Experimental and theoretical investigations of functionalized boron nitride as electrode materials for Li-ion batteries

Abstract

The feasibility of synthesizing functionalized h-BN (FBN) via the reaction between molten LiOH and solid h-BN is studied for the first time and its first ever application as an electrode material in Li-ion batteries is evaluated. Density functional theory (DFT) calculations are performed to provide mechanistic understanding of the possible electrochemical reactions derived from the FBN. Various materials characterizations reveal that the melt-solid reaction can lead to exfoliation and functionalization of h-BN simultaneously, while electrochemical analysis proves that the FBN can reversibly store charges through surface redox reactions with good cycle stability and coulombic efficiency. DFT calculations have provided physical insights into the observed electrochemical properties derived from the FBN.

---

1. Introduction

The ultimate goal of electrochemical energy storage development is the simultaneous achievement of several attractive properties of an ideal energy storage device: high gravimetric and volumetric energy and power densities, safety of operation, economic composition, rechargeability, and cycle life.¹ Covalently functionalized monolayer 2D materials (F2D-s) composed of light elements, such as derivatives of graphene (G) or hexagonal boron nitride (h-BN), are attractive platforms that have potentials to achieve all the above mentioned properties simultaneously.^2–5 F2D-s eliminate the need of cations to diffuse in 3D crystals because these are monolayer materials and reactions occur on their surfaces. Under these conditions cations can hop directly from the electrolyte to the reactive/intercalating surface sites, allowing for very fast charging/discharging rates. There are many layered materials, e.g., MoS₂, WS₂, MoSe₂, WSe₂, and TiS₂, that can potentially be exfoliated to be monolayers. However, in contrast to graphene and monolayer h-BN, these transition metal-containing monolayers cannot offer high specific energy and power densities because of the presence of heavy transition metal elements.

Although functionalized BN (FBN) materials have never been studied for electrode applications, oxygen-functionalized graphite nanoplatelets (OfGNP-s) have been investigated previously by Liu, et al.² as cathode materials for Li-ion batteries, and have demonstrated extremely high specific power in the order of 10–100 kW kg⁻¹ combined with specific energy of 100–1000 W h kg⁻¹. Stable operation with little capacity loss has been demonstrated for 1000 cycles. A similar approach using a sodium anode and OfGNP cathode has been studied by Kim, et al.³ demonstrating specific energies of 100–500 W h kg⁻¹ and specific powers of 50–100 kW kg⁻¹ with 300 cycles at a current rate of 1 A g⁻¹. In the OfGNP cathodes the reactive surface species are either oxo (C=O) or epoxide (C–O–C) functional groups that become reduced to C–O–(Li/Na) during discharge.³ The capacity and energy density of an OfGNP cathode depend on the concentration of oxygen on the surface of the nanoplatelets. The greater the oxygen concentration, the higher the capacity and energy density.^2,3 Unfortunately, the high concentration of oxygen increases the risk of thermal runaway processes in which the OfGNP may explosively decompose to carbon monoxide,⁶ questioning whether OfGNP will ever be safe enough for mass-applications. Liu, et al.² also discuss other than oxo or epoxide functionalizations of GNP-s, such as –COOH, –OH, –NH₂, –OR, and –COOR functionalization. However, these functionalizations have not been tested and thus the reactivity and cycle-stability of these species are uncertain.

A potentially safer, thermally more resistant alternative to OfGNP may be FBNs, as proposed by Nemeth.^4,5 Monolayer h-BN is resistant to oxidation up to 850 °C (ref. 6) at which oxidation and cleavage of h-BN monolayers happen. In contrast, monolayer graphene can be oxidized at as low as about 200 °C in a stream of oxygen containing about 1% water² and graphene oxide explosively decomposes at about 300 °C.⁷ No explosive decomposition of oxidized h-BN has been observed.⁶ Therefore, it is worthwhile to explore the possibility of utilizing FBNs as electrodes for Li-ion battery applications. No such studies have ever been conducted before, although layered BN has been used previously as an additive to graphene oxide electrodes to prevent re-stacking of graphene oxide.⁸

The covalent functionalization of h-BN with –OH,^9–12 –NH₂,¹³ =CBr₂,¹⁴ –F,¹⁵ and substituted phenyl¹⁶ functional groups has been realized through radicals^{9–11,13–16} or nucleophilic/electrophilic attack^2,17,18 of anions/cations on the boron/nitrogen atoms. Both boron and nitrogen atoms can be covalently functionalized. Radicals and anions based nucleophilic attacks create bonds to the electron-deficient boron atoms (Lewis acid), while molecular cations may functionalize the electron-rich nitrogen atoms (Lewis base). Molten alkaline salts have been used for decades to etch h-BN whereby they covalently functionalize and dissolve h-BN.^17,18 Non-covalent functionalization of h-BN is also possible, for example, by molten amines that utilize a boron-amine Lewis acid–base interaction.^19,20 To the best of our knowledge, although many functionalized h-BN species exist, so far only a single one is known in both alkalinated (reduced) and non-alkalinated (oxidized) forms, namely hydroxyl functionalized boron nitride, for which both the BN(OH)_x^9–11 and BN(OH)_xNa_x (where 0 < x ≤ 1)¹² forms have been synthesized.

In this study, we investigate, for the first time, the synthesis of FBN powder via the reaction between molten LiOH and solid h-BN and its first ever application as an electrode material in Li-ion batteries. Various materials characterizations reveal that the melt-solid reaction can lead to exfoliation and functionalization of h-BN simultaneously, while electrochemical analysis and density functional theory (DFT) calculations prove that FBN can reversibly store charges through surface redox reactions.

2. Materials and methods

Synthesis of FBN powder

FBN powders were prepared by melt-solid reaction between LiOH (MC/B Manufacturing Chemists) and h-BN (Alfa Aesar) mixed in 1-to-1 molar ratio. The starting materials were high-energy ball milled for 3 h using a SPEX machine (SPEX Mixer 8000M) under an argon atmosphere to ensure uniform and intimate mixing. Stainless steel balls in a stainless steel canister with a weight ratio of ball-to-powder 20 : 1 were utilized for the ball milling. The ball milled mixture was then pressed into a dense pellet with a crimping machine (MTI MSK-110 Hydraulic Crimping Machine) at ∼5 MPa in an argon-filled glovebox (MBraun) with both moisture and oxygen levels <1 ppm. The pellet was then put into a tightly sealed stainless steel autoclave with a copper gasket and heated to 500 °C (above the melting point of LiOH, 462 °C) for 8 hours. The reacted pellet was collected and then ground into small particles with the aid of a mortar and pestle in the glovebox. The ground powder was ball milled for 10 min using the SPEX machine with a charge ratio of ball-to-powder 20 : 1. The ball-milled powder was then collected in the glovebox and marked as “FBN” for later use.

