Progress on First-Principles-Based Materials Design for Hydrogen Storage

ﬁ Progress on rst-principles-based materials design for INAUGURAL ARTICLE hydrogen storage

Noejung Parka, Keunsu Choib, Jeongwoon Hwangb, Dong Wook Kimc, Dong Ok Kimc, and Jisoon Ihmb,1

aInterdisciplinary School of Green Energy, Low Dimensional Carbon Materials Center, Ulsan National Institute of Science and Technology, Ulsan 689-798, Korea; bDepartment of Physics and Astronomy, Seoul National University, Seoul 151-747, Korea; and cHanwha Chemical R&D Center, Daejeon 305-804, Korea

This contribution is part of the special series of Inaugural Articles by members of the National Academy of Sciences elected in 2011.

Contributed by Jisoon Ihm, October 9, 2012 (sent for review July 30, 2012) This article briefly summarizes the research activities in the field of vehicles and limited forms of zero-emission cars, larger-scale hydrogen storage in sorbent materials and reports our recent works energy operations such as stationary back-up power systems and and future directions for the design of such materials. Distinct full commercial scale automotive applications are very difficult to features of sorption-based hydrogen storage methods are described realize through the pure electric power storage devices. Hydro- compared with metal hydrides and complex chemical hydrides. We gen storage systems have obvious advantages in certain important classify the studies of hydrogen sorbent materials in terms of two aspects (e.g., energy density and durability), and the search for key technical issues: (i) constructing stable framework structures safe and efficient hydrogen storage materials continues for such with high porosity, and (ii) increasing the binding affinity of hydro- wide application purposes. gen molecules to surfaces beyond the usual van der Waals interaction. The recent development of reticular chemistry is summarized as Hydrogen Storage Methods and Sorption-Based Storage a means for addressing the first issue. Theoretical studies focus Systems mainly on the second issue and can be grouped into three classes Methods for hydrogen storage may be divided into two categories: according to the underlying interaction mechanism: electrostatic physical confinement inside vessels and binding onto host mate- interactions based on alkaline cations, Kubas interactions with open rials. The physical confinement of hydrogen gas involves either SCIENCES transition metals, and orbital interactions involving Ca and other high-pressure compression or cryogenic liquefaction and has been nontransitional metals. Hierarchical computational methods to en- used for special purposes or on the laboratory scale. The safety (for APPLIED PHYSICAL able the theoretical predictions are explained, from ab initio studies high-pressure) and cost (for liquefaction) concerns associated with to molecular dynamics simulations using force field parameters. We such methods suggest that large-scale commercial applications also discuss the actual delivery amount of stored hydrogen, which of stored hydrogen might be unlikely via such physical storage depends on the charging and discharging conditions. The usefulness methods. Regarding the method to bind hydrogen onto an absor- and practical significance of the hydrogen spillover mechanism in bent medium, a variety of host materials have been investigated (3, increasing the storage capacity are presented as well. 5). These materials may be classified into three distinct groups depending on the microscopic mechanism related to the charging Renewable Energy and the Significance of Hydrogen and discharging properties: metal hydrides that contain atomic Storage hydrogen as a constituent of the bulk material, complex chemical he prosperity of modern civilization is inextricably based on en- hydrides that react chemically upon charging and discharging, and Tergy supply networks (on-grid or off-grid) that deliver petroleum, the adsorption of molecular hydrogen onto the surfaces of highly natural gas, and electricity to residential, public, commercial, and porous materials. Each method has its own advantages and dis- industrial facilities, as well as transport systems. Concerns are grow- advantages. For example, metal hydride systems (in particular the ing over the limited availability of natural resources and the envi- AB5-type heavy metal alloys) can provide greatly satisfactory vol- ronmentally hazardous byproducts of the energy conversion process, umetric storage density, easy kinetics, and good reversibility, but such as greenhouse gases. From this perspective, renewable energy the gravimetric density is fundamentally limited (6). The re- harvesting technologies based on natural energy flows, including versibility and reaction kinetics present the largest hurdle to solar power, wind power, geothermal energy, and tidal energy, are complex chemical hydride storage systems. In a sense, the com- increasingly gaining momentum. The storage of the harvested energy plex chemical hydride can perhaps be used with a compartment system that is subject to off-board regeneration processes (4, 7). remains a fundamentally important factor for the utilization of the A marked drawback of sorption-based storage in porous mate- renewable energy because the sources of the renewable energy usu- rials is the weak strength with which hydrogen is physisorbed ally undergo significant spatial and temporal variations (1). The issue onto surfaces. Without a particular enhancement of the ad- of hydrogen storage can be understood in this context. Among sorption strength, sorption-based storage systems are practically chemical fuels, pure hydrogen (i.e., in the gas form) has the highest operative only around the liquid nitrogen temperature (77 K). mass density but the poorest volumetric density (2, 3). As a result, Nevertheless, sorbent materials and sorption-based storage sys- hydrogen storage research traces a long history of studies toward the tems deserve a more intensive investigation because they present reversible condensation of hydrogen into a limited volume using a variety of inherent advantages over other storage systems. Mi- lightweight devices. In recent decades a globally recognized objective croscopically, adsorption does not require the dissociation of is the development of a stored hydrogen carrier that can power hydrogen molecules and thus does not involve energy barriers. vehicles through fuel cells (or, perhaps, combustion engines). For Macroscopically, this approach permits the rapid charging example, the guideline published by the US Department of Energy and discharging of hydrogen, and the corresponding thermal states that, for a commercially competitive vehicle, hydrogen storage systems need to achieve an overall capacity of 5.5 wt% hydrogen with

