Cycloparaazine, a full-azine carbon nanoring

Abstract

Cycloparaphenylenes (CPPs) and related carbon nanorings (CNRs) represent iconic molecular entities in molecular nanocarbon science. While theoretical studies predict that the introduction of nitrogen atoms (N-doping) onto CPP frameworks would add a number of fascinating properties, only a handful of partially N-doped carbon nanorings have been synthesized. We herein report the synthesis of a long-awaited cycloparaazine (CPA), where every para-connected aromatic moiety consists of a N-heterocycle, and two other highly N-doped CNRs. The evaluation of optoelectronic and structural properties coupled with theoretical studies uncovered the impact of both the amount and positioning of N-doping onto the nanorings properties; far less ring strain, red-shifted UV–vis absorption and fluorescence, smaller HOMO–LUMO gaps and both higher reduction and oxidation potentials than pristine CPPs. Ultimately, new potential applications of highly N-doped nanorings were examined in non-covalent supramolecular property engineering with Lewis acids and as energy storage materials.

Theoretical studies predict that the introduction of nitrogen atoms onto cycloparaphenylene frameworks would add fascinating properties but few partially N-doped carbon nanorings have been synthesized. Here, the authors report the synthesis of a cycloparaazine, where every para-connected aromatic moiety consists of a N-heterocycle, and two other highly N-doped carbon nanorings.

Introduction

Since their discovery in 1991¹, carbon nanotubes (CNTs) have found wide applications in various fields of material science such as in organic electronics, polymers or energy storage devices, etc^2–6. Along these lines, N-doped structural congeners (NCNTs) have also been successfully prepared⁷ and studied^8–10. However, accessing them by either post-treatment or in situ-synthesis only give rise to randomly N-doped structures. It is known that electrical, magnetic, mechanical and optical properties of CNTs and NCNTs heavily depend on their structure, the alignment and the diameter of the tubes^11,12. Still, the structurally uniform control of the tubes key parameters during their construction remains an unsolved problem and a straightforward way of separating CNTs has also not been developed to date (Fig. 1a)¹³.

Fig. 1. Nitrogen-containing carbon nanotubes and nanorings.

a State-of-the-art of nitrogen-incorporation in carbon nanotubes and rings. b Synthesized N-doped carbon nanorings and the use of their nitrogen lone pairs. c Our approach of synthesizing a cycloparaazine.

Cycloparaphenylenes (CPPs) are considered as simplest sidewall segments of an armchair CNT¹⁴ and they possess unique radially oriented cyclic π-systems, which cause outstanding optical, electronical, self-assembly and charge transport properties (Fig. 1a)^15–19. Despite the synthetic interest in this molecule class since the 1930s²⁰, first preparative accesses were only achieved in the late 2000s by three groups independently in different ways (Jasti/Bertozzi²¹, Itami²² and Yamago²³). In contrast to CNTs, CPPs are atom-precisely defined wherefore their structure dependent properties can be synthetically altered straightforwardly via changing their architecture¹⁶. Over the past 16 years, diverse derivatives of this molecule class were disclosed²⁴, including different sized rings ([n]CPP; n = 5–16, 18, 20, 21)^{14,23,25–32}, donor- and/or acceptor-substituted congeners¹⁵, polyaromatic^33–36 and heteroaromatic^37–41 containing cycles, as well as chiral^33,42,43, antiaromatic⁴³, bridged^44,45, branched^46–49, connected^50–52 or interlocked carbon nanorings^53,54. These findings led to the examination of CPPs and derivatives in electronic applications^55,56, as (biological) fluorophores^57,58 or as solid state emitters^59,60. Additionally, CPPs were envisioned as template structures in the size- and structure-selective bottom-up synthesis of CNTs⁶¹.

Despite these achievements, the selective incorporation of heteroatoms and especially nitrogen atoms have only been sparsely investigated¹⁵. In 2010, Bachrach and Stück proposed in a theoretical study that N-containing nanorings should show highly interesting properties (Fig. 1b). For example, the ring strain energy should be significantly lower, as compared to their all-carbon congeners for steric reasons⁶². Two years later, we reported the synthesis of the N-containing carbon nanoring (CNR) 1. The nitrogen lone pair of the bipyridine entity enabled both protonation, which alters the photophysical properties by a bathochromic effect, qualifying the respective derivative as pH-responsive probe, as well as Pd-ligand⁶³. Jasti and co-workers later synthesized other pyridine-containing CNR-congeners 2. It was found that moderate N-doping alters the carbon nanorings properties only slightly. However, N-methylation generates donor-acceptor nanoring 3 with the phenylene backbone as donor and the pyridinium salt as acceptor moiety, causing both a bathochromic shift in absorption and fluorescence, as well as an increase in reduction potential^38,64. Additionally, the same group reported a bipyridine-doped [8]CPP, which could be applied to construct the dimeric metal-bridged structure 4 and the CNR-containing Ru(bpy)₃²⁺ derivative 5 (Fig. 1b)⁶⁵. Furthermore, Quinton, Poriel and co-workers investigated the structural and electronic properties of CNRs in dependence of varying azines (pyridine, pyrimidine) in the rings peripheries⁶⁶. Recently, the Jasti group also facilitated the coordinative ability of pyridinic nitrogen lone pairs in an active template strategy towards N-doped CNR-catenanes⁶⁷. During the review-process of this manuscript, the preparative access to a cone-shaped all-azine carbon nanobelt was reported by Tong, Wang and co-workers⁶⁸. However, the true cycloparaazines (CPAs)⁶², where every para-connected aromatic moiety consists of a N-heterocycle, remains an unachieved goal today (Fig. 1a).