Materials characterizations

Fourier transform infrared (FTIR) spectra were collected on a Nicolet, Nexus-IR 470 spectrometer using the potassium bromide (KBr) pellet method. The powder was mixed with KBr in a 1 : 100 charge ratio and the mixture was ground using a mortar and pestle in the glovebox filled with Ar of 99.98% purity. The ground mixture was then pressed into a pellet of 1 cm in diameter. The measurement range of the spectrum was set to be from 4000 to ∼670 cm⁻¹. The X-ray diffraction (XRD) patterns were recorded using Bruker X2 with Cu Kα radiation in the 2θ range of 5–100°. To avoid oxidation during data collection, all powders were loaded into glassy capillary tubes in a glovebox and sealed before XRD data collection. The sizes and morphologies of the powders were examined through scanning electron microscopy (SEM, Jeol JSM 5900-LV). The SEM images were obtained by adding the powders onto the carbon tape which served as the substrate, followed by gold coating which could improve the conductivity of the powders so that we can get images without charging.

X-ray photoelectron spectroscopy (XPS) was carried out with a Physical Electronics 5000 VersaProbe II system equipped with a monochromatic aluminum Kα (15 kV) X-ray source. The excitation beam size employed was 100 μm and the power was 25 W. The background pressure in the analysis chamber was around 1.0 × 10⁻⁷ Pa. XPS spectra were recorded in fixed analyzer transmission mode. A survey spectrum was acquired using a pass energy of 117.4 eV, a step size of 1 eV, and an acquisition time of 150 ms per step. Individual element regions were acquired using a pass energy of 23.5 eV, a step size of 0.2 eV, and an acquisition time of 3.0 s per step. In all cases Ar⁺ and e⁻ neutralization were employed. Adventitious carbon on the sample was used to calibrate the binding energy (B.E.) scale with respect to C–H environments at 284.8 eV. Individual element regions were further deconvoluted into components by least square fitting to a sum of Gaussian or Gaussian–Lorentzian components and a Shirley background using the Multipack package by Physical Electronics, provided with the system.

¹H, ⁷Li and ¹¹B solid state NMR experiments were performed on a 500 MHz Bruker Advance III spectrometer (11.7 Tesla superconducting magnet) with a 2.5 mm MAS probe operating at 30 kHz spinning speed. A rotor synchronized echo pulse sequence (90°τ–180°τ_acq), where τ = 1/ν_r (spinning frequency) was used for all the acquisitions. A π/2 pulse width of 2.0 μs and pulse delay of 2 s was used for ¹H experiments and the chemical shift was referenced to TMS at 0 ppm. ⁷Li experiment was performed with a π/2 pulse width of 2.5 μs and pulse delay of 5 s and was referenced with 1 M LiCl at 0 ppm. For ¹¹B MAS NMR, a π/2 pulse width of 2.6 μs and a pulse delay of 2 s were used. 0.1 M boric acid was used as a secondary reference at 18.8 ppm. The simulation and the fit of spectrum was performed using a shape analysis package included in Topspin software. All the experiments were performed at room temperature and the sample was packed under an inert atmosphere in a glovebox.

Electrochemical characterizations

The as-synthesized FBN powder was mixed with carbon black (CB) and PVDF (Alfa Aesar) in a mass ratio of 4 : 5 : 1 in the argon-filled glovebox, and then mixed with 1-methyl-2-pyrrolidinone (NMP, Alfa Aesar) solvent to make a uniform slurry which was subsequently pasted onto an Al foil current collector. In this first ever study, high CB content (50 wt%) was used in the FBN electrode so that the intrinsic charge/discharge properties of FBN can be studied without interference of low electronic conductivity. The pasted Al foil was then transferred to a vacuum oven in the argon-filled glove box (H₂O < 1 ppm and O₂ < 1 ppm) and heated to 120 °C for 12 h. The dried electrode was punched into circular disks of ∼15 mm in diameter and the mass loading of active material on each electrode ranges from 1.6 to 2.8 mg cm⁻². The electrolyte consisted of 1 M lithium hexafluorophosphate (LiPF₆, Sigma-Aldrich) in EC/DMC (3 : 7 v/v) (Sigma-Aldrich). The FBN half cells with a lithium metal anode were fabricated in the form of CR2032 coin cells, and the cycling performance was evaluated using an Arbin System 2000. The charge/discharge processes were started with charge to 3.3 V and then cycled between 1.0 and 3.3 V at a current density of 28 mA g⁻¹ FBN for most of the cells. Several cells were also cycled between 0.4 and 3.3 V or between 0.01 and 3.0 V to study the effect of the voltage window for charge/discharge cycles. Cyclic voltammetry (CV) data were obtained using a Parstat 4000 (Princeton Applied Research) with FBN half cells initially at the open circuit voltage (OCV) condition and examined within a voltage window of 0.5 V and 3.5 V at a scan rate of 0.1 mV s⁻¹.

Since the amount of CB in the electrode was high (50 wt%), pure CB half cells (i.e., 90 wt% CB + 10 wt% PVDF versus a lithium metal anode) were also fabricated in order to take the charge storage effect of CB in the calculation of the specific capacity of FBN. In making these half cells, the procedure of making FBN half cells was followed precisely. Furthermore, the mass loading of CB in each electrode was set to be ∼2 mg cm⁻², similar to that used in the FBN electrodes. The CB half cells were charged and discharged between 1.0 and 3.3 V with a current density of 28 mA g⁻¹ CB which again was similar to that used for charging/discharging FBN half cells.

3. Density functional theory calculations

DFT calculations have been carried out using the Quantum Espresso program package,²¹ following the methodology in ref. 4. A plane wave basis set with 50 Ry wave-function cut-off has been used along the PBEsol exchange-correlation functional^22,23 with ultrasoft pseudopotentials as provided by the software package. Functionalized h-BN monolayers were separated from each other by about 30 Å vacuum layers and 3D periodic boundary conditions were applied. Each simulation cell involved a 2 × 1 supercell of h-BN, functionalized by two identical functional group of either epoxide (–O–) or hydroxyl (–OH), where one of the functional groups was bound to one of the B atoms from the top of the monolayer, while the other functional group was bound to the bottom of the monolayer at the other B atom, such as indicated in Fig. 8. Both spin-polarized and spin-unpolarized calculations have been carried out leading to only spin unpolarized solutions at the converged geometries in all cases. The k-space grid was 4 × 4 × 2, this allows for a mRy convergence of the electronic energy with respect of k-space saturation and geometry convergence. This accuracy is satisfactory for the prediction of accurate cell voltages within the chosen exchange-correlation functional and surface model. The surface models have been relaxed until residual forces on the atoms became smaller than 0.0001 Ry/bohr and residual pressure on the simulation cell was less than 1 kbar. Energies of cell reactions were calculated from electronic energy differences of products and starting materials, all in the crystal phase. Cell voltages are computed as the negative of the cell energies divided by the number of electrons transferred in the reaction. The methodology has been validated on experimental data of lattice parameters and enthalpies of formation of α-Li₃BN₂, Li₃N, and h-BN. Enthalpies of formations were estimated as the change of electronic energy during the formation of the compounds from the corresponding elements at T = 0 K, i.e., from crystalline Li and B and N₂ gas. Experimental lattice parameters have been reproduced within 2.5% error, while experimental enthalpies of formation were with 4%.²⁴ Both spin-polarized and spin unpolarized calculations have been carried out leading to non-magnetic solutions in the surface models used that included two functionalized BN units, with one B atom functionalized from above and the other from below the surface of the monolayer.

