Introduction

The 2D ferromagnetic materials, which are amenable to atomic-scale charge and spin manipulation, have been proposed as ideal candidates for next-generation memory and logic devices1,2,3. In addition to conventional transition metal or rare earth based 2D ferromagnetic materials4,5,6, great efforts have been made to explore 2D metal-free ferromagnetic materials, which are more chemically benign and affordable. However, lacking of localized spins, metal-free materials are intrinsically diamagnetic. Therefore, the state-of-the-art tuning of the s/p electronic structures are required to render metal-free material ferromagnetism7,8,9. Specifically, the electronic structure tuning of 2D metal-free material can be achieved by precisely modulating the magnetic coupling through structural optimizations.

Graphite carbon nitride (g-C3N4), a structural analog of graphene and typical 2D metal-free material, has been extensively investigated due to its unique optical and electronic properties, as well as good chemical and thermal stability10,11. Various strategies including heteroatoms doping12,13, carbon or nitrogen defect constructing14,15 and chemical functionalization16 have been proposed to optimize the electronic structure of g-C3N4 to acquire ferromagnetism. For instance, fabrication of N defects on g-C3N4 nanosheets can facilitate the emergency of net magnetic moments, leading to the establishment of room-temperature ferromagnetic ordering17. Substitution of carbon by boron in the triazine-based g-C3N4 monolayer can modulate the electronic density of states, generating half-metallicity ferromagnetism order18. However, ferromagnetic g-C3N4 prepared from the aforementioned methods suffered from excessive low magnetization (0.015 emu/g).

To enhance the ferromagnetism, a plausible strategy is to induce long-range magnetic sequence in the plane, which requires short-range exchange interactions between the magnetic moments19,20. Ideally, the distance of the induced magnetic moment originated by the defect is expected to be less than 1 nm19,21. However, the actual distance between the nearest neighboring defects is usually about 10 nm19, hindering the formation of long-range magnetic sequence in conventional defective 2D materials.

In this work, one strategy of establishing magnetic moment exchange interaction through Boron decoration is proposed. It is well-known that sp2 hybridized boron atoms are prone to adapt trigonal planar geometry, which matches well with the planar 2D geometry of g-C3N4. Additionally, when the sp2 hybridized boron was introduced to the g-C3N4 motif, an empty p orbital perpendicular to the 2D plane is available for electron exchange. Prior to this work, planar B has been successfully introduced into carbon nanoribbons at atomic scale through a bottom-up approach, leading to a well-defined B-doped 2D nanoribbon with a modulated electronic configuration22,23. Based on the aforementioned theoretical analysis and literature precedents, in this work, sp2 hybridized B was chosen to be introduced into the g-C3N4 motif to facilitate the magnetic moment exchange interactions to optimize its ferromagnetism (Fig. 1a, b). Specifically, the introduction of planar B into the g-C3N4 motif is accomplished by supercritical CO2 treatment, which has been well demonstrated in heteroatom doping24,25,26. To our knowledge, introducing bridging B to establish the exchange of local magnetic moments and form long-range magnetic order, tuning magnetic property of 2D materials is proposed for the first time.

Fig. 1: Schematics and preparation strategy of borate decorated 2D g-C3N4 nanosheets with in-planar bridging -B(OH)- group.
figure 1

a Structure of sp2 hybridized B with an empty p orbital. b Proposed g-C3N4 structural motif with in-planar boron bridge. c Proposed reaction route of borate decorated 2D g-C3N4 nanosheets (B-C3N4-X MPa) with in-planar bridging -B(OH)- group.

Results

To fabricate the borate-functionalized 2D amorphous g-C3N4 nanosheets (B-C3N4-X MPa), wherein B refers to the introduced boron bridges and X refers to the pressure of SC CO2, one-step SC CO2 treatment was utilized, where H3BO3 is used as the boron source. In this process, the straining and anisotropic acidic etching effect of SC CO2 is expected to cleave the weak van der Waals forces between layers and the in-planar hydrogen bonds of tri-s-triazine units in the layers, thereby exfoliating bulk C3N4 into 2D nanosheets, facilitating the formation of the amorphous and defective structure with -NHx groups26,27,28,29,30. Meanwhile, since H3BO3 is presented over the preparation process, the empty sp2 orbital on B of H3BO3 is expected to be nucleophilic attacked by the lone pair of -NHx groups of the defected g-C3N426, forming in-planar bridging -B(OH)- groups (Fig. 1, detailed mechanism in Scheme S1). Theoretical calculations were conducted to analyze the feasibility of the reaction in Fig. 1c. Specifically, free energy pathway for -B(OH)n- incorporation to the g-C3N4 framework (Scheme S1S2 and Figure S1) was analyzed. As results, the incorporation of terminal -B(OH)2 is endergonic with an energy barrier of 1.07 eV, which could be achieved over the SC CO2 treatment. The formation of the in-planar bridging -B(OH)- groups is thermodynamically more favorable to occur with a reduced energy barrier (0.64 eV).