Herein we report the preparative access to a CPA 6 through a gold-mediated macrocyclization of the bisboronic ester-functionalized N-heterocyclic linker 7, that is subsequently para-connected throughout reductive elimination from the Au-complex 8 (Fig. 1c). Additionally, two other highly N-doped CNRs could be prepared applying the same approach and all three rings were examined regarding their optical, electronical, structural and theoretical properties. Comparison of those properties with the all-carbon [9]CPP sheds light on the impact of both the amount and positioning of the N-doping onto the ring’s properties. Ultimately, new potential applications were outlined by examining these NCNRs in supramolecular complexation with Lewis acids and as charge storage materials.

Results

Retrosynthetic analysis

We first evaluated whether established strategies to access CPPs¹⁵ can also be applied for the synthesis of CPAs. Generally, those can be subdivided in two main strategies (Fig. 2a): One strategy relies on the application of proaromatic L-shaped building blocks, which are first connected to the less-strained macrocycles 9 or 10, that are afterwards aromatized to strained CNRs^21,22. Necessarily, the L-shaped units are converted into phenylenes that are incorporated into the nanocarbon framework (Incorporation strategy, Fig. 2a red), which disqualifies these protocols for the construction of CPAs. The other strategy relies on the formation of metalorganic macrocycles 11–13 as intermediates that undergo reductive C–C bond formation to the desired compounds^23,69,70. Therein, linear aromatic bismetalated (SnR₃, B(OR)₂ or Li) precursors undergo transmetalation with preformed metal complexes (Pt(COD), Au(dcpm), Ni(dnbpy)) whose chelating ligands induce the formation of macrocyclic structures 11–13. Thus, the metal complexes inducing the cyclic nature are formally traceless (Traceless strategy, Fig. 2a blue) rendering this class of reactions eligible for the general synthesis of CPAs.

Fig. 2. Synthesis of CPPs and CPAs.

a State-of-the-art of CPP syntheses. b Retrosynthetic considerations for CPA synthesis.

Nitrogen lone pairs are prone to coordinate metals. However, to obtain the desired metalorganic macrocycles 11–13, the respective ligands (COD, dnbpy, dcpm) are inevitably necessary for the transformation. Therefore, metal ligation to the N-atoms of CPA building blocks must be suppressed. The Yamago method²³ using cyclooctadiene (COD) and our method⁶⁹ that operates with bipyridine (dnbpy) ligands will not be compatible, due to interference of the coordination abilities through a pyridine substrate (see Supplementary Information (SI) 3.21 for details). However, phosphine-gold coordination as applied in the Tsuchido-Osakada approach⁷⁰ is generally strong, rendering this strategy promising to access CPAs. Following this approach, stereoisomerically symmetric bisboronic precursors needed to be synthesized (see SI chapter 3 for details) for the self-assembly with bis-Au-dcpm complex 14 in order to avoid regioselectivity problems. Since boronates at the 2-pyridine position are unstable under aqueous conditions^71,72, a suitable starting material should contain the boronic moiety at the pyridine 3-position. Thus, the CPA-forming reaction would then represent a 3,3’-bipyridine bond formation (Fig. 2b).

Synthetic investigations

The Tsuchido-Osakada method⁷⁰ that is derived from earlier work of the Toste group⁷³ has not yet been used for pyridine-couplings. Pleasingly, we found that the bis-gold-bisphosphine complex 14 reacted with 3-pyridine boronic acid (15) to give the desired complex 16 in 56% yield⁷⁴. Noteworthy, the analogous reaction with a 2-pyridine boronate failed (see SI 3.3.). Oxidation of 16 with PhICl₂ (17) in CDCl₃ afforded 3,3’-bipyridine (18) in 81% yield (see SI 3.4 for details). This model reaction clearly revealed that the P-Au complexation is not interfered through monodentate pyridine complexation and the 3,3’-bipyridinic bond formation is generally feasible (Fig. 3a).

Fig. 3. Synthetic efforts towards CPA and other N-doped derivatives.

a Au-mediated 3,3’-bipyridine synthesis through 16. b Unsuccessful synthesis of [6]CBPy (21). c Successful synthesis of [9]CPA (24), [9]CPP-Py (25) and [9]CPP-Pyr (26). The conformation of the Au-macrocycles 20, 23, 28 and 30 were assigned tentatively based on the reported all-carbon congeners (see SI 6.6 for full details)^70,74.