4. Results and discussion

Fig. 1(a) compares the FTIR spectra of h-BN, LiOH, and the melt-solid reaction product (i.e., the as-synthesized FBN). Pure h-BN has two major IR active modes: one at ∼1370 cm⁻¹ involving B–N stretching and the other at ∼817 cm⁻¹ involving out-of-plane bending, in good accordance with the previous studies of pure h-BN.^10–12,25 The reaction product is expected to be Li(HO)BN, Li(O)BN, Li₂(O)BN, (HO)BN, (O)BN, or a mixture of these compounds, depending on the nature of the melt-solid reactions at 500 °C. In any case, most of the B atoms are expected to be either converted to sp³ hybridization from the planar sp² states destroying the conjugation of the pi-electron system or become edge B atoms by cleaving the h-BN monolayers. This is supported by the drastic decrease in the intensity of the out-of-plane bending, similar to the HO-functionalized BN.¹¹ The edge B atoms may or may not be functionalized.

Fig. 1 (a) The FTIR spectra, (b) XRD patterns of h-BN, LiOH, and the melt-solid reaction product, and (c) comparison of the XRD pattern of the melt-solid reaction product with those of possible reaction products as indicated. The intensities of all XRD patterns in (b) are plotted with the same scale and thus the intensities of the new peaks from melt-solid reaction product are very small. However, these peaks are magnified in (c) for easy observation.

The peak at ∼1155 cm⁻¹ is associated with the B–O deformation vibrations, similarly to an analogous band observed in (HO)_xBN.^11,12 While ref. 11 reports a very broad and shallow peak between 3000 and 3600 cm⁻¹ that is assigned to hydrogen bonds between and within (HO)_xBN layers, we could not identify its analogue in our product. This suggests that the FBN formed through the melt-solid reaction may contain little Li(HO)BN or (HO)BN because some inter- and intra-molecular hydrogen bonds are expected from Li(HO)BN and (HO)BN. However, as we will see later, XPS and NMR data indicate the presence of OH-functional group in the as-synthesized FBN. Thus, we hypothesize that one of the possible reasons for the lack of intra-FBN hydrogen bonding in the FTIR data might be due to the fact that only about 1/3 of B atoms are functionalized (as indicated by XPS and NMR analyses to be discussed later) and thus most of the surface –OH groups are too far from each other to form hydrogen bonds. In addition, the presence of the –OLi group on the surface (identified using XPS and to be discussed later) can also prevent the formation of intra-FBN hydrogen bonding. The lack of inter-FBN hydrogen bonding is likely due to the relatively random orientations of FBN monolayers which does not allow for the formation of a sufficient number of inter-FBN hydrogen bonds to be detected by FTIR.

Fig. 1(b) compares the XRD patterns of h-BN, LiOH, and the as-synthesized FBN. Note that all the peaks of h-BN and LiOH are absent in the melt-solid reaction product, indicating that exfoliation of h-BN has taken place during the melt-solid reaction. This is consistent with the expectation because the completely exfoliated h-BN has no XRD peaks,²⁶ similarly to graphene and graphene oxide.³ The complete disappearance of XRD peaks can be expected for any monolayer FBNs as well, such as for Li(HO)BN, Li₂(O)BN, Li(O)BN, and (O)BN. Another possibility for disappearance of h-BN peaks is amorphization. However, the FTIR analysis above reveals that the h-BN bonding is still present after the melt-solid reaction. Therefore, disappearance of h-BN peaks cannot be attributed to amorphization, but is due to exfoliation of h-BN into monolayers with functionalization on the surface of monolayers. In spite of functionalization and exfoliation, it is noted that accompanied with disappearance of h-BN peaks and LiOH peaks, several new yet small intensity peaks appear. The intensities of these peaks are very small since the XRD patterns in Fig. 1(b) are collected under the same conditions and presented in the same scale. Thus, these small peaks suggest the formation of a small quantity of crystalline phase(s) along with FBNs. The nature of these trace amounts of crystalline phases is unknown at this stage, but they are not the possible compounds of crystalline Li₃BO₃, LiBO₂, Li₂BNO and BNO since these compounds do not match the crystalline peaks of these compounds as shown in Fig. 1(c) where the peak intensities of these trace amounts of crystalline phases have been magnified for easy observation and their peak positions are marked as well.

Fig. 2 shows XPS spectra of the as-synthesized sample. Fig. 2(a) contains a survey spectrum showing that, other than small amounts of F and C (∼1 at% and ∼5 at%, respectively), the sample does not contain detectable amounts of any atomic species not expected in the as-synthesized FBN. Fig. 2(b) through Fig. 2(f) contain detailed XPS scans of the B1s, N1s, O1s, Li1s, and C1s regions of the same sample. Each of these figures contains the experimental data (black dots) after subtraction of a Shirley background, as well as fit and individual components (continuous lines) resulting from a least squares deconvolution of the data. Reported binding energies (B.E.), taken from references listed in the NIST XPS Database,²⁷ are also indicated by discontinuous vertical lines in Fig. 2(b) through Fig. 2(f), for reference. Table 1 summarizes the sample composition estimated from the XPS data presented in Fig. 2(b) through Fig. 2(f), while Table 2 lists the fit parameters resulting from the least squares deconvolution. Note that both Tables 1 and 2 do not contain H element because XPS cannot detect H element directly.

Fig. 2 XPS spectra of the as-synthesized FBN: (a) survey spectrum, (b–f) B1s, N1s, O1s, Li1s, and C1s regions, respectively. Data points in panels (b) through (f) correspond to (Shirley) background-corrected data, while solid lines represent a least squares fit to the data and its components.