a volumetric ratio of 40 g/L within a few years (4). Author contributions: D.W.K., D.O.K., and J.I. designed research; N.P., K.C., J.H., D.W.K., However, the objectives of hydrogen storage need to be defined D.O.K., and J.I. performed research; N.P., D.W.K., and D.O.K. contributed new reagents/ from a wider perspective of the energy storage and energy back-up analytic tools; N.P., K.C., and J.H. analyzed data; and N.P., K.C., and J.I. wrote the paper. system. Although supercapacitors and rechargeable batteries The authors declare no conflict of interest. have shown an optimistic outlook in the field of hybrid electric 1To whom correspondence should be addressed. E-mail: [email protected].

www.pnas.org/cgi/doi/10.1073/pnas.1217137109 PNAS | December 4, 2012 | vol. 109 | no. 49 | 19893–19899 Downloaded by guest on September 25, 2021 management is expected to be easier than chemical or metal various MOFs and ZIFs. An extensive review of them is provided hydrides (8). A fundamental requirement for sorbent materials is in refs. 11 and 23. Furukawa et al. (10) described the inclusion of the presence of a stable framework structure with sufficient po- a long and slim linker as conveying ultrahigh porosity. They recently rosity and surface area. In this respect, recent experimental ad- reported the preparation of an extremely porous structure, named vancement in the field of reticular chemistry, that is, metal– MOF-210, which shows an extremely high specific surface area. In organic frameworks (MOFs) and covalent organic frameworks terms of the Brunauer–Emmett–Teller area, MOF-210 reaches the (COFs), is particularly noteworthy (9–11). On the other hand, highest value so far (6,240 m2/g), which exceeds that of other many theory-based studies have addressed the issues of increasing members of the family of MOFs [e.g., MOF-5 (3,800 m2/g) and binding affinity of hydrogen molecules onto surfaces by in- MOF-177 (5,640 m2/g)]. Its excess and total hydrogen uptakes (86 troducing various strategies, as will be described in more detail in mg/g and 176 mg/g at 77 K and 80 bar, respectively) are higher than later sections. those of other porous crystals (10). These results showed that an increase in the specific surface area of a porous material directly Computational Methods Applied to Hydrogen Storage benefits the uptake of hydrogen storage at low temperatures, and Problems also the storage of heavier gases like CO2 or methane. The gov- Several hierarchical computational methods have been used in this erning thermodynamics associated with the gas uptake by MOFs is field. An accurate evaluation of the binding strength between hy- derived from the physisorption of gas molecules onto a chemically drogen molecules and the host material surface is central to this inert MOF surfaces. The polarity of a building unit and the ap- study. Among various ab initio theories, the density functional propriate spatial separation between the organic units can partially theory (DFT) has been most widely used. The Kohn–Sham equa- increase the binding strength of hydrogen molecules to some degree tion, which is an effective single-particle equation, is usually solved (24). However, these binding strengths still fall in the regime of with the local one-body operators that approximate the electron– physisorption and are on the order of several kJ/mol (a few tens of electron many-body effects (12). Commonly made choices are meV). Thus, these binding strengths are not relevant for hydrogen the form of a local density approximation or a generalized gradient storage at room temperatures. approximation (13, 14). Such local approximations of the exchange–correlation potentials can have remnant (surplus) electron Decoration of Metal Atoms onto Porous Structures self-interactions and occasionally require a cautious treatment. The The success of low-temperature hydrogen storage methods using over-delocalization problems of the local functionals can arise in microporous materials, as stated in previous paragraphs, can lead to some cases of adsorptions (15). In this regard, effects of the exact a certain utility (10). Nevertheless, higher operational temperatures exchange and the explicit inclusion of the Hubbard U term deserve for the hydrogen storage systems would unarguably convey a much detailed theoretical considerations (16, 17). Accurate and practical wider impact, for example by permitting vehicular applications un- calculations of van der Waals (vdW)-type dispersion interactions der ambient conditions. For such a purpose, a large specificsurface are important for calculating hydrogen–surface interactions (18). area in a sorbent material must be accompanied by an enhanced The direct treatment of such interactions using the Schrödinger binding affinity of the surface to hydrogen molecules. As discussed equation is impractical; therefore, wavefunction theories based on below, the interactions between open transition metal atoms or al- the