Encouraged by these preliminary results, we extended this strategy towards the CPA synthesis and chose the 2,2’-bipyridine bisboronic ester 19 as linker moiety. The triangular gold-bipyridine macrocycle 20 was obtained in 51% yield. However, after oxidation of macrocycle 20, only traces of the desired [6]cycloparaazine (21, [6]CBPy) could be observed by MALDI-TOF MS analysis (see SI 3.8 for details) and its isolation was not possible (Fig. 3b). DFT calculations of the ring strain of [6]CBPy (21) via the homodesmotic reaction method⁷⁵ (B3LYP/6-31 G(d) level of theory, see SI 5.1 for details) revealed that 89.7 kcal/mol additional ring strain has to be overcome during the demetalative cycle formation. With this, the ring strain of [6]CBPys (21) is significantly smaller than that of all-carbon [6]CPP (96.0 kcal mol^–1), which is yielded in 76% under comparable conditions⁷⁰. Therefore, we assumed that not just the strain alone but also the intrinsic nature of the azines plays a crucial role in the unsuccessful synthesis of 21 (see SI 3.8 for full discussion). However, after examining numerous parameters — including reaction conditions, possible intermediates, and various electronic factors (see SI 3.8. for details) — we shifted our focus to ring-enlarged nine-membered derivatives. Since the bisboronic precursor 7 needs to be symmetric, we formally inserted a 2,5-pyrazine in between two pyridine units and selected bispyridinyl pyrazine 22. To this end, we developed a new synthetic route to the designed linker 22 which facilitates a specific Miyaura borylation-Stille coupling sequence (see SI 3.22. for full details). The self-assembly towards the corresponding Au-macrocycle 23 proceeded well, but an inseparable mixture with a structurally related side product resulted reproducibly, which could not be separated. This suggests that the Tsuchido-Osakada approach is also slightly interfered when chelating substrates such as 22 are applied instead of non-coordinative hydrocarbons (see SI 3.12. for details). Fortunately, the sequential oxidation with PhICl₂ (17) and subsequent reductive elimination cascade of the mixture yielded the desired [9]CPA (24) in 8% yield over two steps.

To check the general applicability of this method and to evaluate the effect of the different degree of nitrogen incorporation and positioning on the properties of the compounds, we additionally synthesized two other nine-membered CNRs 25 and 26 through analogous approaches. Both heterocycles bear six nitrogen atoms. In [9]CPP-Py (25), these are located on six different arene entities applying the 1,4-di(pyridin-2-yl)benzene bisboronate 27 as the precursor (self-assembly: 28, 62% yield, reductive elimination: 25, 66% yield). [9]CPP-Pyr (26) carries six nitrogen atoms equally positioned on three heteroarenes when using the 2,5-diphenylpyrazine bisboronate 29 (self-assembly: 30, 88% yield, reductive elimination: 26, 74% yield) (Fig. 3c). Since the rings 25 and 26 do not consist entirely of azines and neither the starting material nor the product contain chelating bipyridine moieties, alternative methods for synthesizing CNRs (see Fig. 2) could also be viable. While the Tsuchido-Osakada approach was found to be both straightforward and highly productive, additional attempts to synthesize other cycloparaazines applying this strategy were unsuccessful due to either an unproductive reductive elimination from the respective Au-macrocycle or failed syntheses of the bisborylated polyazine linkers (see SI 3.8. & 3.24 for details), leaving untouched synthetic space for future endeavors.

Structural properties

With the synthesized NCNRs 24–26 in hand, we set out to examine their structural parameters via X-ray diffractometric analysis of single crystals and compared those to calculated data. Proper single crystals of [9]CPA (24), [9]CPP-Py (25) and [9]CPP-Pyr (26) were obtained from CH₂Cl₂, tetrahydrofuran (THF)/n-hexane and benzene/n-pentane solutions, respectively (Fig. 4a). [9]CPA (24) crystallizes in an overall tubular packing driven by intermolecular CH−N interactions in the tubular alignment and intermolecular π−π interactions between the tubes (see noncovalent interaction (NCI) plot). Interestingly, [9]CPP-Py (25) crystallizes in a 1:1 ratio in two forms: in one dipole-minimalized structure with three nitrogen pointing to both sides (gray structure) and the other structure has a ratio of 4:2 pointing in both directions (orange structure). This results in a herringbone overall packing with intermolecular CH−CH and CH−π interactions. Instead, the constitutional isomer [9]CPP-Pyr (26) crystallizes in a tubular fashion driven by intermolecular CH−N and π−π interactions. Noteworthy, the NCI plots of all three rings 24−26 show no intramolecular N‒H interaction between adjacent pyridines/pyrazines and phenylenes which is an experimental hint of the lower ring strain energy compared to pristine [9]CPP due to lower steric repulsion (see intramolecular NCI plots, SI 6.5.). The ring strain in all N-doped nanorings 24–26 is significantly lower than for their all-carbon congener (see SI 5.2. for details) (Fig. 4b). Comparison of calculated structural data (on the B3LYP-D3/6-31 G(d) level of theory) with the crystallographic data for NCNRs 24−26 and CPP shows generally good agreement. Typically for CNRs with an odd number of incorporated rings³⁰, geometric parameters (α, β, θ and Ø) of 24−26 show characteristic broad distributions indicated by the respective standard deviations given in parenthesis (Fig. 4b) (for a full overview of all measured parameters see SI 6.4, tables 11−18). Most interestingly, the average torsional angles θ of all N-doped rings 24−26 are significantly smaller compared to those of all-carbon [9]CPP. The most heavily N-doped [9]CPA (24) shows the smallest θ overall due to less intramolecular CH−CH steric repulsion (see postulation of Bachrach and Stück⁶², Fig. 1b).

