Nitrogen doped graphene with diamond-like bonds achieves unprecedented energy density at high power in a symmetric sustainable supercapacitor

Supercapacitors have attracted great interest because of their fast, reversible operation and sustainability. However, their energy densities remain lower than those of batteries. In the last decade, supercapacitors with an energy content of ∼110 W h L−1 at a power of ∼1 kW L−1 were developed by leveraging the open framework structure of graphene-related architectures. Here, we report that the reaction of fluorographene with azide anions enables the preparation of a material combining graphene-type sp2 layers with tetrahedral carbon–carbon bonds and nitrogen (pyridinic and pyrrolic) superdoping (16%). Theoretical investigations showed that the C–C bonds develop between carbon-centered radicals, which emerge in the vicinity of the nitrogen dopants. This material, with diamond-like bonds and an ultra-high mass density of 2.8 g cm−3, is an excellent host for the ions, delivering unprecedented energy densities of 200 W h L−1 at a power of 2.6 kW L−1 and 143 W h L−1 at 52 kW L−1. These findings open a route to materials whose properties may enable a transformative improvement in the performance of supercapacitor components.


Synthesis
Graphite fluoride (1 g) was dispersed in 15 ml DMF and stirred with a teflon coated magnetic stirrer at 500 rpm for 3 days under nitrogen atmosphere in a glass flask. Afterwards, the mixture was sonicated for 4 hours (Bandelin Sonorex, DT255H type, frequency 35 kHz, effective power 160 W), and then left for stirring overnight. Next day, 3 g of NaN 3 was added and the dispersion mixture was heated at 130 ºC for 72 h with a condenser and stirring at 800 rpm in an oil bath.
After the end of reaction, the mixture was left to cool down and washing was performed with DMF, acetone, ethanol, hot water (80 o C), water, and acidified water (3% solution of HCl). The product was separated from the solvents using centrifugation (Sigma 4-16K) at 13 000 rcf.
Finally, washing with water was performed until the material stopped precipitating with centrifuge. Then, the dispersion was subjected to dialysis against ultrapure water, until the conductivity of the dispersion was bellow 0.2 mS cm -1 .