Table 1 Chemical composition of the FBN sample determined from the XPS regions in Fig. 2(b)–(f)

Element	B	N	O	Li	C	F
Concentration (at%)	33	29	21	12	5	1

---

Table 2 Fit parameters for XPS regions in Fig. 2(b)–(f)

Region	Centroid (eV)	FWHM (eV)	Area (% of region)	Gaussian ratio (%)	χ2
B1s	187.22	2.63	4.64	100	1.02
	188.71	1.63	9.08	100
	190.60	1.96	71.75	100
	192.66	1.58	14.53	100
N1s	395.89	2.43	13.9	77	2.50
	397.86	2.00	65.95	87
	398.35	1.43	20.15	77
O1s	530.42	4.23	28.72	92	0.73
	531.44	2.43	49.11	100
	532.99	2.01	22.17	100
Li1s	55.70	2.89	100	60	1.25
C1s	284.90	2.71	77.10	78	1.25
	288.70	3.25	22.90	68
F1s	685.38	2.88	100	57	1.13

---

The B1s spectral region, as shown in Fig. 2(b), is dominated by a Gaussian contribution at B.E. ∼190.6 eV, close to the reported 190.5 eV value of B in BN. Two additional components centered at 192.66 eV and 188.71 eV can be respectively attributed to electron-deprived and electron-enriched B environments. The component centered at 192.66 eV is shifted from BN toward B(OH)₃ and B₂O₃ environments, reported at 193.6 eV and 193.7 eV, respectively, consistent with electron withdrawing OH-functionalization or O-functionalization of B atoms resulting in HO–BN and O–BN environments, respectively. Note that LiO–BN will also result in a similar shift as HO–BN, and thus the peak at 192.66 eV is assigned to HO–BN, LiO–BN and O–BN (Table 3). In contrast, the component centered at 188.71 eV is shifted toward B₁₀H₁₄ at 187.8 eV, consistent with H functionalized B in H–BN environments. Finally, a minor, broad component centered at 187.22 eV can be mainly attributed to metallic B, perhaps with some contributions of boron hydrides and carbides.

Table 3 Sample-level atomic concentration and summary interpretation of peak components listed in Table 2. See the text for details

Region	Centroid (eV)	Concentration (at% of sample)	Assignment
B1s	187.22	1.5	B
	188.71	3.0	H : BN
	190.60	23.7	BN
	192.66	4.8	HO : BN/O : BN/LiO : BN
N1s	395.89	4.0	BN : H
	397.86	19.1	BN
	398.35	5.8
O1s	530.42	6.0	Organic
	531.44	10.2	Li2CO3/Li2O/LiOH
	532.99	4.6	HO : BN/O : BN/LiO : BN
Li1s	55.70	12.0	Li2CO3/Li2O/LiOH/LiF/LiO : BN
C1s	284.90	3.4	C–H/Organic
	288.70	1.0	Li2CO3
F1s	685.38	1.0	LiF

---

The N1s spectral region, Fig. 2(c), is well described by three Gaussian–Lorentzian peaks centered at 395.89 eV, 397.86 eV, and 398.35 eV. The main peak (at 397.86 eV), which accounts for ∼65% of the signal in the N1s region, is very close to the reported N1s B.E. of BN (398.1 eV) and is broad enough (FWHM – 2 eV) to comfortably cover the BN range and to encompass the narrower peak (FWHM – 1.4 eV) centered at 398.35 eV, which itself accounts for ∼20% of the signal in the N1s region. This secondary peak is shifted towards N environments with lower electron density with respect to BN, such as lithium azide and ammonia, reported at 398.7 eV. Conversely, the peak centered at 395.89 eV is shifted towards higher electron density environments, consistent with H functionalization of the N atom in BN; i.e., BN–H environments. The sum of the atomic concentrations associated with the two higher B.E. components of the N1s region (Table 3) accounts for ∼25% of the observed atoms in the sample. This is in good agreement with the amount of atoms accounted for by the B1s contribution at 190.6 eV (∼24%) and BN stoichiometry. However, the breadth of the N1s contributions and their lack of clear resolution suggest that, rather than representing three clearly distinguished N environments in BN (i.e., BN and two different N functionalizations), they constitute a mathematical approximation to a more complicated N1s peak shape stemming from a distribution of N environments in the sample. For example, it is not difficult to imagine that an HO–BN functionalized B atom will pull some additional charge from the neighboring N atoms to partially compensate for the charge donated to the OH group. In this case, the N atoms neighboring this B atom will show up at higher B.E. than in pure BN (398.1 eV). This interpretation is somewhat supported by the fact that the total amount of N in the sample detected by XPS (∼29 at%) is close to the total amount of B attributed to BN, H–BN, HO–BN, LiO–BN and O–BN-like environments (∼31 at%). The small discrepancy from overall BN stoichiometry is within standard estimates of XPS accuracy of a few at%, but could be indicative of a slight preference for B termination at the edges of the BN sheets, or stabilization of N vacancies due to the functionalization.

The O1s spectral region, Fig. 2(d), is described by three Gaussian–Lorentzian components centered at 530.42 eV, 531.44 eV, and 532.99 eV. The main component, which is centered at 531.44 eV and accounts for half of the area of the region, is close to the reported B.E. of lithium carbonate (531.5 eV), lithium oxide (531.3 eV) and lithium hydroxide (531.2 eV). This peak probably represents remaining unreacted LiOH and/or contamination from its reaction with adventitious atmospheric hydrocarbons. The highest B.E. contribution in the O1s region, centered at 532.99 eV, is shifted towards H₂O (533.2 eV) and B(OH)₃ (533.4 eV) positions, consistent with O in HO–BN, O–BN and LiO–BN environments. This assignment is supported by the close match between the number of O atoms (4.6 at%) and B atoms (4.8 at%) accounted for this and the corresponding B1s contribution (at 192.66 eV) assigned to B atoms in HO–BN, O–BN and LiO–BN moieties (see Table 3). A final O1s contribution, centered at 530.42 eV and accounting for ≲30% of the O1s signal, is close to the reported B.E. of aminobutyric acid (530.7 eV) and several metal oxides. Lacking evidence of the presence of metals other than boron and lithium, we attribute this peak to organic oxygen.

The Li1s spectral region, Fig. 2(e), can be described by a single, broad (FWHM – 2.89 eV) Gaussian–Lorentzian contribution centered at 55.7 eV. The Li1s signal spans reported B.E. values for several lithium compounds, indicated in the figure, including lithium carbonate (55.2 eV), oxide (55.6 eV), hydroxide (54.9 eV) and fluoride (55.7 eV). Similarly, the F1s region (not shown here because of its low concentration, 1 at% only) can be described by a single broad (FWHM = 2.88 eV) Gaussian–Lorentzian component centered at 685.38 eV, close to the reported 685.1 eV value for LiF. Note that there are only 1 at% LiF and 10 at% Li₂CO₃/Li₂O/LiOH, while the total amount of Li is 12 at%. This suggests that LiO–BN is about 1 at%. Here, the LiOH could be a remnant of the precursor used to functionalize the BN, and the Li₂O a product of its reaction to donate H to H–BN and BN–H; however, the origin of the fluorine is not clear at this point.

