Graphdiyne (GDY), composed of periodically arranged sp- and sp2-hybridized carbon atoms, is an emerging two-dimensional (2D) material and a novel carbon allotrope.1-3 Theoretical studies predicted numerous intriguing properties of GDY,4-9 which can be modulated by tuning the layer number.10,11 Monolayer GDY (ML-GDY) has attracted particular interest because, unlike zero-bandgap graphene, it exhibits a natural direct bandgap and ultrahigh carrier mobilities on the order of 104 cm2 V-1 s-1.12-14 ML-GDY’s excellent semiconducting properties, combined with its extreme thinness, mechanical robustness,5,15 and inherent nanoporous structure,12,16 make it a promising material platform for various applications, including flexible electronics,17,18 energy storage,19 and molecular-sieving membranes.20,21 However, due to the difficulty of preparing GDYs with precisely controlled layer numbers and long-range structural order,11 the theoretical properties of ML-GDY have not been demonstrated experimentally.
Acetylenic homocoupling of hexaethynylbenzene (HEB) is the most common method for the synthesis of GDY,11,22 and the instability and proneness to side reactions of HEB can be largely addressed by using the Eglinton reaction conditions.23-25 However, synthesizing ML-GDY remains challenging for two main reasons (Supplementary Fig. 1). First, the alkyne–aryl and alkyne–alkyne single bonds can freely rotate during the coupling of the HEB monomer, leading to out-of-plane random growth of the framework and eventually to the formation of a three-dimensional disordered structure rather than the desired 2D crystalline structure.17,26,27 Second, free GDY monolayers tend to stack into multilayers via van der Waals forces and π–π interactions to lower the surface energy.28,29
Several synthetic strategies have been developed to overcome these obstacles. Chemical vapor deposition was used to grow GDY on an Ag substrate to reduce the degrees of freedom of the HEB monomers. However, only an ill-defined carbonaceous framework was obtained.30 Confining the monomer coupling to graphene surfaces,25,29 oil–water or solid–liquid interfaces,24,31 and electric double layers17 yielded GDYs with ordered structures, but in the form of multilayers rather than monolayers. A bulk GDY material was used to prepare ML-GDY by mechanical exfoliation.32 However, given the limited crystallinity of the bulk GDY, as evidenced by the lack of in-plane X-ray diffraction peaks, the as-prepared ML-GDY may be structurally disordered. Moreover, the physical properties of the obtained ML-GDY, which could in turn provide evidence for the material quality, have never been explored. Therefore, the reliable preparation of crystalline ML-GDY remains a synthetic challenge.
This study reports a confined-space synthesis method that enables the preparation of crystalline ML-GDY by using MXene as a template. In this method, HEB monomers diffuse into the periodic interlayer gaps of MXene, where they subsequently polymerize to form the GDY structure in the presence of Cu ions. The sub-nanometer interlayer space of MXene acts as a confinement reactor to effectively suppress the random out-of-plane growth or vertical stacking of GDY. Diffraction, atomic-resolution imaging, and spectroscopic characterizations provide comprehensive and unambiguous evidence for the diffusion, intercalation, and polymerization of HEB monomers within the MXene interlayer space as well as the formation of crystalline ML-GDY therein. The MXene template can be removed by ion intercalation-assisted liquid phase exfoliation to obtain free-standing ML-GDY flakes with micrometer-scale lateral dimensions; this allows the fabrication of field-effect transistor (FET) using ML-GDY for the first time. The ML-GDY device exhibits substantially higher carrier mobility and electrical conductivity than the previously reported multilayer GDY materials.
Growth of ML-GDY in MXene
The preparation of ML-GDY began with selectively etching Al from a dense MAX (Ti3AlC2) phase using HF to obtain a layer-structured MXene (Ti3C2Tx; T: OH or F) phase (Fig. 1a).33 Note that although MXenes prepared in this way contain mixed OH and F groups at interlayer surfaces,34 the structural models in this report only display OH terminal groups for simplicity.
Scanning electron microscopy (SEM) images revealed that the as-prepared MXene maintained the lateral dimensions (~10 µm) of the pristine MAX crystals while exhibiting lamellar features along the vertical direction (Supplementary Fig. 2). X-ray photoelectron spectroscopy (XPS) results confirmed the complete removal of Al and the formation of the MXene structure (Supplementary Fig. 3).35 Powder X-ray diffraction (PXRD) patterns indicated that the conversion of MAX to MXene resulted in a peak shift of the (002) reflection from 9.53° to 8.82° in 2θ, corresponding to an increase in the d spacing from 0.93 to 1.00 nm (Fig. 1b).
