid circle and a star (B3LYP/cc-pvdz; tBu and TIPS groups have been replaced by hydrogen atoms). d Canonical forms of the dianions of trimers 3, 6, and of an infinite radiaannulene graphyne nanoribbon (GyNR). tBu = tert-butyl, TIPS = triisopropylsilyl

Full size image

The lowest-energy absorption energy λmax and HOMO-LUMO gap (λonset) also decrease in the Type 2 series versus oligomer length: from monomer 4a (λmax = 450 nm, λonset = 466 nm), to dimer 5 (λmax = 504 nm, λonset = 532 nm), to trimer 6 (λmax = 522 nm, λonset = 570 nm (2.66, 2.33, 2.18 eV, respectively, for 4a, 5–6). While the spectral characteristics of 6 fit a trend established by 4a and 5, it should be noted that the broadening and tailing of the low-energy absorption could result from the presence of impurities due to the limited amount of sample (vide supra). Notably, the shift in HOMO-LUMO gap upon extension from 5 to 6 is relatively small (0.15 eV, based on λonset), which indicates that at the stage of the trimer, the effective conjugation length for this series of 6,6,12-graphyne model compounds has been nearly reached. This premise is further supported by a comparison of HOMO-LUMO gaps of Type 1 trimer 3 and Type 2 trimer 6, which show similar values for λonset (2.25 eV and 2.18 eV, respectively).

Redox properties

The potential of 6,6,12-graphyne and of GyNRs to function as useful semiconductors is expected to correlate with their redox potentials. Electrochemical analysis of selected molecules and reference compounds by cyclic voltammetry (CV) was therefore conducted, including the series of 1–3 (Table 1 and Fig. 5b). Upon reduction of 1–3, a well-defined one-electron transfer takes place leading to the reversible formation of the radical anion; the peak heights are of similar size for the compounds. Upon further reduction, the dianion is formed, which for 2 and 3 is stable within the experimental timeframe (while a follow-up reaction occurs for the dianion of 1, probably protonation with residual water, forming products that are reduced at a similar potential as where the radical anion is reduced to the dianion). Compounds 2 and 3 exhibit the smallest reduction potentials, with trimer 3 being 40 mV easier to reduce than dimer 2, which again is 110 mV easier to reduce than the bis-dehydrobenzannulene 1. A similar influence of macrocyclic oligomerization is observed when proceeding from Type 2 radiaannulene 4a (−1.50 V vs. Fc/Fc+) to dimer 5 (−1.33 V vs. Fc/Fc+), and from radiaannulene 17 (−1.60 V vs. Fc/Fc+) to dimer 18 (−1.46 V vs. Fc/Fc+), i.e., monomer-to-dimer differences of 170 mV and 140 mV, respectively (voltammograms were measured under slightly different conditions, see Supplementary Fig. 83).

Table 1 Electrochemical data—reductionsa
Full size table

It is clear that the size of the π-system plays a role for the acceptor strength. Thus, the first reduction potential of 2 and 3 differs by only 40 mV, while the second reduction differs by as much as 250 mV; this is reasonable based on electrostatic repulsion, in which the two negative charges of the dianion can be further separated in 3 than in the smaller 2. Moreover, the facile reduction of the radiaannulene oligomers is attributed to the inherent acceptor strength of the TEE unit, which has been documented in the literature33,34. Notably, the radiaannulenes are significantly stronger acceptors than tribenzocyclyne35 (−2.21 V vs. Fc/Fc+) that can be seen as a macrocyclic fragment of γ-graphyne.

Oxidations of Type 1 molecules occur irreversibly, and the data are therefore more difficult to interpret. Oxidations occurred for 1–3 at peak potentials of 1.13 V, 1.10 V, and 0.82 V (vs. Fc/Fc+), respectively. For trimer 3, this would correspond to an electrochemical HOMO-LUMO gap of 2.15 eV, which is slightly lower than the optical gap of 2.25 eV (based on λonset).

In summary, the electrochemical studies show that radiaannulene oligomers of both Type 1 and Type 2 are excellent electron acceptors, and we predict that 6,6,12-graphyne would stand out as a particularly good electron acceptor in comparison with other graphyne allotropes.

