Among hydrogen storage materials, activated carbons, graphene, carbon nanotubes, and metal-organic frameworks are of great interest due to a combination of large surface area, low mass density, and high structural stability [
1]. However, the typical H
2 binding energy of these high-surface-area carbon materials is weak (–5.1 kJ/mol H
2) [
2]. The weak binding energy is attributed to physi-sorption dominated by van der Waals attraction, such that adsorbed molecules of H
2 do not remain attached at ambient temperatures and require liquid nitrogen temperatures to achieve adsorption [
3,
4]. For practical applications, such as H
2 fuel–cell electric vehicles, the operating temperatures are much higher, ranging from −40 to 85 °C, and require binding energies on the order of 15 to 25 kJ/mol for pressures ranging from 5 to 100 bar [
5,
6]. Unsaturated metal centers (e.g., Mg and Ni in metal–organic frameworks) provide an effective approach for enhancing the strength of binding between H
2 and nanostructured porous materials [
7,
8]. Other approaches have involved templating porous carbons that can provide impressive H
2 sorption properties and gravimetric and volumetric H
2 storage capacities of approximately 5 wt% and 35 g/L, respectively, at 77 K and 20 bar [
9]. Confinement effects have further been investigated in H
2 binding in potassium-graphite-intercalation compound KC
24, in which H
2 is found to adsorb with an enthalpy of 8.4 kJ/mol H
2 between layers of graphite with an interplanar spacing of 3.35 to 5.4 Å [
10]. Similarly, graphite nanofibers have been studied as potential materials with enhanced storage capacity due to their ability to capture H
2 molecules between layers, being estimated to be ~3.4 Å apart. Ahn et al. [
11] reported that the performance of graphite nanofibers is not superior to other activated carbon materials, even though other reports have given ranges of storage between ~3 wt% and up to 13 wt% [
12,
13,
14]. Calculations of H
2 adsorption on slit pore models of graphite nanofibers by Wang et al. [
15,
16] also suggest that although the pores show increased adsorption relative to modeled single-wall carbon nanotubes, slit pores with a distance of 9 Å had the highest adsorption capacity. In addition, doping carbons with electron–deficient boron (B) has been investigated because of the lower mass of B compared to metal. Substituting B for a metal provides a potential approach to meeting gravimetric density targets for storing hydrogen onboard a fuel–cell electric vehicle [
17]. In a pioneering computational study [
18], doping fullerene C
36 with B was predicted to lead to a significant increase in the hydrogen adsorption energy to −19.2 kJ/mol H
2. The authors of that study suggested that the increase in binding energy is due to a partial charge transfer from the sigma-bond of an H
2 molecule to the free localized
pz–orbital of a B atom. However, a notable feature is the intrinsic non-planarity of the fullerene that induces distortion of the B center, allowing for better matching of the orbitals for enhanced interactions. In contrast, Zhou et al. [
19] employed periodic one-dimensional calculations on carbon nanotubes using density functional theory (DFT) with a generalized gradient approximation (GGA) functional and supercells that range from 64 to 80 atoms and reported H
2 binding and pristine carbon material binding energies that ranged between −1.8 and −10.3 kJ/mol H
2, depending on the position of the H
2. Furthermore, they observed that B doping reduces the binding energy by between 0.7 and 3.6 kJ/mol H
2 and suggested that doping with electron–deficient B disrupts the conjugated six–membered ring electronic system. They also suggested that the polarity induced in the nanotubes by doping might play a role in the adsorption of the H
2 molecule, resulting in decreased binding energy. More recently, a theoretical study on the adsorption of various gas molecules on graphene that was doped with B [
20] reported energies of adsorption of −1.35 kJ/mol H
2 using GGA, acknowledging that, for non-covalent interactions, the consequent neglect of long–range van der Waals (vdW) interactions, does not accurately describe the physi-sorption of H
2. Computational work by Sha et al. on H
2 storage in bulk BC
3 material shows significant overestimation of the binding energies and spontaneous dissociation of H
2 when local density approximation (LDA) functionals were used [
21]. It has been established that LDA and GGA functionals are insufficient for capturing the vdW interactions of H
2 with doped carbon materials [
19,
20]. These shortcomings have led to the development of the vdW–DF correlation functionals (vdW–DF1 [
22,
23] and vdW–DF2 [
24]), which are parameterized to capture the vdW forces without any empirical corrections to the nonlocal dispersion interactions for a variety of systems, including graphene along with other carbon–based polymeric materials. Here, the nonlocal correlations are taken into account through the density–density interaction term [
22]. Klimes et al. [
25] showed that the accuracy of these functionals can be dramatically improved in dispersion bonded complexes. Benchmark studies of the vdW–DF2 functional with revPBE [
26] exchange on H
2 adsorption on coronene and graphene surfaces show physi-sorption energies for H
2 on the hollow site of coronene (–5.7 kJ/mol H
2) and graphene surfaces (–5.9 kJ/mol H
2) to be in good agreement with experimental values [
27]. A benchmark investigation of the physi-sorption of H
2 on pure benzene and graphene shows that the revPBE based vdW–DF2 method with nonlocal dispersion is in good agreement with high–level coupled cluster (CCSD(T)) calculations and experimental values for both benzene and graphene [
28].