The C1s region, Fig. 2(f), contains two Gaussian–Lorentzian components, centered at 284.9 eV and 288.7 eV. The smaller of these two components (at 288.7 eV) accounts for only ∼1% of the atoms in the sample and spans the B.E. reported for lithium carbonate (289.8 eV). Although it is possible that a fraction of its intensity corresponds to Li₂CO₃, it is more likely that most of the C in the sample corresponds to adventitious atmospheric hydrocarbons and, potentially, N-containing organic products of its reaction with BN.

Fig. 3 shows ¹H, ⁷Li and ¹¹B solid state NMR data for the as-synthesized FBN sample. ¹H MAS NMR reveals the presence of at least four different proton peaks at around −2, 0.6, 2.8 and 6.2 ppm. The lower frequency peak at −2.0 ppm is due to the presence of LiOH, whereas 0.6 and 2.8 ppm peaks might be due to the presence of H–BN or HO–BN environments and NH–groups, respectively. Higher frequency proton peaks (6 ppm and above) are generally due to either hydrogen bonded OH-groups or the presence of protonated amines. ⁷Li MAS NMR shows one relatively broad peak at ∼0 ppm. Due to low and close chemical shift range of diamagnetic ⁷Li NMR peaks, different lithium bearing species (such as Li₂O, LiF, LiOH and LiO–BN) are expected to overlap and give one broad resonance.

Fig. 3 (a) ¹H, (b) ⁷Li, and (c) ¹¹B MAS NMR data for the as-synthesized FBN. The inset in (c) shows spectrum fit simulation for ¹¹B NMR. The calculated isotropic chemical shift (δ_iso-ppm) and quadrupolar coupling constant (C_Q MHz) values are as follows; site1: 1.4 ppm, 0.4 MHz, site2: 6.8 ppm, 0.2 MHz, site3: 18.3 ppm, 2.5 MHz and site4: 29.7 ppm, 2.8 MHz.

¹¹B MAS NMR data of FBN sample shows at least 4 different boron environments (Fig. 3(c)). ¹¹B is a quadrupolar nuclei and it can give rise to significant broadening and distortions of the NMR resonances of compounds with asymmetrical boron environments. Boron containing oxo-compounds typically contain tetrahedral BO₄ or trigonal BO₃ configurations. The former has a relatively narrow single resonance with δ_iso of 2 to −4 ppm and a C_Q of 0 to 0.5 MHz whereas the latter has larger C_Q of around 2.5 MHz and δ_iso values ranging from 12 to 19 ppm. For boron nitrides, the cubic phase (c-BN) generally shows a symmetric ¹¹B NMR resonance at around 1.6 ppm and h-BN gives a single quadrupolar lineshape peak with δ_iso of ∼30 ppm and C_Q of 2.8 MHz.^28,29

As seen in Fig. 3(c), the FBN sample shows at least two distorted boron environments with isotropic chemical shift of 18.3 and 29.7 ppm mostly due to the presence of 3-coordinated B–O and/or B–N environments. The presence of the two peaks at 1.4 and 6.8 ppm suggests the formation of 4-coordinate B–O or B–O–N environments as these peaks are not expected to be present in pristine h-BN sample and a cubic BN phase is unlikely. The relative intensities of these peaks are 100, 69.9, 40.6 and 57.6 at 29.7, 18.3, 6.8 and 1.4 ppm, respectively. This means that the 4-coordinated B has a relative amount of 36% of all boron. A trivial explanation of the 4-coordinated B would be the presence of glassy B₂O₃ which could be in an amorphous state and will not show up in the XRD pattern. However, if this 4-coordinated B was due to B₂O₃, it would have produced significant summation and overtone vibrations in the FTIR spectrum, between 2000 and 3000 cm⁻¹.³⁰ Since the FTIR spectrum is completely flat at this region, we have assigned this 4-coordinated B to O atoms attached to the B-s in the basal plane of h-BN, i.e., O–BN as revealed in the XPS analysis.

In summary, FTIR, XRD, XPS and NMR analyses have indicated unambiguously the functionalization of h-BN and the products of the functionalization are HO–BN, O–BN, LiO–BN and H–BN. However, about two-third h-BN do not appear to be functionalized even though its long-range order has disappeared (as evidenced by the disappearance of its crystalline peaks in the XRD pattern). A small amount of LiOH may remain unreacted in the sample. Furthermore, the XPS analysis reveals that the melt-solid reaction condition (500 °C) has resulted in some loss of LiOH, most likely due to the vaporization of LiOH during synthesis since the melting point of LiOH is only 462 °C.

SEM images of the as-synthesized FBN particles and the starting h-BN powder are shown in Fig. 4. Many particles of the as-synthesized FBN are agglomerated and the particle sizes range from 200 nm to 3 μm. This morphology is quite different from that of the starting h-BN powder which contains a large number of h-BN sheets with some particles. Further, the starting h-BN particles have larger sizes (∼2 to 20 μm) than that of the final product, suggesting that molten LiOH not only exfoliates h-BN, but also reacts with h-BN in the direction perpendicular to the basal plane, thereby breaking large h-BN sheets into multiple pieces with smaller basal planes. This is quite different from the work by Li, et al.¹⁸ who have formed transparent h-BN nanosheets (∼4 nm thick) through exfoliation of h-BN using molten NaOH/KOH mixture at 180 °C for 2 h. The exact reason for the different morphologies is not clear at this stage, but it is likely due to the higher temperature used in this study (500 °C in order to create molten LiOH). One of the possible mechanisms is that high temperature could result in several reactions more than just exfoliation and substantial agglomeration of FBN sheets, leading to the formation of small particles rather than nanosheets.

Fig. 4 SEM images of (a) and (b): the FBN particles and (c) and (d): the as-received h-BN powder. The gray background in (a) and (b) is the carbon tape and thus should be ignored. The typical morphology of FBN particles are highlighted with arrows. Different magnifications are used to highlight different sizes of the FBN and as-received h-BN powders.