Characterization techniques
Infra-red spectra were measured on an iS5 FTIR spectrometer (Thermo Nicolet), using the Smart Orbit ATR accessory with ZnSe crystal. A drop of a dispersion of the sample in ethanol or water was placed on a ZnSe crystal and left to dry and form a film at ambient environment. Spectra were recorded by summing 50 scans. Nitrogen gas was flowing through the ATR accessory during the background and sample measurements. ATR and baseline correction were used for processing the collected spectra.
X-ray photoelectron spectroscopy (XPS) was carried out with a PHI VersaProbe II (Physical Electronics) spectrometer using an Al K α source (15 kV, 50 W, spot size 100 μm). All binding energies were referenced to the C1s core level of the C-C bond set to the nominal value of 284.8 eV. The obtained data were evaluated and deconvoluted with the MultiPak (Ulvac-PHI, Inc.) software package. Spectral analysis included a Shirley back-ground subtraction and peak deconvolution employing mixed Gaussian-Lorentzian functions.
Images from transmission electron microscopy were obtained with a JEOL 2100 TEM, equipped with an emission gun of LaB 6 type, operating at 160 kV. The high-resolution TEM images were obtained using an FEI Titan electron microscope operating at 80 kV. The samples were also analysed with scanning electron microscopy using Hitachi SU6600 instrument with accelerating voltage of 5 kV. For these analyses, a small droplet of a material dispersion in ultrapure water (concentration approximately 0.1 mg ml -1 ) was placed on a carbon-coated copper grid and left for drying.
The 1H→ 13 C and 19 F→ 13 C CPMAS NMR measurements were performed using a JEOL spectrometer JNM-ECZ400R with superconducting coil having a magnetic field of 9.4 T (working frequency: 399.8 MHz for 1 H, 376.2 MHz for 19 F and 100.5 MHz for 13 C) equipped with a 3.2 mm MAS probe. The spectra were collected at ambient temperature at spinning rate 18 kHz, using contact time 6 ms for all measurements and relaxation delay 6 s for samples GN3 and GN3-4h, and 5 s for GF, while the number of scans was 13000 for sample GN3, 16000 for sample GN3-4h and 88 for pristine GF.
Raman spectra were obtained on a DXR Raman microscope, the 633 nm excitation line diode laser was used.
Thermal analyses were performed in open α-Al 2 O 3 crucibles with a Netzsch STA 449C Jupiter instrument with an adapted quadrupole mass spectrometer (QMS 403C Aëolos) at a heating rate of 10 °C min -1 , under an argon or synthetic air flow (100 ml min -1 ) in the sample compartment.
The concentration of sodium was determined using electrothermal atomization atomic absorption spectroscopy ETA-AAS technique, equipped with a graphite furnace (ContrAA 600; Analytik Jena AG, Germany), with a high-resolution Echelle double monochromator (spectral band width, 2 pm at 200 nm). A continuum radiation source was provided by xenon lamp.
XRD measurements were performed using X'Pert PRO MPD diffractometer (PANalytical) in the Bragg-Brentano geometry equipped with a Co X-ray tube (iron filtered Co Kα radiation with λ=0.178901 nm), fast X'Celerator detector, and programmable divergence and diffracted beam antiscatter slits. Samples were placed on a zero-background Si slide, gently pressed and scanned with a step size of 0.0334°, and the 2θ range of 5°-120° (2θ resolution of 0.017°) was used to record the pattern. Commercially available standards SRM640 (Si) and SRM660 (LaB 6 ) from NIST were used for line positions and instrumental line broadening evaluation, respectively. The crystalline phases identification was performed employing the High Score Plus (PANalytical) software in conjunction with the PDF-4+ and ICSD databases.
The lateral dimensions of the graphene sheets (La: the in-plane correlation length) and the stack size (Lc) were evaluated using the Scherrer formula (Equation 1).
where λ is the X-ray wavelength (0.179 nm for CoKa irradiation), θ is the diffraction angle in radians, β is the full width at half maximum (FWHM) of the deconvoluted peak in radians, and K is a constant (K = 1.84 for La (from the 100 reflection) and K = 0.89 for Lc ( from 002 reflection). [1][2][3] Raman spectroscopy. The lateral size of the sp 2 areas (or the in-plane correlation length) was where C(λ) is a wavelength-dependent constant equal to: where C0 and C1 are also constants, estimated to be −12.6 nm and C1= 0.033, respectively  For the two-electrode system, symmetrical full-cell supercapacitor device was assembled to evaluate the performance, rate stability and cyclic stability of the product. Note, that the electrodes for the CVs did not have the same mass loading as the electrodes for the GCD tests, because during electrode pasting it is not always possible to get the exact same deposited mass. Briefly, active material was homogeneously dispersed in N-methyl-2-pyrrolidone (p.a. ≥ 99%, Sigma-Aldrich) with binder PTFE (Sigma-Aldrich) and conductive carbon (TimCal from MTI) at a ratio of 85:10:5, the mixture was sonicated for 1 hour and mixed using a planetary Before actual testing of material, conditioning was performed as follows: Hold of potential for 5 minutes at 1.2 V, 10 cycles at current density 0.5 A g -1 up to 2 V, 10 cycles at current density 1 A g -1 up to 3.7 V, 5 cycles of current densities 3 A g -1 , 5 A g -1 and 7 A g -1 .
Specific capacitance of the active material (amount of GN3 material with binder and conductive additive deposited on the electrodes) (C s , F•g −1 ) was calculated from GCD discharge curve according to previous reports 6 : Gravimetric and volumetric energy density and power density were calculated as follows: Comparative testing was also performed at 60 °C using the same conditions as for the low mass loading.
All energy and power densities mentioned in the manuscript from materials reported in the literature comply to the equations reported above. All numbers reported were verified independently from the discharge times given in every published article.