Transmission electron microscopy (TEM) and high-resolution TEM (HRTEM) images (Figure S2a, b) reveal the transparent sheet-like morphology. No lattice fringes are detected according to the selected area electron diffraction (SAED) pattern (Figure S2b, inset), indicating the as-prepared B-C3N4−16 MPa is amorphous. According to the atomic force microscope (AFM) and the associated height images (Figure S2c), the thickness of the ultrathin nanosheet morphology of B-C3N4−16 MPa is 2–3 nm, corresponding to 6~9 single atomic layers.

To characterize the crystal structure of B-C3N4−16 MPa, X-ray powder diffraction (XRD) pattern were investigated (Fig. 2a). The bulk C3N4 exhibits two typical diffraction peaks at 13.1° and 27.3°, originating from (100) in-plane long-range atomic order and (002) interlayer-stacking motif, respectively. In contrast, only a broad (002) peak was observed for the B-C3N4-16 MPa, the (100) peak almost disappeared31,32. The difference in Fig. 2a suggests that the supercritical CO2 could destruct the periodic atomic structure in tri-s-triazine framework, leading to amorphous structure.

Fig. 2: Characterizations of bulk C3N4 and B-C3N4-16 MPa samples.
figure 2

a XRD patterns of bulk C3N4 and B-C3N4-16 MPa. bd FTIR spectra of bulk C3N4 and B-C3N4-16 MPa.

To verify the structural change in B-C3N4-16 MPa, Fourier transform infrared (FTIR) spectroscopy was performed. As shown in Fig. 2b, the bulk C3N4 exhibits several characteristic peaks in the region of 3000–3400, 1200–1700, and 810 cm−1, which are assigned to the vibrational absorption of N-H, aromatic g-C3N4 heterocyclic units and the heptazine rings, respectively33,34. However, after the SC CO2 treatment, several new peaks emerge on the spectrum of B-C3N4-16 MPa, the peak at 2480 cm−1, 2140 cm−1, and 2000–2040 cm-1, which are assigned to the stretching vibration of C-O35, N = C = O36, and C–O or C–N37, respectively (Fig. 2c). The FTIR results demonstrate the destruction of partial N = C–N bonds and form new covalent bonds. Additionally, the absorption intensity of N-H (3000–3400 cm−1) is substantially enhanced, suggesting an increased concentration of -NHx groups38. Overall, the IR characterizations indicate that SC CO2 could cleave the chemical bonds, leading to the formation of abundant -NHx groups. The –NHx groups are expected to react with H3BO3 to introduce B bridges into the g-C3N4 motif. In addition to the structural changes, the formation of -B(OH)- moieties is characterized by IR spectroscopy, as shown in Fig. 2d. The B-C3N4−16 MPa exhibit new peaks in the enlarged region of 500–1200 cm−1. The peaks at 1011 and 885 cm−1 are corresponding to the deformation vibration of N-B-O and bending vibrations of B-O-H, respectively, and the peaks at 731 and 554 cm−1 are assigned to the torsional vibration of O–H bond39,40,41,42. The presence of these peaks confirms that the expected -B(OH)- groups are successfully introduced into the g-C3N4 framework with the assistance of the SC CO2.

To further elucidate the construction and surface chemical states of B-C3N4−16 MPa, X-ray photoelectron spectroscopy (XPS) characterizations were performed. Both XPS survey spectrum and B 1 s spectrum with binding energy at 193.3 eV indicate the introduction of boric acid (Fig. 3a and Figure S3)40,43,44. In addition to B analysis, the surface N/C ratio decreases from 1.28 (bulk C3N4) to 0.70 (B-C3N4-16 MPa) according to the XPS results (Table S1), indicating the loss of lattice nitrogen atoms. As shown in Fig. 3b, c, the characteristic peaks of CN heterocycle frameworks have been detected in the C 1s and N 1s spectra, wherein the typical peaks of N═C─N, N─(C)2, and N─(C)3 located at 288.1, 398.5 and 399.0 eV, respectively40,45. For B-C3N4-16 MPa, the higher binding energies of the C 1s and N 1s are associated with the structural change induced by the N defect. It is notable that the integrated peak area ratio between N-(C)2 and N-(C)3 in N 1s is significantly decreased from 2.69 (bulk C3N4) to 1.29 (B-C3N4-16 MPa), indicating the preferential loss of N-(C)2 sites46. Additionally, the decreasing of N-C = N peak in the C 1s spectrum confirms the loss of N atoms in N-(C)2 sites. Notably, Table S1 shows that the -NHx components of B-C3N4−16 MPa increase significantly (from 17.16% to 28.30%), suggesting more -NHx groups are generated after the SC CO2 treatment, consistent with the results of FTIR. Additionally, an apparent peak corresponding to B-O bond arises at 531.4 eV in the O 1s spectrum of B-C3N4−16 MPa, confirming the successful introduction of -B(OH)- to the g-C3N4 framework (Fig. 3d).