Fig. 4. Experimental and theoretical structural examination of N-doped carbon nanorings.

a X-ray crystal structures of [9]CPA (24), [9]CPP-Py (25) and [9]CPP-Pyr (26), two-dimensional packing and NCI plot (nitrogen atoms are highlighted in blue). Solvent molecules and hydrogen atoms are omitted for clarity. b Observed and calculated structural characteristics of N-doped carbon nanorings 24–26. Calculations were performed at the B3LYP-D3/6-31 G(d) level of theory. Standard deviation is given in parenthesis. Parameters of [9]CPP were measured from a published crystal structure⁹². Source data are provided as a Source Data file.

Electronic properties

To understand the heteroatomic influence on the physical properties of 24–26, we performed optical, electrochemical and theoretical investigations and compared those to literature-known [9]CPP.

The UV–vis absorption and fluorescence emission of all the N-doped nanorings 24–26 are red-shifted compared to [9]CPP (λ_max,abs = 341 nm, ɛ = 1.41 × 10⁵m^–1cm^–1; λ_max,em = 503 nm, Φ_F = 0.62⁷⁶, τ = 8.12 ns⁷⁶) (Fig. 5a). The all-azine ring 24 absorbs at λ_max,abs = 363 nm, but its extinction coefficient (ɛ = 2.13 × 10⁴ m^–1cm^–1) is nearly seven-times lower than that of [9]CPP. The emission of the dodeca-N-doped ring 24 is also bathochromatically shifted at λ_max,em = 527 nm. Its fluorescence quantum yield is decreased to Φ_F = 0.26, but the fluorescence lifetime (τ = 7.67 ns) is similar to that of pristine [9]CPP. In contrast to 24, [9]CPP-Py (25) absorbs at a significantly lower wavelength of λ_max,abs = 355 nm (ɛ = 9.55 × 10⁴m^–1cm^–1), but still red-shifted compared to [9]CPP. The emission occurs at λ_max,em = 512 nm with a fluorescence quantum yield of Φ_F = 0.18 in a lifetime of τ = 4.07 ns. Surprisingly, its constitutional isomer [9]CPP-Pyr (26) shows completely different properties with the strongest overall red-shift in absorption at λ_max,abs = 372 nm (ε = 4.97 × 10⁴m^–1cm^–1) as well as in fluorescence λ_max,em = 532 nm (Φ_F = 0.12, τ = 4.00 ns). These results stand in contrast to those of modestly N-doped carbon nanorings, where only slightly altered photophysical properties compared to pristine CPPs were reported³⁸. As [9]CPP-Pyr (26) shows the strongest effect where six nitrogen atoms are equally centered on only three of the nine incorporated aromatics, it is obvious that also the positioning of the N-doping is highly relevant. The N-doped carbon nanorings 24, 25 and 26 show only a weak solvatochromic effect in the tested solvents (dichloromethane, THF, toluene and chlorobenzene) (see SI 7.1. for details). Noteworthy, the fluorescence quantum yields Φ_F and lifetimes τ of nanorings 24–26 are significantly lower than those of pristine [9]CPP, resulting in nonradiative decay being much favored over radiative decay (Fig. 5c).

Fig. 5. Photophysical properties of nine-membered carbon nanorings.

a UV–vis (plain lines) and fluorescence (dashed lines) spectra recorded of approx. 10 µm solutions in CH₂Cl₂. b Macroscopic appearance of all four analyzed CNRs in solid and liquid state. Excitation was performed with a Kessil lamp (λ_max = 370 nm, 8 W). c Overview of measured photophysical data and calculated HOMO–LUMO gaps. d Frontier molecular orbitals of N-doped carbon nanorings 24–26 and results of time-dependent DFT calculations. Calculations were performed at the B3LYP-D3/6-31 G(d) level of theory. The main transitions with their major contributions (coefficients >0.4) are shown. See SI 5.3., Tables S3−S5 for all transitions. Source data are provided as a Source Data file.