The representative charge/discharge behavior of CB half cells is shown in Fig. 5(a). The open circuit voltage (OCV) is 2.55 V, and the cell is charged first to 3.3 V and then discharged to 1 V. The first discharge gives rise to two plateaus with one at ∼2.2 V and the other below 1.3 V. The specific capacity for the first discharge is large (∼123 mA h g⁻¹), but it decreases drastically to 38 mA h g⁻¹ in the second discharge and continues to decrease gradually beyond the second cycle. However, the capacity decrease becomes extremely small after the tenth cycle and the specific capacity stays at ∼28 mA h g⁻¹. The two plateaus in the first discharge are attributed to oxide impurities in CB which react with Li⁺ ions irreversibly. In a recent study on activated carbon³¹ we have found that activated carbon half cells with a lithium metal anode also exhibit such plateaus in the first discharge. The EDS analysis³¹ reveals that the main impurities in activated carbon are SiO₂, CaO, Al₂O₃ and FeO (see Table S1 in ESI†). All of these oxide impurities can be reduced by Li⁺ ions. For example, SiO₂ reacts with Li⁺ ion irreversibly to form Li₄SiO₄ alloy according to the reaction of 2SiO₂ + 4Li⁺ + 4e⁻ = Li₄SiO₄ + Si.³² After the first cycle the plateau near 2.2 V disappears completely (Fig. 5(a)), clearly indicating that this plateau is caused by irreversible reactions. In contrast, the plateau below 1.3 V is still present but with much smaller capacities in the subsequent cycles.

Fig. 5 Voltage profiles of (a) a CB half cell and (b) a FBN half cell, charged and discharged from 1.0 to 3.3 V at a current density of 28 mA g⁻¹ CB and FBN, respectively.

Fig. 5(b) shows the representative charge/discharge behavior of FBN half cells. The specific capacity in this figure is calculated through subtracting the total capacity of the entire electrode by the capacity derived from the given amount of CB in the electrode (e.g., 28 mA h g⁻¹ of 50 wt% CB in the electrode when the cell voltage is at 1.0 V). The resulting capacity after the subtraction is then divided by the weight of FBN in the electrode to obtain the specific capacity of FBN. Note that the first discharge profile has a plateau at ∼2.2 V which disappears in the following cycles. This plateau is ascribed to the effect of 50 wt% CB in the electrode because it has the same behavior as the one displayed by the CB half cell (Fig. 5(a)). Similarly, the second plateau below 1.6 V in the first cycle is partially due to CB, but it is also partially from some irreversible reactions of FBN because CB cannot account for all of the irreversible capacity in the first discharge. However, the most interesting phenomenon in Fig. 5(b) is that FBN can offer a specific capacity at ∼78 mA h g⁻¹ with a good cycle stability and coulombic efficiency (Fig. 6(a)).

Fig. 6 (a) The specific capacity and coulombic efficiency of the FBN half cell in Fig. 5(b) as a function of charge/discharge cycles, and (b) the differential capacity vs. voltage of the charge/discharge profiles of CB and FBN half cells shown in Fig. 5(a) and (b), respectively. The vertical dashed lines indicate the location where dQ/dV starts to deviate from constants.

To better understand the charge/discharge behavior of FBN, Fig. 5 has been re-plotted as the dQ/dV vs. V curves (where Q is the quantity of charge and V is the voltage), as shown in Fig. 6(b). It can be seen that dQ/dV is a constant between 1.85 and 2.80 V for FBN and between 1.60 and 2.80 V for CB, indicating that both half cells store charges through electrical double layer (EDL) mechanism in these voltage ranges.^33,34 However, redox reactions do take place below 1.85 V and 1.60 V for FBN and CB, respectively, because dQ/dV is not a constant below these respective voltages.

To further understand the nature of redox reactions, cyclic voltammetry (CV) tests were conducted. As shown in Fig. 7, there are two reduction reactions in the first scan, but the reduction reaction R1 at ∼2.0 V potential disappears in the subsequent scans, clearly indicating that R1 is not a reversible reaction. This behavior is in excellent agreement with charge/discharge behavior of CB half cells and FBN half cells shown in Fig. 5, and this reduction has been attributed to oxide impurities in CB. It is interesting to note that R2 reduction at ∼0.6 V potential is reversible since a corresponding oxidation peak O₂ is present. Furthermore, in the first scan R2 reduction contains some irreversible components because the current of R2 reduction decreases significantly in the subsequent scans. Again, these phenomena match the charge/discharge behavior of FBN half cells very well (Fig. 5(b)), unambiguously proving that redox reactions are present at the cell voltage below ∼1.8 V.

Fig. 7 CV curves of FBN half cells. Two reduction reactions are observed in the first scan, but the reduction reaction R1 disappears in the subsequent scans, whereas the reduction reaction R2 exhibits some reversibility.

One of the possible mechanisms for redox reactions of FBN is Li reacting with surface-bound oxygen radicals of O_x–BN, as shown in Fig. 8, with the electrochemical reaction shown below.

O_x–BN + 2xLi⁺ + 2xe⁻ → Li_2xO_x–BN (x ≤ 1) (1)

Fig. 8 A fragment of oxidized h-BN (O–BN) 3 × 3 supercell, as cut out from an infinite 2D monolayer. Color code: B – pink, N – blue, O – red, Li – magenta.

Our DFT calculations predict that this redox reaction gives an electrochemical potential of 2.08 V (vs. Li/Li⁺) for x = 1. In this reaction, the h-BN sheet may be viewed as a 2D monolayer that captures ·O· radicals and binds them to one B and one N atom through an epoxide bond on the surface. Therefore, the cell reaction may be simplified to

·O· + 2Li⁺ + 2e⁻ → 2Li⁺ + O²⁻ (2)

which is identical with the reaction of an oxygen radical with Li. As a result of this reaction, the N–O bond of the formerly mentioned epoxide will open up, while the B–O bond remains intact and the oxygen atom absorbs two electrons. This reaction is similar to that happening on the surface of graphene oxide when an epoxide O reacts with Li and Na, as reported in ref. 2 and 3. However, while the epoxide of graphene oxide can absorb only one electron without breaking off from the surface, FBN can theoretically absorb two electrons while still maintaining a single covalent bond to the surface. The reason for the absorption of the second electron in FBN is the one-electron deficiency of boron as compared to carbon, i.e. one electron absorbed on O is donated to B to form a charged B–O single covalent bond with a negative charge on the B atom, while the other electron absorbed on ·O· will complete the dangling bond on the other side of the B–O· unit assigning a single negative charge to the O. Both negative charges will be counterbalanced by Li⁺ ions diffusing close to the surface. If the captured radical is a hydroxyl (·OH), or a methoxyl (·OCH₃), or any other radical species with a single unpaired electron, only one electron will be absorbed during discharge, namely the one that completes the electron pair in the single covalent bond between the electron deficient B and the radical. In reality, the effects of the surface-captured radicals extend beyond the B atom: the hole, i.e. the missing electron from the B–O bond in (HO)_x–BN is redistributed in the environment of the B atom and leads to decreased electron density on the N atoms, as compared to pure h-BN,³⁵ therefore, during discharge some electron density will flow to the N atoms as well.