The as-prepared MXene powders were dispersed in pyridine to expand their interlayer spaces through molecular intercalation, as evidenced by the increase in the d(002) spacing to 1.30 nm (Supplementary Fig. 4). Next, HEB monomers were introduced into the MXene suspension in pyridine (Fig. 1a). The addition of HEB results in a further increase in the d(002) spacing to 1.37 nm (Fig. 1b and Supplementary Fig. 5), indicating that HEB molecules could also enter the interlayer space of MXene. The Raman spectrum of the resulting material showed a band at 2107 cm-1, which is characteristic of the alkyne groups in HEB,25 confirming the intercalation of HEB into MXene (Supplementary Fig. 6). Because the limited interlayer space of MXene cannot accommodate “standing” HEB molecules, the intercalated HEB molecules must adopt a horizontal in-plane configuration (Supplementary Fig. 5). Subsequently, Cu ions were added into the suspension to initiate the Eglinton coupling reaction between HEB monomers (Supplementary Fig. 7). A portion of the Cu ions diffused into the MXene interlayer space, where they initiated the reaction governed by the steric confinement effect to form crystalline ML-GDY. Meanwhile, free Cu ions and HEB monomers in the solution yielded amorphous GDY, which could be largely removed from the GDY-containing MXene by centrifugation due to its much lower density.
The purified GDY-containing MXene (denoted as GDY-MXene) exhibited the same crystal morphology as the MXene template (Supplementary Fig. 8a), indicating that growth of GDY was confined within the interlayer space. The PXRD result revealed that that the in-situ growth of GDY resulted in a further increase of the d(002) spacing to 1.54 nm (Fig. 1b). This d value corresponds to an available interlayer space of 0.66 nm, which is just enough to accommodate one layer of GDY (Supplementary Fig. 5). Notably, GDY-MXene exhibited a diffraction peak at 2θ of 10.5o with a d spacing of 0.83 nm, which was larger than the d(004) of MXene while perfectly matching the d(100) of the ideal ML-GDY structure (Fig. 1b and Supplementary Fig. 8b). This result indicates that GDY grown in the MXene interlayer space has in-plane structural order (i.e., is crystalline).
Control experiments demonstrated that intercalated solvent (i.e., pyridine) molecules could be removed by vacuum heating, as manifested by the restoration of the interlayer spacing of MXene; likewise, intercalated Cu ions could be removed by washing with dilute HCl solution (Supplementary Fig. 9). In contrast, after the sequential addition of pyridine, HEB, and Cu ions, the significantly expanded interlayer spacing of MXene could not be restored by vacuum heating or HCl washing; this provides additional strong evidence for the formation of an extended GDY framework within MXene.
Scanning transmission electron microscopy (STEM) was used to probe the structural evolution from MAX to MXene and, finally, to GDY-MXene. For each material, high-angle annular dark-field (HAADF) and integrated differential phase-contrast (iDPC)-STEM images were acquired simultaneously to obtain comprehensive structural information; HAADF-STEM can clearly identify the layered structure composed of Ti, whereas iDPC-STEM is more sensitive to the interlayer light elements (Al, O/F, and C).36 To clearly resolve the atomic columns, the images were all acquired along the [110] zone axis, using specimens prepared with a focused ion beam (Supplementary Fig. 10).
The HAADF-STEM image of MAX showed alternating Ti3 and Al layers with strong and weak contrast, respectively (Fig. 1c). The corresponding iDPC-STEM image showed the Al layers more clearly and even identified the C columns near the Ti columns (Fig. 1c and Supplementary Fig. 11), due to the enhanced contrast of light elements. The HAADF-STEM image shows that, compared with MAX, MXene has larger but empty spaces between Ti3 layers (Fig. 1d). This observation is consistent with the PXRD results and indicates the complete removal of Al by selective etching. The iDPC-STEM image of MXene confirmed the absence of Al layers, while identifying terminal O/F columns on the surfaces of Ti3C2 layers (Fig. 1d and Supplementary Fig. 12). Although the HAADF-STEM image of GDY-MXene was blurred by the electron beam-induced carbon contamination, it revealed a significantly expanded interlayer distance (Fig. 1e), consistent with the PXRD result. Notably, the iDPC-STEM image clearly showed continuous linear contrast between adjacent Ti3C2Tx layers (Fig. 1e, 1f), providing the most direct evidence for the successful growth of ML-GDY within the interlayer space of MXene. Figure 1g compares the intensity profiles of the marked regions in the iDPC-STEM images, clearly illustrating the differences in the interlayer structure between the three materials. The large-area STEM image of GDY-MXene demonstrates the uniform distribution of ML-GDY throughout the specimen (Supplementary Fig. 13).