As the red-shift of N-doped CNRs is generally attributed to the electronegative nitrogen atoms contributing with a greater effect to the molecules LUMOs than HOMOs³⁸, we performed time-dependent (TD)DFT calculations on 24–26 at the B3LYP-D3/6-31 G(d) level of theory (see SI 5.3. for details) (Fig. 5d). Noteworthy, the orbitals of the strongly electronically depleted [9]CPA (24) lay significantly lower than for those of the other NCNRs 25 and 26. The calculated HOMO–LUMO gaps of 24–26 rise from [9]CPP-Pyr (26, 2.98 eV) to [9]CPA (24, 3.03 eV) and [9]CPP-Py (25, 3.13 eV) (Fig. 5c). The HOMOs and LUMOs of all N-doped CNRs 24−26 are equally distributed over the whole ring structures. This suggests that the electronical difference of the azines (pyridine/pyrazine) and phenylenes is surprisingly not large enough for displaying a donor-acceptor type character in 24−26 since no characteristic distinctive HOMO–LUMO orbital separation is visible. This is supported by only minor solvatofluorochromic effects observed for 24–26 in different solvents (see SI 7.1. for details). For [9]CPA (24), HOMO–1 and HOMO–2 as well as LUMO +1 and LUMO +2 are fully degenerate. Instead, for [9]CPP-Py (25) and [9]CPP-Pyr (26), the orbitals are only nearly degenerate. Figure 5d shows the NCNRs 24−26 main transitions with their major contributions (coefficients >0.4) color-coded (for a full overview about all transitions, see SI 5.3., Tables S3−S5). Although the absolute values deviate slightly, the calculations are overall in good agreement with the observed experimental properties (for full discussion, see SI 5.3.). Despite being Laporte-forbidden, the HOMO→LUMO transition also contributes very slightly in all nanorings causing small shoulders between 400–450 nm in the visible region.

Additionally, the observed spectroscopic properties are also clearly macroscopically visible. The solid-state color under ambient light of all four nanorings is quite similar yellow since the absorption difference in the visible region is quite marginal. All N-doped nanorings 24–26 show only a slight solid-state fluorescence. Contrary to this, the pristine [9]CPP shows stronger solid-state fluorescence in agreement with its much higher quantum yield. Additionally, the N-doping-induced red shift is clearly macroscopically visible in solution shifting the emitted light from green to yellow (Fig. 5b).

The electrochemical properties of nanorings 24–26 were investigated by cyclic voltammetry (CV) (Fig. 6a). To ensure comparability, reductions were evaluated in THF and oxidations in CH₂Cl₂, respectively. As anticipated, all NCNRs 24–26 showed higher reduction and oxidation potentials than [9]CPP due to their electron-deficient nature. Importantly, while literature reports the reduction of [9]CPP to be irreversible⁷⁷, we found that both oxidation and reduction of [9]CPP are reversible (see peak current ratios, Fig. 6c). This might be attributed to the measurement being conducted under ultra-inert conditions in a glovebox. [9]CPP-Py (25) shows a first reversible reduction at E_red^1/2_rev: –2.05 V, a second irreversible reduction at E_red,irrev: –2.38 V and an irreversible oxidation at E_ox,irrev: 0.96 V. Its constitutional isomer [9]CPP-Pyr (26), with a lower HOMO–LUMO gap (see Fig. 5c), shows a first reversible reduction proceeding at E_red^1/2_rev: –2.00 V, a second irreversible reduction at E_red,irrev: –2.37 V and an irreversible oxidation at E_ox,irrev: 0.91 V. Contrary, [9]CPA (24) is irreversibly reduced at E_red,irrev: –1.65 V, followed by multiple additional reductions at insignificantly smaller potentials. These measurements show that [9]CPA (24) is electrochemically rather unstable due to multiple irreversible reductions proceeding at similar potentials. The oxidation of the most electronically depleted ring 24 proceeds appropriately also at the largest overall potential E_ox,irrev: 1.08 V irreversibly (Fig. 6c). Next, the separation of oxidation and reduction peaks (ΔE_p,red) belonging to the reversible reductions E_red^1/2_rev was examined. This value is, for reversible redox processes, inversely proportional to the number of transferred electrons. The data revealed those to be in the range of 51–55 mV (only reversible events). By comparing these values to the one-electron-process of the ferrocene/ferrocenium couple (101 mV), it was revealed that the reversible reductions of the N-doped nanorings 25 and 26 as well as [9]CPP are two-electron reductions from the neutral species to their dianionic derivatives (Fig. 6b).

Fig. 6. Electrochemical properties of nine-membered carbon nanorings.

a Cyclic voltammograms (CV) of N-doped nanorings 24–26 and [9]CPP in THF (reduction) and CH₂Cl₂ (oxidation). Initial scan directions are indicated by arrows and irreversible reduction potentials are determined on inflection point. Details of electrochemical measurements are described in the SI, chapter 8. b Evaluation of the number of transmitted electrons via the peak separation of reversible reductions of ferrocene and [9]CPP-Py (25). c Overview of measured CV data in THF and CH₂Cl₂, respectively. d Overview of measured CV data in CH₃CN. Source data are provided as a Source Data file. rev.: reversible.

In cyclic voltammetry, testing the reductive and oxidative behavior in the same solvent is generally highly desirable or even mandatory for some electrochemical applications. Therefore, we investigated the redox behavior of [9]CPP-Py (25), [9]CPP-Pyr (26) and [9]CPP also in CH₃CN as its electrochemical stability allows the measurement of both oxidation and reduction in the same solvent. Unfortunately, however, [9]CPA (24) does not show sufficient solubility in CH₃CN. The table in Fig. 6d summarizes the observed electrochemical parameter of those rings in CH₃CN. Most importantly, the observed redox potentials in CH₃CN are slightly shifted compared to the ones measured in THF and CH₂Cl₂, respectively. This illustrates that for appropriate comparability values measured in the same systems (solvent, conducting salt) should be considered.