Fig. 3: Chemical state and coordination information for B-C3N4−16 MPa.
figure 3

af High-resolution XPS of B 1s a, C 1 s b, N 1s c, and O 1s d for B-C3N4−16 MPa. e B K-edge XANES spectra and f solid-state 11B CP-MAS-NMR spectrum of B-C3N4−16 MPa.

As discussed above, two scenarios of B introduction can be expected: bridging -B(OH)- and terminal -B(OH)2 (Scheme S1S2). Thus, X-ray absorption near-edge structure (XANES) spectra and solid-state 11B MAS NMR measurements were conducted to probe the coordination environment of the B atoms. As shown in Fig. 3e, the B K-edge spectrum displayed the spectral fingerprints of B–N bonds with a sharp B 1s → π* at about 194.0 eV and B-O bonds with B 1s → σ* resonances at 196.6 eV47,48, suggesting the -B(OH)−/−B(OH)2 groups were introduced and coordinated to the N atoms of g-C3N4. Subsequently, 11B NMR was utilized to investigate the binding mode of boron groups (bridging -B(OH)- or terminal -B(OH)2). In Fig. 3f, the 11B NMR spectrum of B-C3N4-16 MPa demonstrated two signals located at −4 to −2 and 3–15 ppm, suggesting both bridging -B(OH)- and terminal -B(OH)2 groups exist in the sample49,50. Importantly, XANES and the 11B NMR results confirmed the existence of the proposed bridging -B(OH)- in the B-C3N4-16 MPa material, which is critical to establish the exchange of local magnetic moments and optimizing the magnetic properties.

After the structure of defective 2D g-C3N4 nanosheets containing bridging -B(OH)- groups were fully characterized, the magnetic properties of the as-obtained B-C3N4-16 MPa were subsequently investigated by a superconducting quantum interference device magnetometer. The ferromagnetic nature was confirmed by the appearance of magnetic hysteresis, saturation magnetization (MS), and coercivity (HC). As shown in Fig. 4a, b, bulk C3N4 exhibits a negligible ferromagnetic response with Ms of 0.002 emu g−1 at 300 K, consistent with the previously reported values17. After SC CO2 treatment in the presence of H3BO3, the magnetism is significantly enhanced compared to the bulk C3N4. The MS and HC of B-C3N4-16 MPa at 300 K are 0.043 emu g−1 and 95 Oe, respectively, superior to most of the reported carbon-based materials (Table S2). Similar to B-C3N4-16 MPa, B-C3N4-12 MPa and B-C3N4-20 MPa at 300 K also exhibit room-temperature ferromagnetism with saturation magnetization of 0.035 and 0.026 emu g−1, respectively (Figure S4S6 and S8). The enhancement of the ferromagnetism could be attributed to the introduction of in-planar bridging -B(OH)- groups, which facilitates exchange interactions of the magnetic moment located at the N defects. Since magnetic impurity elements of B-C3N4-X MPa could lead to ferromagnetism as well, inductively coupled plasma mass spectrometry (ICP-MS) was employed to ensure that the magnetism observed originated from the C3N4 material (Table S3). According to the ICP-MS characterization, all samples contain similar and negligible metal-based impurities. Thus, the ferromagnetism observed in Fig. 4 is attributed to the B-C3N4-X MPa instead of magnetic impurities.

Fig. 4: Ferromagnetic characterizations of bulk C3N4 and B-C3N4-X MPa samples.
figure 4

a M–H curve of bulk C3N4 and B-C3N4-X MPa at 300 K; b The corresponding magnified M-H curves of bulk C3N4 and B-C3N4-X MPa near H = 0. c FC-ZFC magnetization curve of B-C3N4−16 MPa in external magnetic field of 100 Oe. d ΔM = MFC − MZFC curve of B-C3N4−16 MPa.