Computational details
To explore possible reaction pathways in the initial phase of the N-doping of fluorographene with NaN 3 in DMF (Fig. S3, ESI The impact of N-doping on the development of C-C sp 3 bonds in smaller bi-layered models was studied at the PBE0-D3/6-31++G(d,p) level of theory 12,13 , which describes intermolecular interactions consistently with the DFT/plane wave approach employing the periodic-boundary conditions which was used to study larger bi-layered models as described below.
The first-principles spin-polarized DFT calculations were performed by applying the Perdew-Burke-Ernzerhof functional (PBE) 14 Fig. S1 (a) HR-XPS deconvoluted spectra for the N1s region of the GN3 derivatives after 4h (GN3-4h) and 72h of reaction (GN3). The increase of the pyridinic type nitrogens is in keeping with the lower energy state and higher stability of the pyridinic configurations. 11 (b) HR-TEM image of GN3 material, showing its holey structure.  Comments: Raman spectra of GN3 were analyzed more elaborately, in order to provide further macroscopic information on the structure. Typically, the first order Raman scattering spectra from both amorphous and graphene derivatives [14][15][16][17] are successfully interpreted with deconvolution of the D (around 1350 cm -1 ) and G (around 1590 cm -1 ) bands. Both D and G bands are connected to the sp 2 rings (D to sp 2 rings with adjacent sp 3 defects, and G to crystalline/conjugated sp 2 islands), [18] and since Raman spectroscopy is several-fold more sensitive to sp 2 areas than to sp 3 ones, the D and G bands dominate even in amorphous/active carbons with small content in sp 2 areas. [18] The GN3 spectrum is also dominated by these bands (Fig. S3) including the 4h GN3 intermediate (GN3-4h), and GN3 after annealing in inert atmosphere at 1000 °C (GN3-1000). The deconvolution of the D and G bands and estimation of lateral size, L a , (see experimental part bellow for calculation details) corroborate the small lateral size (4 nm) of the aromatic areas, as obtained from XRD. Very interestingly, the comparison of GN3 before and after annealing at 1000 °C, revealed the same Raman features (also reflected in the respective I D /I G and L a values in Fig. S3). This indicates that no healing process of the graphene backbone takes place during annealing and that the same amount and type of defects are preserved before and after the annealing. This corroborates the extensive holey structure observed from the TEM characterization, giving rise to permanent defects due to large vacancies and sp 3 bonds. Comments: X-ray diffraction (XRD) was performed on GN3, and on the commercial porous carbon, for comparison (PC, Fig. S4). In both cases, the broad bump at the 30° region of the (002) reflection was fitted with two Gaussian peaks, typical for disordered carbons, [11] originating from the diffraction from aromatic graphenic layers (narrow diffraction), and from amorphous sp 3 carbon areas (broader diffraction). From the narrow diffraction and the Scherrer formula, the L c average stack height was obtained, while from the (100) diffraction the L a average lateral size of the graphenic platelets was obtained. L c and L a values of GN3 (Fig. S4b) were very similar with that of other disordered carbons, [12] pointing to the small extent of ordering and limited planarity of the aromatic areas, in agreement with the high content of sp 3 carbons deduced from NMR. Thus, GN3 displays a significantly disordered structure with vacancies (holes) and tetrahedral carbon bonds developed randomly, without showing an interlayer spacing of 0.29 Å. Such disordered XRD patterns have been observed upon the defluorination of fluorinated graphites, [9,13] as well as in carbon fibers, whereby the initial sharp and intense 002 graphitic reflection, after fluorination and defluorination, resulted to very broad, amorphous carbon-like 002 reflection at ca. 0.35 nm (all experimental details and equations used in the analysis of the XRD results are provided later bellow).  After pressing the density increased from 0.5 g cm -3 to 0.6 g cm -3 . (i),(j) High mass GN3 film on Al foil. After pressing the density increased from 0.84 g cm -3 to 2.7 g cm -3 . Smaller electrodes were used in this case because the initial 100 μm film was not stable when placed inside a die for pellet preparation. The low pressure hysteresis observed for GN3 (despite the 40 s equilibration time allowed during sorption/desorption) constitutes the pore size analysis relatively unsafe and can be related to various reasons, including changes of the pore size or strong sorbent/adsorbate interactions, 24 or it can be probably due to requirement of very long equilibration time during the desorption. 25 The shape of the isotherm with the very small knee at the low pressure regions clearly suggests absence of micropores. (c),(d) Pore width distribution curves comparing GN3 and KC, showing more uniform pore-size for GN3 material, compared KC carbon.      S13 (a) Infra-red spectra of the GN3 material after assembly (GN 3-fresh ) and after cycling (GN 3post/washed ). The spectra of the electrolyte (EMIM-BF 4 ) and of the binder (PTFE) which were used are also shown, in order to indicate the bands in the GN 3-post/washed electrode material which do not originate from the material itself. Considering this, it is evident that the spectrum of the GN3 material after cycling is practically identical to the fresh one, verifying its structural stability. (b) Raman spectra of the GN3 electrodes before and after cycling. (c) Infra-red spectra showing the fresh electrolyte and the electrolyte adsorbed on the electrode after cycling 10000 times (spectra obtained without washing the electrode). The two spectra are identical, also verifying the stability of the electrolyte.

Fig. S18
Performance characteristics of the symmetric full-cell built from the GN3 material showing the discharge time (t), specific capacitance (C g ), specific energy and power (E g and P g respectively) as well as the volumetric capacitance (C v ), energy and power densities (E v and P v respectively).