Application of N-doped carbon nanorings

Encouraged by those measured properties, we addressed potential applications for our N-doped nanorings 24–26. Our prior study reports that simple protonation of N-doped CNR 1 with a Brønsted acid alters photophysical properties (see for comparison introduction section and Fig. 1b)⁶³. We now envisioned that our nanorings 24–26 could show affected properties via complexation with different amounts of a Lewis acid such as tris(pentafluorophenyl)borane (BCF) (Fig. 7a)^78,79. We commenced our studies with a ¹H NMR titration experiment with [9]CPP-Py (25) wherein complexation became obvious. The complexation led to characteristic peak broadening and shifting of the signals of all hydrogen atoms attached to a pyridine, while no change of the phenylene hydrogen atoms was observed (Fig. 7b). Following, the UV–vis absorption and fluorescence titration study in fact revealed that property engineering can be triggered in CH₂Cl₂ (Fig. 7c). The addition of BCF causes a continuous red-shift in the UV–vis spectrum, lowering of the extinction coefficient as well as broadening of the absorption spectrum (from λ_max,0equiv = 355 nm to λ_max,10equiv = 375 nm). Interestingly, the fluorescence maximum is bathochromically shifted upon addition of 2 equivalents of BCF (from λ_max,0equiv = 512 nm to λ_max,2equiv = 546 nm) and the fluorescence quantum yield is also decreased (see SI 9.3. for details). However, when more than two equivalents of the Lewis acid are added, the emission undergoes a blue-shift and the quantum yield is increased (see SI 9.3. for details). Macroscopically, this emission color change is also visible from deep green (λ_max,0equiv = 512 nm) over orange (λ_max,2equiv = 546 nm) to yellow (λ_max,10equiv = 534 nm) (Fig. 7d). This unique spectral effect might be attributed to an induced push‒pull effect between the, throughout BCF-coordination electronically depleted, pyridine moieties and the more electronically rich phenylene rings. This causes a narrowed HOMO–LUMO gap with thus energetically lower, red-shifted absorption as supported by (TD)DFT calculations (see SI 9.4. for details). Moreover, the induced push‒pull effect results in reduced orbital overlap causing the decreased fluorescence quantum yield. With larger amounts of BCF being coordinated to the N-doped nanoring, the dihedral angle θ increases due to steric repulsion causing hindered rotation. Therefore, non-radiative decay modes are more unlikely resulting in a higher probability for a radiative decay, thus explaining the enhanced fluorescence quantum yield (see SI 9.5. for full details). A similar optical phenomenon is visible for [9]CPA (24) in CH₂Cl₂, though the blue-shift occurs only after addition of 5 equivalents of BCF (see SI 9.2. for details). Interestingly, when [9]CPP-Pyr (26) is tested under otherwise identical conditions, nearly no supramolecular property engineering effect is visible in neither the ¹H NMR titration nor the UV–vis and fluorescence studies (see SI chapter 9 for details). This suggests that the less Lewis-basic pyrazine moiety is far weaker binding than pyridine. Additionally, performing those titration studies of the N-doped ring 24−26 in different solvents (toluene, THF and chlorobenzene) showed varying property engineering effects. [9]CPP-Py (25) shows the unique red-shifted/blue-shifted emission in CH₂Cl₂, toluene and chlorobenzene, but [9]CPA (24) only in CH₂Cl₂. [9]CPP-Pyr (26) does not show this effect in any solvent. All rings 24−26 show non-affected properties in THF (see SI 9.2. for details). Of note, since BCF forms a super-Brønsted-acidic adduct with traces of water^80,81, we cannot rule out that also partial protonation of the N-doped nanorings 24−26 proceeds. However, a control titration experiment with the Brønsted-acid trifluoroacetic acid (TFA) showed a different spectral change for which much more equivalents (300 equiv) of acid were needed (see SI 9.2. for details). Therefore, we suggest that mostly the Lewis acidity of BCF induces the observed non-covalent band gap engineering.

Fig. 7. Potential application of N-doped carbon nanorings in non-covalent supramolecular property engineering and long-term performance for energy storage.

a Supramolecular [9]CPP-Py BCF adduct formation. b ¹H NMR titration study of [9]CPP-Py (25) with BCF in CD₂Cl₂ (600 MHz, 293 K). c UV–vis (plain lines) and fluorescence (dashed lines) titration study of [9]CPP-Py (25) with BCF recorded of approx. 10 µm solutions CH₂Cl₂. d Emission of [9]CPP-Py BCF adducts in CH₂Cl₂ (10 µm) under irradiation with a Kessil lamp (λ_max = 370 nm, 8 W). e H-cell cycling performance of [9]CPP-Py (25), [9]CPP-Pyr (26) and [9]CPP against the ferrocene/ferrocenium couple in CH₃CN (0.25 mm). f Electron transmissions during charging and discharging events of [9]CPP-Py (25) and CPP. g Spectroelectrochemical analysis of the reduction of [9]CPP-Py (25) in CH₃CN (0.25 mm). h ¹H NMR study of electrochemical products in d₈-THF. i Back-oxidation of [9]CPPy^•– radical anion (25^•–) in CH₃CN. j Proposed electrochemical processes during charging and discharging, energy calculations and NICS_iso values. The calculations were performed on the PW6B95/def2-QZVP (energies) and HF/6-31 + (d,p) (NICS) levels of theory. Source data are provided as a Source Data file.