The proposed surface intercalation/conversion reactions of FBN-s are only distantly related to Li intercalation in bulk h-BN, even though electron absorption by electron deficient valence shells of B atoms happen in both processes. The intercalation of pure Li (without –O– and –OH groups) into bulk h-BN has been experimentally studied in the literature³⁶ and has been theoretically modeled as well.³⁷ However, the intercalation of pure Li into bulk h-BN is a very different process, as the presence of –O– and –OH groups in our case accelerates the exfoliation of h-BN, resulting in functionalized exfoliated monolayers. In contrast, pure Li intercalation between h-BN layers results in 3D crystals instead of exfoliated and functionalized monolayers. In spite of their difference, the absorption of electrons into the electron-deficient valence shell of B atoms of h-BN monolayers also happens even by electron donating pure Li and other alkali metals, e.g. when they are dissolved in 1,2-dimethoxyethane (DME) at room temperature.³⁸ In this case the alkali metal generates solvated electrons in the DME solvent (similarly to liquid ammonia), the solvated electrons are absorbed by the layers of bulk h-BN, and the negatively charged h-BN layers exfoliate due to electrostatic repulsion and will be surrounded by solvated alkali cations in the solution/suspension, as described recently in ref. 38. This latter experiment is another demonstration that h-BN monolayers are capable of absorbing electrons without decomposing, even though when functionalized, the electron absorption happens on sp³ hybridized B atoms, instead of sp² ones (the latter is the case for pure h-BN).

This charge storage mechanism is unique for h-BN and cannot be found in graphene, as the absorption of radicals on graphene leads to the rearrangement of double and single bonds and in general does not preserve the radical character of the absorbed species. Strong magnetism is expected to be characteristic for most radical functionalized h-BNs, and it has been detected in fluorinated h-BN.¹⁵

Since XPS and NMR analyses have indicated the presence of HO–BN in the as-synthesized FBN, we have also performed DFT calculations for hydroxyl (–OH) functionalization and found that this functionalization offers a greater electrochemical potential (∼2.97 V vs. Li/Li⁺) with the following redox reaction.

HO–BN + Li⁺ + e⁻→ Li(HO)–BN (3)

This reaction will provide a theoretical specific capacity of 550 mA h g⁻¹ and a specific energy of 1632 W h kg⁻¹ when coupled with a Li anode. However, experimentally no redox reactions at 2.97 V are observed. The discrepancy may be due to the significant agglomeration of the as-synthesized particles while the DFT calculations assure isolated monolayers.

Experimentally, we have only observed sloping charge/discharge curves from 1.85 to 1.0 V (Fig. 5), which is closer to the predicted voltage of 2.08 V for the redox reaction of O_xBN indicated in eqn (1). The lacking of a voltage plateau at 2.08 V may be attributed to several possible mechanisms. These are (i) multiple surface functional groups (e.g., –O and =O), (ii) different patterns of –O and =O functional groups on the surface of h-BN, (iii) partial surface functionalization, (iv) agglomeration of FBN particles, and (v) impurities in the original h-BN. All of these can result in redox reactions at the surface of FBN particles with different electrochemical potentials and thus sloping voltage profiles in the charge/discharge cycles. This is consistent with many recent studies on carbon and graphene with O- and N-containing functional groups^2,3,39–41 which have shown sloping voltage profiles with a wide voltage range, e.g., from 0.01 to 2.0 V^39–41 or from 1.0 to 4.0 V,^2,3 for Li-ion and Na-ion reactions at the surface of the functionalized carbon and graphene.

It should be pointed out that when the charge/discharge voltage range is expanded from 1.0 to 3.3 V to from 0.4 to 3.3 V, the reversible discharge specific capacity of the FBN half cell is increased from ∼78 mA h g⁻¹ to ∼145 mA h g⁻¹ with good cycle stability and coulombic efficiency (Fig. 9). If the discharge voltage is further reduced to end at 0.01 V, the reversible discharge specific capacity is increased further to ∼400 mA h g⁻¹. This value is compatible with the specific capacity of the state-of-the-art graphite anode (∼350 mA h g⁻¹). Fig. 10(a) shows how the voltage profile varies with the charge/discharge voltage window for the 4th charge/discharge cycle, whereas Fig. 10(b) compares the discharge specific capacities of the 3rd and 4th cycles for three charge/discharge voltage windows. It is very clear from Fig. 9 that the increased specific capacity for the charge/discharge voltage window from 0.4 to 3.3 V is due to redox reactions. This is also the case for the charge/discharge voltage window from 0.01 to 3.0 V (Fig. 10). These phenomena indicate that the FBN synthesized from the melt-solid reaction has the potential for anode applications if nanosheets with little agglomeration can be produced.

Fig. 9 (a) Voltage profiles of a FBN half cell charged and discharged from 0.4 to 3.3 V at a current density of 27 mA g⁻¹ FBN, and (b) its specific capacity and coulombic efficiency as a function of charge/discharge cycles.

Fig. 10 (a) Voltage profiles of the FBN half cells at the 4^th cycle with different charge/discharge voltage windows, and (b) comparisons in the discharge specific capacity of the 3^rd and 4^th cycles among three charge/discharge voltage windows. As indicated, the charge/discharge from 0.01 to 3.0 V give the highest specific capacity (∼400 mA h g⁻¹).

It is worthy of mentioning that the valences of atoms in eqn (3) are as follows: B + 3, N − 3, C + 4, H + 1, and Li + 1 while the O valence changes from −1 to −2 during lithiation in both equations. These are idealized valences that do not take into account that the change of the O valence is partially distributed over the N atoms, as discussed above. The simplified valences are based on a model in which the reversible reduction of hydroxyl radicals happens without the involvement of the atoms of the h-BN substrate to which these radicals are covalently bound. This charge storage mechanism is also supported by the fact that both hydroxyl radical^9–11 and hydroxide ion¹² functionalization of h-BN have been described in the literature. In the case of the hydroxyl radical the oxygen valence is −1, whereas the oxygen valence is −2 in the case of the hydroxide ion.

5. Conclusions

The feasibility of synthesizing FBN via the reaction between molten LiOH and solid h-BN has been demonstrated for the first time. It is found that the melt-solid reaction leads to exfoliation and functionalization of h-BN simultaneously. The FBN products are HO–BN, O–BN, LiO–BN and H–BN; however, these products are agglomerated particles rather than nanosheets. These agglomerated FBN particles can charge/discharge reversibly through surface reactions with Li ions and exhibit a specific capacity of 78 mA h g⁻¹ and 400 mA h g⁻¹ with good cycle stability and coulombic efficiency when the discharge voltage ends at 1.0 and 0.01 V, respectively. DFT calculations suggest that one possible redox reaction for the FBN in the present study is Li reacting with surface-bound oxygen radicals of O–BN. This redox reaction can offer an electrochemical potential of 2.08 V (vs. Li/Li⁺). DFT calculations also predict that HO–BN can provides an electrochemical potential of 2.97 V; however, this redox reaction is not observed experimentally, which may be due to the significant agglomeration of the product particles. This first ever evaluation of FBNs as electrodes for Li-ion batteries, through an integrated experimental and theoretical modeling approach, shows that FBNs can indeed store charges through redox reactions. This study also reveals that it is necessary in the future to optimize the melt-solid reaction or to synthesize FBNs through wet chemical methods to produce non-agglomerated O–BN, LiO–BN, and HO–BN nanosheets with high surface area for further evaluation and to achieve the desired electrochemical potentials for anode applications in Li-ion batteries.