Being encouraged by theoretical predictions of enhanced binding of H
2 to B-doped carbon materials, several experimental studies have been undertaken to dope these materials with boron. Some of these experimental approaches targeted the increased heteroatom substitution, which is the nature of the dopant in frameworks with high surface area and tunable porosity. Wang et al. [
29] used the substitution reactions to prepare B-doped microporous carbons. They reported that B doping led to an increase in the surface area and the isosteric heat of adsorption [Q
st] for H
2 compared to the undoped carbon (i.e., an increase from ca. −6 kJ/mol to −10 kJ/mol at higher coverage). Chung et al. [
30] developed a new class of polymeric B-containing precursors that were thermally transformed to microporous B/C materials with a boron content of 7.2% and surface area of 780 m
2/g. The substitutional
p–type B dopant was shown to polarize the C surface, resulting in higher Q
st of −10.8 kJ/mol for 0.62 wt % at temperatures of 77 to 87 K and a pressure of 1.2 bar for H
2 binding. Tour and coworkers [
31] reported a bottom-up, solution–phase technique for the preparation of pristine and heteroatom-substituted carbon scaffolds that show high surface areas and enhanced H
2 physi-sorption capacities relative to the pure carbon scaffolds. This approach involved heating chorine-containing small organic molecules (precursors) with metallic sodium at reflux in high-boiling solvents, being subsequently followed by the addition of the heterotopic electrophiles to the mixture for dopant incorporation. The substituted carbon scaffolds enriched with 7.3% boron were observed to have higher surface area of around 900 m
2/g and enhanced reversible hydrogen physi-sorption capacity (–8.6 kJ/mol at zero coverage). Jeong et al. [
32] developed a method for preparing a broad range of porous B-doped carbon materials by using B precursors containing inorganic additives. They reported the chemical structure of the doped sheets changed from a disordered (less π–conjugated) state with a B-puckered configuration at 600 to 800 °C to an ordered (highly π–conjugated) state with a planar configuration at 1500 °C. The planar graphitic layers were reported to accommodate <3% of B content, while the amorphous, BC
x-like materials exhibited much higher surface areas (500 to 800 m
2/g) and 12% B content with electron–deficient B moieties, resulting in enhanced H
2 binding energies (–12 to −20 kJ/mol at higher coverage). In a recent joint experimental and computational effort [
33], a site–specific selective incorporation of B into the graphene sublattices was demonstrated to form a tunable band gap controlled by the dopant concentration. Shcherban et al. [
34] reported a nanocasting approach to prepare B-doped CMK-3 using SBA-15 as an exo-template. They reported no increase in H
2 binding energy as a result of incorporating B and a decrease in the overall capacity of H
2 adsorption. In 2012, Bult et al. [
35] reported enhanced hydrogen binding energy for a high-surface-area, B-doped carbon material made by chemical vapor deposition of a B-doped carbon (BCx) layer onto CM-Tec activated carbon. The enhancement was attributed to increased interlayer spacing of the material (~4 Å) and higher concentration of B sites.
Table 1 provides a summary of the previously measured and calculated binding energies.