Additionally, the reversible reduction of [9]CPP-Py (25), [9]CPP-Pyr (26) and [9]CPP itself stimulated our interest to test them as charge storage materials. Recently, Chen, Du and co-workers⁸², and the Esser group⁸³ applied polymer-incorporated carbon nanorings as anode or cathode material in lithium-ion or organic solid-state batteries, respectively. However, the charge storage properties of pristine carbon nanorings themselves in solution are underinvestigated so far. Thus, H-cell cycling in CH₃CN was conducted to gain further insights into the long-term stability of the reduction process. [9]CPA (24) was not evaluated due to irreversibility of its reduction (see Fig. 6a for comparison), and [9]CPP and [9]CPP-Pyr (26) both showed only poor long-time stability. However, [9]CPP-Py (25) exhibited completely different behavior than both its all-carbon congener and its constitutional isomer 26. After 100 cycles the normalized discharge capacity is still at 36% ([9]CPP: 6%, [9]CPP-Pyr (26): 10%) (Fig. 7e). Although not ideal for an application in a battery, yet this constitutes a great improved electrochemical stability and could therefore stimulate further research. We propose that smaller analogs, like our N-doped rings, represent valuable model substrates for elucidating the origins of the frequently observed superiority in electrochemical properties exhibited by nitrogen-doped nanocarbon materials against their all-carbon congeners^84–87. Thus, this result highlights the relevance of precise nitrogen doping over random doping.

We further examined this electrochemical behavior by looking into the number of transferred electrons during reduction and oxidation events. The cyclic voltammograms suggest that the reduction takes place as a two-electron process (see comparison Fig. 6b). This two-electron reduction from [9]CPP-Py (25) to [9]CPP-Py^2– dianion (25^2–) could also be observed during the first charging of the H-cell. However, in the following oxidation and reduction events, only one electron is transferred (Fig. 7f). This suggests that a first charging event is needed to enter into the operating charge storage mode between dianion 25^2– and radical anion 25^•–. In contrast to this, for [9]CPP, roughly four electrons are transmitted in the charging event, but only one electron is transferred during discharging. In the following charging event again, more electrons are needed as received in the former discharging, disqualifying this system as a charge storage material. To get further insight into the electrochemically involved species, we performed spectroelectrochemical analysis of the process. For [9]CPP-Py (25) a new species appeared with a λ_max,abs around 540 nm (Fig. 7g). Comparison of this absorption spectrum with a calculated one suggests that this spectral change arises from the dianionic species 25^2– (see SI 10.2. for details). Additionally, for the H-cell cycling of [9]CPP-Py (25), an ¹H NMR of the charged species shows no signal at room temperature, but clearly showed characteristic⁸⁸, broad, upfield-shifted signals (4.0–6.5 ppm) corresponding to the dianion 25^2– at –80 °C (Fig. 7h) which fit well to calculated data, and a typical EPR-signal was obtained for radical anion 25^•– (see SI 10.3. for details).

The enhanced performance of [9]CPP-Py (25) compared to both other rings might be attributed to the fact that the ΔE between the desired, first, reversible reduction and the undesired, second, irreversible reduction is slightly larger ([9]CPP-Py (25): 0.43 V; [9]CPP-Pyr (26): 0.38 V; [9]CPP: 0.36 V; all values measured in CH₃CN) (see comparison Fig. 6c). This therefore allows more selective reduction to the desired dianion and most likely the smaller ΔE results in unselective overreduction to the respective tetraanion, a process which was chemically already described for CPPs⁸⁹. Therefore, future research endeavors could be concentrated on increasing this ΔE while keeping the reduction potential as low as possible. Our studies indicate that N-doping might be beneficial for enhancing the electrochemical performance, but also show that too much ([9]CPA (24)) or the wrong nature ([9]CPP-Pyr (26)) of the dopant is also deteriorating. Regarding this, pyridinic doping is likely superior over pyrazinic doping.

Furthermore, we also showed that the [9]CPP-Py^•– radical anion (25^•–) can be fully oxidized to neutral [9]CPP-Py (25) when a higher potential is applied (Fig. 7i) and its corresponding ¹H NMR spectrum showed complete back formation of uncharged species 25 (Fig. 7h). Single energy point calculations of all involved species showed that both reduction steps are energetically downhill (see SI 10.4. for details). For [9]CPP-Py^2– dianion (25^2–-), the singlet state is approximately 6 kcal/mol more stable than its triplet state (see SI 10.4. for details). Additionally, it is known that the reduction of carbon nanorings results in a structural change towards quinoidal-type structures⁹⁰, causing in-plane aromaticity with a surprising stability⁹¹. In fact, NICS_iso calculations at the HF/6-31 + G(d,p) level of theory confirmed that also for dianion 25^2–- the aromaticity is increased from NICS_iso = –2.30 for neutral 25 to NICS_iso = –15.29 (see SI for details). Thus, the aromaticity-breaking back oxidation of the charged species is energetically uphill. Upon oxidation, the molecule undergoes substantial conformational rearrangement transforming from the nearly planar quinoidal structure 25²⁻ (θ = 3.4°) via the intermediate radical anion 25^•– (θ = 12.4°) to the neutral form 25 (θ = 20.9°) (see SI 10.4). Notably, the associated increase in ring strain is significantly larger (∆≈ 6 kcal mol^–1) for the second oxidation leading to the selective monooxidation to the [9]CPP-Py^•– radical anion (25^•–) under H-cell conditions (Fig. 7J).