Nucleus independent chemical shift (NICS) calculations

To explore the remarkable ability of the molecules to accommodate electrons, a calculational study was explored using NICS36. According to Hückel’s rule of aromaticity, the radiaannulene cores are 14 πz-aromatic upon reduction. NICS(1) πzz values are listed in Fig. 5c for the various charge states, in which NICS(x) designates an evaluation of aromaticity at a distance of x Å above the molecule. Upon reduction of trimers 3 and 6 to their dianions, the NICS values transform from positive to negative, indicating gain in aromaticity of all three rings. As shown in Fig. 5d, canonical forms with three 14 πz-aromatic rings can be drawn for these dianions37. In fact, for an infinite oligomer, a two-electron reduction could afford an entirely aromatized structure, as illustrated in Fig. 5d. Extrapolating from this observation suggests that a neutral GyNR would also begin to build up diradical character at the edges, stabilized by the resultant aromatic character imparted to the core. This is analogous to observed behavior of GNRs38. It should be noted, however, that this “aromatization” disturbs the benzenoid character of the system, and resonance forms with intact benzene rings (Clar sextets) are expected to contribute as well to the resonance hybrid. Interestingly, the NICS calculations reveal that, upon two-electron reduction, the rings of 3 gain more aromaticity than the rings of 6. The efficient π-electron delocalization skeleton of 3 upon reduction is confirmed by geometry calculations, which show a distinct decrease in bond length alternation through shortening of single bonds and lengthening of double and triple bonds, as schematically suggested by the resonance form shown in Fig. 5d (see Supplementary Fig. 84 and Supplementary Table 8).

Calculations also allow an estimation of molecular size for the radiaanulene oligomers. Trimer 3 has calculated length of 1.7 nm (distance calculated from center of the two butadiyne edges), and for trimer 6 the length is estimated as 2.6 nm (the distance calculated between distal carbon atoms of the terminal alkyne units bearing the TIPS groups).

Discussion

In conclusion, radiaannulene oligomers that represent significant sections of 6,6,12-graphyne have been synthesized by iterative acetylenic cross-coupling reactions and studied in either solution or the solid state. Only the Type 2 trimer shows limited stability in solution (chloroform), while the Type 1 trimer was stable and shows particularly easy reduction to mono- and dianions. When proceeding from neutral trimer to its dianion (3 and 6), the generation of three aromatic rings is suggested from NICS calculations. For an oligomer composed of n radiaannulene rings, the same number n of aromatic octadehydro[14]annulene rings could be generated upon two-electron reduction. In fact, we speculate that a long GyNR may be described also in the neutral state as a sequence of 14π-aromatic rings, with an unpaired electron at each of the edges (diradical). X-ray crystallography of Type 2 mono- and dimers predict a planar framework for 6,6,12-graphyne, as would be expected, in spite of the potential of small deviations from ideal interatomic angles, most notably the alkylidene enediyne angles which appear to tend toward 118°. Furthermore, the interplanar packing for 6,6,12-graphyne (≤3.27 Å) is predicted to be slightly less than that of graphene (3.35 Å).

From systematic optical studies and λonset values, we predict the HOMO-LUMO gap of an infinitely long GyNR to be not much smaller than 2.2 eV. This experimentally predicted HOMO-LUMO gap is significantly larger than that predicted from theoretical models of 6,6,12-graphyne, for which two Dirac points are expected (i.e., no band gap). A comparison of the two types of oligomers reveals clear edge effects in the electronic properties, but in an infinite polymer, such effects are expected to be eliminated.

The synthetic protocols that have been developed in this study would allow for vertical extension to longer GyNRs. It seems more relevant, however, to target expansion in the horizontal direction. UV–vis absorption studies show a lower HOMO-LUMO gap for bis-dehydrobenzannulene 1 than for the Type 1 dimer 2, despite the larger π-system of the latter. Thus, fusing radiaannulene oligomers along the lateral edges would introduce the bis-dehydrobenzannulene fragment (Fig. 1a), leading to increased electronic delocalization as expected for extended GyNRs. Concurrently, the question remains, how many repeat units are required in the horizontal direction to make a GyNR analogue of 6,6,12-graphyne? The answer to this is not yet known and will clearly challenge current synthetic methods to introduce the additional functionality needed to allow expansion of the GyNR in two dimensions. Indeed, the complicated, rectangular symmetry of 6,6,12-graphyne calls for a classical synthetic approach rather than the otherwise convenient on-surface polymerization of suitable molecular precursors that has attracted attention in recent years for synthesis of GNRs4,5,6 and recently also for 30,30,30-graphyne39.

Methods

Synthesis, optical and electrochemical studies

The syntheses of all target molecules and precursors are provided in Supplementary Methods, which also gives details on the optical and electrochemical studies.

Computations

Geometry optimizations on neutral molecules, radial anions, and dianions were performed on structures where the tBu and TIPS groups had been substituted for hydrogen atoms at the B3LYP/cc-pVDZ level of theory (an optimization of 14 with TIPS groups, but without tBu groups, was also performed). Frontier molecular orbitals are shown in Supplementary Fig. 85, and were obtained at the B3LYP//cc-pVDZ//CAM-B3LYP/6-31 g(d, p) level of theory.