Electron energy loss spectroscopy (EELS) was performed during STEM imaging of GDY-MXene (Fig. 2a). The high-energy resolution (~50 meV) allowed probing different chemical states of carbon species. As shown in Fig. 2b, the C-K edge EELS spectra acquired from the Ti3C2Tx layer and the interlayer space (i.e., ML-GDY) differ in the energy-loss near-edge structure (ELNES). Compared with Ti3C2Tx,37 the ELNES of ML-GDY shifts slightly to the high energy region, where the primary peak at 285.52 eV and the shoulder peak at 285.93 eV can be attributed to the 1s→π* excitation of carbon−carbon double bonds and the 1s→π* excitation of carbon−carbon triple bonds, respectively.38 An additional peak is observed at 287.45 eV, which may originate from the C-O bonds generated by the partial oxidation of ML-GDY.31 The elemental map based on the C-K edge (285.5-287.5 eV) EELS (Fig. 2c) shows stripes with alternating strong and weak intensities, which matches the HAADF-STEM image (Fig. 2a) but with the reversed contrast. This result indicates that the carbon content of the interlayer space is higher than that of the Ti3C2Tx layer, demonstrating a high filling rate of ML-GDY in MXene.
To avoid possible interference from GDY formed in solution, a focused ion beam was used to remove the outer surface of a GDY-MXene crystal and then characterized the exposed inner structure using Raman spectroscopy (Supplementary Fig. 14). The obtained Raman spectrum showed a band at 2174 cm-1, which is characteristic of the diacetylenic linkages in the extended network (Fig. 2d).17,25,39 The absence of the band associated with terminal alkynes at 2107 cm-1 indicates a high degree of HEB polymerization, while the appearance of G and D bands provides additional evidence for the formation of GDY within MXene (Fig. 2d). Moreover, the Raman intensity map shows a uniform distribution of diacetylenic linkages over the investigated region (Supplementary Fig. 15). The spectral fitting results of C 1s XPS show that, compared with MXene, GDY-MXene had an additional peak related to the carbon–carbon triple bond22,24 and a significantly lower C-Ti peak (Fig. 2e). The presence of amorphous carbon with carbon–carbon single and double bonds in MXene has been previously reported and ascribed to contaminants introduced during synthesis.40 The detection of sp-hybridized carbons further evidences the formation of GDY in MXene.
Selected area electron diffraction (SAED) was performed on the periphery of a GDY-MXene crystal, where the crystal was thin enough to allow electron beam penetration. The acquired SAED pattern shows the coexistence of two sets of hexagonal lattices (Fig. 2f). The lattice composed of strong diffraction spots with small d values can be well indexed based on the [001]-projected MXene structure (a = b = 0.31 nm). The lattice composed of weak diffraction spots with large d values perfectly matches the SAED pattern simulated from an ideal 2D hexagonal ML-GDY structure (a = b = 0.96 nm) (Fig. 2f and Supplementary Fig. 16).24 The SAED result not only reinforces the conclusion from PXRD that the ML-GDY grown in the interlayer space of MXene is crystalline, but also reveals the 30-degree orientation relationship between ML-GDY and the MXene template (Fig. 2f). It was found that the diffraction spots associated with ML-GDY rapidly disappeared during SAED, suggesting that the structure of ML-GDY is extremely sensitive to electron beam irradiation even under the protection of the MXene template.
Free-standing GDY monolayers
MXene can be easily exfoliated into Ti3C2Tx monolayers by sonication in solution (Supplementary Fig. 17),41 whereas the same treatment cannot decompose GDY-MXene (Supplementary Fig. 18), implying that the growth of ML-GDY led to stronger interlayer interactions. In order to obtain free-standing GDY monolayers from GDY-MXene (Fig. 3a), Li2SiF6 was used to facilitate the liquid-phase exfoliation via Li+ intercalation.32 The Li+ intercalation into GDY-MXene increased its interlayer spacing (Supplementary Fig. 19), after which GDY-MXene crystals were decomposed by sonication to form a homogeneous suspension in water (Fig. 3b). The powder collected from the suspension was dispersed on a silicon wafer and then characterized using grazing incidence x-ray diffraction. Three reflections were observed, attributed to the (002) and (004) planes of GDY-MXene after Li+ intercalation and the (100) planes of ML-GDY, respectively (Fig. 3b). This result demonstrates that the ordered in-plane structure of ML-GDY was preserved during the sonication-assisted exfoliation process, but also indicates that GDY-MXene crystals were not completely exfoliated into GDY and Ti3C2Tx monolayers.