References

1. J. M. Tarascon and M. Armand, Nature, 2001, 414, 359–367.
2. C. Liu, A. Zhamu, D. Neff and B. Z. Jang, 2010, US Pat. #: 8,795,899.
3. H. Kim, Y. U. Park, K. Y. Park, H. D. Lim, J. Hong and K. Kang, Nano Energy, 2014, 4, 97–104.
4. K. Nemeth, Int. J. Quantum Chem., 2014, 114, 1031–1035.
5. K. Nemeth, Functionalized boron nitride materials as electroactive species in electrochemical energy storage devices, WO/2015/006161, priority date 2013.
6. L. H. Li, J. Cervenka, K. Watanabe, T. Taniguchi and Y. Chen, ACS Nano, 2014, 8, 1457–1462.
7. D. Krishnan, F. Kim, J. Luo, R. Cruz-Silva, L. J. Cote, H. D. Jang and J. Huang, Nano Today, 2012, 7, 137–152.
8. H. Li, R. Y. Tay, S. H. Tsang, W. Liu and E. H. T. Teo, Electrochim. Acta, 2015, 166, 197–205.
9. V. E. Fedorov, A. S. Nazarov and V. N. Demin, Method of producing soluble hexagonal boron nitride, Russian Patent #: 2,478,077, 1996.
10. A. S. Nazarov, V. N. Demin, E. D. Grayfer, A. I. Bulavchenko, A. T. Arymbaeva, H. J. Shin, J. Y. Choi and V. E. Fedorov, Chem.–Asian J., 2012, 7, 554–560.
11. T. Sainsbury, A. Satti, P. May, Z. Wang, I. McGovern, Y. K. Gunko and J. Coleman, J. Am. Chem. Soc., 2012, 134, 18758–18771.
12. D. Lee, B. Lee, K. H. Park, H. J. Ryu, S. Jeon and S. H. Hong, Nano Lett., 2015, 15, 1238–1244.
13. T. Ikuno, T. Sainsbury, D. Okawa, J. M. J. Frechet and A. Zettl, Solid State Commun., 2007, 142, 643–646.
14. T. Y. Sainsbury, A. O'Neill, M. K. Passarelli, M. Seraffon, D. Gohil, S. Gnaniah, S. J. Spencer, A. Rae and J. N. Coleman, Chem. Mater., 2014, 26, 7039–7050.
15. M. Du, X. Li, A. Wang, Y. Wu, X. Hao and M. Zhao, Angew. Chem., Int. Ed., 2014, 53, 3645–3649.
16. G. P. Rajendran, Surface Modified hexagonal Boron Nitride particles, US 8258346, 2009.
17. Y. Kumashiro, Electric refractory materials, CRC Press, Boca Raton, FL, USA, 2000, p. 530.
18. X. Li, X. Hao, M. Zhao, Y. Wu, J. Yang, Y. Tian and G. Qian, Adv. Mater., 2013, 25, 2200–2204.
19. Y. Lin, T. V. Williams and J. W. Connell, J. Phys. Chem. Lett., 2010, 1, 277–283.
20. Z. Gao, C. Zhi, Y. Bando, D. Golberg and T. Serizawa, Nanobiomedicine, 2014, 1, 7,  DOI:10.5772/600001.
21. P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. D. Corso, S. Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari and R. M. Wentzcovitch, J. Phys Condens. Matter, 2009, 21, 395502.
22. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865.
23. J. P. Perdew, A. Ruzsinszky, G. I. Csonka, O. A. Vydrov, G. E. Scuseria, L. A. Constantin, X. Zhou and K. Burke, Phys. Rev. Lett., 2008, 100, 136406.
24. K. Nemeth, J. Chem. Phys., 2014, 141, 054711.
25. R. Geick, C. H. Perry and G. Rupprecht, Phys. Rev., 1966, 146, 543.
26. P. Thangasamy and M. Sathish, CrystEngComm, 2015, 17, 5895–5899.
27. A. V. Naumkin, A. Kraut-Vass, S. W. Gaarenstroom and C. J. Powell, NIST X-ray photoelectron spectroscopy database, Version 4.1, http://srdata.nist.gov/xps/Default.aspx.
28. K. J. D. Mackenzie and M. E. Smith, Multinuclear solid-state NMR of inorganic materials, Pergamon Material Series, Pergamon, 2002, p. 6.
29. P. S. Marchetti, D. K. Kwon, W. R. Schmidt, L. V. Interrante and G. E. Maciel, Chem. Mater., 1991, 3, 482–486.
30. J. L. Parsons and M. E. Milberg, J. Am. Ceram. Soc., 1960, 43, 326.
31. C. Liu, L. Li and L. Shaw, unpublished results.
32. H. C. Tao, X. L. Yang, L. L. Zhang and S. B. Ni, Ionics, 2014, 20, 1–6.
33. B. E. Conway, Electrochemical supercapacitors: scientific fundamentals and technological applications, Kluwer Academic: New York, NY, USA, 1999.
34. C. Liu, Z. Yu, D. Neff, A. Zhamu and B. Z. Jang, Nano Lett., 2010, 10, 4863–4868.
35. H. M. Wang, Y. J. Liu, H. X. Wang, J. X. Zhao, Q. H. Cai and X. Z. Wang, J. Mol. Model., 2013, 19, 5143–5152.
36. A. Sumiyoshi, H. Hyoudou and K. Kimura, J. Phys. Chem. Solids, 2010, 71, 569–571.
37. B. Altintas, C. Parlak, C. Bozkurt and R. Eryigit, Eur. Phys. J. B, 2011, 79, 301–312.
38. H. Feng, Z. Hu and X. Liu, Chem. Commun., 2015, 51, 10961–10964.
39. Y. X. Wang, S. L. Chou, H. K. Liu and S. X. Dou, Carbon, 2013, 57, 202–208.
40. W. H. Shin, H. M. Jeong, B. G. Kim, J. K. Kang and J. W. Choi, Nano Lett., 2012, 12, 2283–2288.
41. Z. Wang, L. Qie, L. Yuan, W. Zhang, X. Hu and Y. Huang, Carbon, 2013, 55, 328–334.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra03141b

---