Additionally, the zero-field-cooled/field-cooled (ZFC/FC) signals were measured under an applied field of 100 Oe. As shown in Fig. 4c, the ZFC and FC curves exhibit the observable divergence, confirming the ordered spin alignment that appeared in B-C3N4-16 MPa. Since no blocking temperature in ZFC curve is observed and ZFC/FC curves are divergent until 550 K, it can be concluded that the thermal energy is incapable of disturbing the magnetically ordered state51,52 and the Curie temperature is around 550 K12 (Fig. 4d).

Density functional theory (DFT) calculations were further conducted to investigate the origin of the room-temperature ferromagnetism of B-C3N4-X MPa. Although the introduction of in-planar bridging B are proposed to be critical to the enhanced ferromagnetism of B-C3N4-X MPa, the formation of N defects over the SC CO2 is expected as well, which could lead to ferromagnetism18,53,54,55. Therefore, the effect of N defects on magnetism is investigated first. As expected, B-C3N4-X MPa with N defects are found to be ferromagnetic with local magnetic moment according to theoretical calculation (see Supplementary Information for detailed information).

After the local magnetic moment formed by N defects is investigated, DFT calculation was conducted to investigate the pivotal role of the in-planar bridging B on electronic structure and ferromagnetism of g-C3N4. As for g-C3N4 with N defects, an asymmetric distribution of electrons around the tri-s-triazine motifs is found according to the charge density distribution (Fig. 5a), and the magnetic moment of this supercell is 1.9892 μB. Since both in-planar bridging -B(OH)- and terminal -B-(OH)2 groups (Fig. 1c, Scheme S1S2) are introduced according to 11B NMR and EXANES, their magnetic properties were theoretically investigated separately. Interestingly, since the in-planar bridging -B(OH)- groups connect two tri-s-triazine units, more spin charge distribution in the non-magnetic regions is achieved via electron and magnetic moment delocalization. Thus, local magnetic moment induced by N defects successfully established the exchange interaction, which could form long-range magnetic sequence through the introduction of B bridges (Fig. 5b). In sharp contrast to the bridging -B(OH)- scenario, the simulation demonstrates that the introduction of terminal -B(OH)2 groups didn’t contribute positively to the magnetism and the magnetic moment of the supercell was reduced to 1.9758 μB (Fig. 5c).

Fig. 5: Spin-polarized DFT-calculations for the magnetism of samples.
figure 5

The corresponding top-view of spin charge density of a g-C3N4 with N defects. b g-C3N4 with N defects and in-plane bridging -B(OH)- groups. c g-C3N4 with N defects and terminal -B(OH)2 groups. d g-C3N4 with N defects and sp3 hybridized -N(OH)- groups. The yellow and blue equipotential profiles represent the majority-spin up and minority-spin down, respectively. The values in the figure correspond to the magnetic moment of the supercell.

To further confirm the critical role of the sp2 hybridized electronic structure of the introduced B atoms, we compared the magnetism of analogous g-C3N4 material with a bridging -N(OH)- moiety, where N atom is trigonal pyramidal with sp3 hybridization instead. In contrast to the planar bridging -B(OH)- groups, the non-planar geometry of -N(OH)- groups obstruct the electron distribution and the magnetic moment delocalization, resulting in a reduction of the magnetic moment to 1.9620 μB (Fig. 5d). Overall, the calculations suggest the planar geometry of the introduced bridging boron are critical to achieve the magnetic properties of B-C3N4, consistent with the experimental observations.

Discussion

In summary, to facilitate the magnetic moment exchange interaction and acquire ferromagnetic 2D metal-free nanosheets, in-planar bridging boron is introduced into the g-C3N4 motif, where the geometry of boron bridges perfectly matched to the 2D g-C3N4 motif. As expected, the as-prepared B-C3N4 nanosheets exhibit strong ferromagnetism with a Curie temperature around 550 K. Subsequently, the critical role of the in-planar bridging -B(OH)- groups to the B-C3N4 ferromagnetism is revealed through theoretical calculation. In contrast, the introduction of non-planar -N(OH)- or terminal -B(OH)2 groups didn’t contribute to the ferromagnetism of g-C3N4 according to calculations, indicating the bridging planar -B(OH)- group is crucial to the magnetic moment exchange in B-C3N4. Therefore, this work proposes a strategy to enhance the ferromagnetism of 2D materials, which is introducing a geometric matched planar bridging group to build magnetic moment exchange interaction. Such strategy is expected to be applied to prepare additional 2D magnetic materials for next-generation memory and logic devices.

Methods

Materials and reagents

Commercially melamine, Hydrogen peroxide (H2O2, 30% aqueous solution), and ethanol were purchased from Sinopharm Chemical Reagent Co., Ltd. (China). Boric acid (H3BO3) was purchased from STREM CHEMICALS (Fluka, CAS Number: 10043-35-3).