Discussion

In conclusion, we have developed a synthetic access to an all-N-heterocyclic nanoring (24) through an Au-macrocyclization method. Additionally, other highly N-doped CNRs 25 and 26 were synthesized using the developed method. Although additional attempts to synthesize other CPAs were unsuccessful using the Tsuchido-Osakada approach, it remains the only CNR construction strategy that allowed the synthesis of a CPA derivative. In our hands, CPA derivatives were not accessible through other established methods (Jasti/Bertozzi, Itami, Yamago)^21–23. The measurement of UV–vis and fluorescence spectra showed red-shifts for all N-doped structures compared to [9]CPP. However, quantum yields and fluorescence lifetimes are decreased. This is due to the effect of electronegative nitrogen atoms contributing stronger to a LUMO-lowering than on their HOMOs resulting in overall smaller HOMO‒LUMO gaps, as investigated by (TD)DFT-calculations. Cyclic voltammetry experiments showed that both oxidation and reduction potentials are higher compared to [9]CPP, albeit only the reduction is reversible. The strongly electronically diminished [9]CPA (24) shows very poor electrochemical stability by multiple irreversible reductions at close potentials. The synthesized nanorings 24–26 were additionally examined in non-covalent supramolecular property engineering with BCF, wherein only pyridine-containing nanorings 24 and 25 showed affected properties by coordination with the Lewis acid. Instead, the properties of pyrazine-derivative 26 were not affected. Furthermore, the reduction of nanorings 25, 26 and [9]CPP was analyzed in more detail to assess their potential as energy storage materials. It became evident that selective N-doping has a significant impact on the long-term electrochemical stability, which is crucial for integrating nanocarbon materials into real-world applications. Of note, N-doped carbon nanotubes have already found applications in various batteries^82–85, but their controlled atom-precise synthesis remains highly problematic today. We therefore hope that our analysis of N-doped nanorings could help to get a broader general understanding of N-doping in nanocarbon chemistry.

Most importantly, all experiments in our comparison study revealed that next to the bare number of dopants their positioning is highly relevant for inducing property changes. This stands in contrast to the fact that most N-doping in nanocarbon chemistry proceeds randomly. Therefore, we herewith would like to encourage further research in related fields and expect impactful breakthroughs since the heteroatoms allow the formation of structures inaccessible for pristine nanocarbons.

Methods

Under an argon atmosphere, a bis-borylated linker 7 (1.0 equiv), bis-Au-dcpm complex 14 (1.0 equiv) and Cs₂CO₃ (6.0 equiv) were suspended in a mixture of PhMe/EtOH/H₂O (4:1:1, 0.02 m) and stirred at 50 °C for 24−72 h. The mixture was allowed to cool down to ambient temperature, CH₂Cl₂ was added and the formed precipitate was filtered off. The filter cake was washed with H₂O and EtOH, and dried on high vacuum to yield Au-macrocycle 8. Under an argon atmosphere, the Au-macrocycle 8 (1.0 equiv) was dissolved in C₂H₂Cl₄ (approx. 2 mM) and chilled to 0 °C. PhICl₂ (6.0 equiv) was added as solid in one portion or as solution dropwise, and the resulting mixture was stirred for two hours while the temperature was allowed to rise to ambient conditions. Afterwards, all volatiles were evaporated under reduced pressure. The remaining yellow solid was purified via preparative thin layer chromatography (PTLC) to yield the respective N-doped carbon nanoring as bright yellow solid.

Author contributions

T.D., E.S.H., L.L., D.L., A.Y., A.S. & K.I. conceived all studies and experiments. T.D. and D.I synthesized all compounds. D.I., H.S., and C.G.D. measured and analyzed the X-ray data. D.I., L.L., H.S. and T.K. performed all optical measurements. D.I., L.L. and H.S. performed all calculations. The band gap engineering study was performed by D.I., T.D. and L.L. All electrochemical measurements and applications were performed by E.S.H. Temperature-dependent NMR studies were performed by K.B. Spectroelectrochemical measurements were performed by E.S.H. and L.L. All authors discussed the results and contributed to the preparation of the manuscript. All authors have given approval to the final version of the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Reaction optimization, experimental procedures, characterization data, copies of NMR and EPR spectra, X-ray crystallographic details, photophysical studies, electrochemical measurements, and calculations of all involved compounds and processes are provided in the Supplementary Information. Source data are provided with this paper. CIF crystallographic data files and xyz coordinates of the optimized structures are available as Supplementary Files. Crystallographic data for the structures reported in this article have been deposited at the Cambridge Crystallographic Data Center, under deposition numbers CCDC 2386478 (25) and 2409124 (26). Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures/. All data are available from the corresponding authors upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-59934-5.