Although the obtained powder is a mixture of GDY monolayers, Ti3C2Tx monolayers, and residual GDY-MXene sheets, GDY monolayers could be identified by atomic force microscopy (AFM). As shown in Fig. 3c, AFM revealed flakes of different thicknesses, the thinnest being ~0.77 nm (Fig. 3d). Given that the thickness of Ti3C2Tx monolayer is about 1.5 nm (Supplementary Fig. 20), flakes with sub-nanometer thicknesses can be unambiguously assigned to GDY monolayers.32,42,43 Statistics based on AFM observation of 150 GDY monolayers give a lateral dimension distribution of 0.3-2.4 μm centered at 0.75 μm (Supplementary Fig. 21). The Raman spectra collected from different regions of a sub-nanometer-thick flake (pre-identified using AFM) all showed the band characteristic of diacetylenic linkages, confirming its homogeneous GDY structure (Figs. 3e, 3f). The gradually decreasing signal intensity in the Raman spectral series can be attributed to the damaging effect of prolonged laser irradiation (Fig. 3f and Supplementary Fig. 22).
Under transmission electron microscopy, free-standing GDY monolayers can be identified from the weakest image contrast combined with the lack of Ti element determined by EELS or energy-dispersive X-ray spectroscopy (Supplementary Fig. 23). Free-standing GDY monolayers are extremely sensitive to the electron beam; thus, our attempts to verify their crystallinity using high-resolution TEM imaging were unsuccessful. We compared the EELS spectra collected from a free-standing ML-GDY and the amorphous carbon film of the TEM grid (Fig. 3g, 3h). In the low-loss region, the spectrum of ML-GDY reveals a bandgap of 1.04 eV, which is in good agreement with the theoretical prediction using the first-principles calculations;44 in contrast, the spectrum of amorphous carbon film showed no bandgap excitation (Fig. 3g). In the core-loss region, ML-GDY exhibits a shoulder peak at 285.91 eV, associated with the 1s→π* excitation of sp carbon bonded carbon–carbon triple bonds,38 which is not observed in the spectrum of amorphous carbon (Fig. 3h). Moreover, the amorphous carbon has stronger σ* excitation than ML-GDY, suggesting a higher density of sp3 carbon (Fig. 3h). These results demonstrate that, despite the inevitable partial structural damage during EELS, free-standing ML-GDY displays distinctly different properties from amorphous carbon, including a well-defined bandgap and the presence of sp-hybridized carbon.
Electronic properties of ML-GDY
Semiconducting 2D materials are considered critical for the development of next-generation electronic devices. Theoretical studies indicate that GDY is a semiconductor with electronic properties varying with number of layers. To date, only multilayer GDY materials have been tested.14,25,29,31,45-49. The confined-space synthesis reported here enables the preparation of isolated ML-GDY, thus allowing the first measurement of its electronic properties by fabricating a series of FET devices.
The ML-GDY FET was fabricated by depositing Au/Ti on ML-GDY as source and drain terminals and using single-crystal Si as the bottom gate and SiO2 as the dielectric layer (Fig. 4a and Supplementary Fig. 24). The AFM and Raman characterization results confirmed that the devices were fabricated on ML-GDY (Supplementary Fig. 25). The gate leakage current density of the fabricated ML-GDY FET was as low as 10-3 A cm-2 at ± 30 V (Supplementary Fig. 26a), confirming the excellent insulating properties of the dielectric SiO2 layer. The linear Ids–Vds curve measured at 298 K indicated an ohmic-like contact between ML-GDY and electrodes (Fig. 4b), and the calculated conductivity of ML-GDY was 5.1×103 S m-1. The conductivity of ML-GDY increased with the measurement temperature and reached 7.1×103 S m-1 at 398 K (Fig. 4c and Supplementary Fig. 26b). This behavior suggests that ML-GDY is a semiconductor while demonstrating its good thermal stability.
The transfer characteristic curve (at Vds = 0.05 V) of the ML-GDY FET showed a spurt in current growth at Vg = ~ 2.5 V. The current reduction during the forward sweep of Vg indicated hole-dominated conduction and thus the p-type nature of ML-GDY (Fig. 4d).29,31 The output characteristic curves of the device exhibited an obvious evident gate control property (Fig. 4d, inset), which was not observed in previously reported multilayer GDY-based FETs. The carrier mobility of ML-GDY was calculated using the equation μ = [dIds/dVg][L/(WCgVds)],50 where L is the channel length of the FET, W is the width of ML-GDY (see Fig. 4a), and Cg = 34.5 nF·cm-2. The calculated carrier mobilities from three independent FET devices were 247.1, 213.7, and 233.3 cm2 V−1 s−1 (Supplementary Fig. 27 and Supplementary Tab. 1), respectively, with an average value of 231.4 cm2 V−1 s−1.
As summarized in Fig. 4e, ML-GDY exhibits higher electrical conductivity and carrier mobility than previous multilayer GDY materials prepared using various methods (see details in Supplementary Tab. 2). ML-GDY’s outstanding electronic properties can be attributed to its high structural order achieved by the confined-space synthesis and single-atom thickness. Nevertheless, the measured carrier mobility of ML-GDY is still considerably lower than theoretical prediction values, which may be due to the partial structural damage of ML-GDY and the introduction of contamination during the fabrication of the FET devices.