Synthesis of bulk g-C3N4

The bulk C3N4 were fabricated by a typical synthesis route24. In brief, melamine (10.0 g) in a covered combustion boat was heated at 550 °C for 3 h using a heating rate of 10 °C min−1 in a muffle furnace. The resulting yellow product was grounded into power using an agate mortar for further use.

Synthesis of borate decorated 2D amorphous g-C3N4 via SC CO2

Bulk C3N4 (50 mg) and H3BO3 (100 mg) were added to the solution mixture of anhydrous ethanol (5 mL) and 30% H2O2 solution (5 mL). After the solution was sonicated for 30 min to form a homogeneous solution, the mixture was transferred into the supercritical CO2 apparatus with a heating jacket and a temperature controller. The autoclave was heated to 80 °C, and then CO2 was charged into the autoclave to the desired pressure (16 MPa). The sample was treated for 4 h under stirring. After the SC CO2 treatment, the sample was cooled down to room temperature and the CO2 gas was released. Finally, the dispersion was sonicated for 1 h, and then centrifuged at 3823 ×g for 15 min to remove aggregates.

Subsequently, the supernatants obtained from 3823 ×g centrifuge were further centrifuged at 10619 ×g for 20 min and the precipitate was washed with deionized water until the solution became neutral. The purified precipitate was dried in a vacuum oven at 60 °C overnight. For control experiment, samples were prepared under different CO2 pressures (12 and 20 MPa) with the same procedures. C3N4-20 MPa was prepared at 20 MPa, 80 °C without the addition of H3BO3.

Material characterizations

The morphology and structure of the samples were characterized by tapping mode AFM (Nanoscope IIIA), TEM (JEOL JEM-2100), and HRTEM. X-ray diffraction (XRD) patterns were collected on a XPERTPRO (Netherlands). X-ray photoelectron spectroscopy was performed using a Thermo ESCALAB 280 system with Al/K (photon energy = 1486.6 eV) anode mono-X-ray source. All binding energies were calibrated by using the contaminant carbon (C 1 s = 284.8 eV) as a reference. Fourier transform infrared (FTIR) spectra were obtained on a Nicolet Nexus spectrometer. The magnetic properties of the samples were measured by a SQUID magnetometer with a sensitivity of <1 × 10−8 emu. Specifically, the magnetic properties were measured by compressing the samples into linear diamagnetic plastic capsules, which were carefully handled using nonmagnetic tweezers, capsules and tapes to avoid unintentional magnetic contamination. Solid-state 11B magic-angle-spinning nuclear magnetic resonance (11B MAS-NMR) measurements were performed on Bruker AVANCE III 400 MHz WB solid-state NMR spectrometer. The spectra were acquired using 4 mm MAS NMR probes with a spinning rate of 10 kHz. The detection of residual metal content in the bulk C3N4, B-C3N4−12 MPa, B-C3N4−16 MPa and B-C3N4-20 MPa samples was performed by inductively coupled plasma mass spectrometry (ICP-MS). The exact amount of the sample (10 mg) was immersed in a concentrated nitric acid (≥99.999% trace metals basis) and heated for 2 h at 100 °C. Afterward, the mixture was transferred into 100 mL volumetric flask, diluted with water, and the undissolved samples were caught by a 200 nm Millipore filter. The obtained concentration of metals in the solution was recalculated to the amount of the tested sample (analogically, diluted nitric acid was used as a blank). X-ray absorption near edge structure (XANES) measurements were conducted at the insertion-device beamline of the Materials Research Collaborative Access Team (MR-CAT) at the Advanced Photon Source located within the National Synchrotron Radiation Laboratory. The energy resolution for the absorption spectra is around 40 and 80 meV for the boron edges, respectively. The B K-edges XANES spectra were collected using the sample drain current mode.

Computational methods

All the calculations were based on Density Functional Theory (DFT) as implemented in the Vienna Ab-initio Simulation Package (VASP) code utilizing the projector augmented wave method (PAW)56,57,58. The exchange-correlation energy of generalized gradient approximation proposed by Perdew, Burke, and Ernzerhof (GGA-PBE) was adopted59. A vacuum of 20 Å perpendiculars to the sheets was applied to avoid the interaction between layers. A kinetic energy cut off of 450 eV was used for the plane-wave basis set. The sampling in the Brillouin zone was set with 5 × 5 × 1 by the Monkhorst-Pack method60,61. Convergence criteria employed for both the electronic self-consistent relaxation and ionic relaxation were set to be 10−4 and 0.02 eV/Å for energy and force, respectively.