Hartree–Fock (HF) self-consistent field (SCF) approximation kaline cations with hydrogen molecules can yield a desirable binding provide a reference for or complement to the DFT results. For strength. In an effort to use such an advantage, a few postsynthetic example, the Moller–Plesset second-order perturbation applied to treatments have been considered to incorporate alkali, alkaline the HF-SCF wavefunctions provides a reasonably practical and earth, and transition metal atoms onto the organic linkers of MOFs accurate approach (19). or COFs (25–27). Transition metal atoms have a strong tendency to More realistic adsorption isotherms may be calculated using aggregate, whereas alkaline cations tend to oxidize. Therefore, the molecular dynamics simulations at given pressures and temper- preparation of dispersed under-coordinated metal elements for the atures. Such studies require transferrable force parameters that adsorption sites in the porous material accessible to gas molecules account for the chemical environment, such as the reactive force remains an extremely challenging task in synthetic chemistry (28). field. If reliable force parameters are available, a grand canonical Instead of decoration with metal elements, as a somewhat different Monte Carlo (GCMC) simulation is particularly useful for cal- approach, people considered modification of porous structures with culating hydrogen adsorption isotherms associated with a porous embedded metal nanoparticles (29–31). In such cases, chemisorp- material (20). Porous materials are characterized by the surface tion into the destabilized nanoparticles and adsorption onto surfa- area and the free pore volume. Experimental storage measure- ces can lead to a synergic increase of storage capacity. ments, either through gravimetric or volumetric methods, typi- Focusing on methods of surface treatment, we explored mod- cally count only the so-called excess amount of hydrogen uptake. ifications of framework structures that help anchor additional However, the hydrogen gas present in the free volume of the metal atoms. In our previous publications, we demonstrated that pores should be included in calculating the total amount of hy- boron substitution in the organic building units of COFs sub- drogen. GCMC calculations provide a practical tool for a direct stantially improved the stabilization of Sc, Ti, and Ca coordinated evaluation of the total as well as excess amounts of stored hy- to the linker moieties (32). We also searched for functional groups drogen in a porous material (21). capable of preferential binding to each transition metal atom so as to prevent the aggregation. In particular, the presence of OH Porous Framework Structures terminal groups on the MOF linkers was found to be efficient in Solid materials intended for gas uptake must possess a high specific stabilizing transition metal atoms (33). Our ab initio calculations surface area and porosity. Microporous and mesoporous materials, showed that the under-coordinated Fe(II) on the OH terminals including zeolites and silica derivatives, have been developed could bind up to four H2 molecules, and the transition from a high throughout the history of materials sciences with a variety of spin state to a low spin state took place upon adsorption of H2 applications, such as gas sorbents, gas separators, and catalysts (22). molecules. Such transition in the spin state obviously indicates the Recent advances in the reticular chemistry of MOFs and related orbital interaction between the 3d states and H2 molecular orbitals. materials, including zeolitic imidazolate frameworks (ZIFs) and Other groups suggested that the decoration of Li on the benzenyl COFs, are relevant to the design and synthesis of materials with rings of MOF carboxylate linkers could lead to a substantial in- a tunable pore size and chemical functionality derived from selected crease in the H2 uptake capacity near room temperature (34–37). building units (9). The reversible hydrogen storage, CO2 uptake, This effect was attributed to the strong electrostatic field centered and methane uptake are used to benchmark the performances of at the cation site.

19894 | www.pnas.org/cgi/doi/10.1073/pnas.1217137109 Park et al. Downloaded by guest on September 25, 2021 Delivery Amount

Before we examine the detailed interaction mechanism between INAUGURAL ARTICLE hydrogen molecules and these surface-modified structures, we want to precisely define the target value for the binding energy in the context of the actual vehicle-operating conditions. In this sit- uation, the volumetric and gravimetric storage capacity should be assessed in terms of the reversibly deliverable amount rather than the maximum hydrogen uptake. The delivery amount is defined as the difference in the amount of hydrogen present under the charging and discharging conditions. Therefore, the delivery amount, also called the usable amount, depends on the macro- scopic charging and discharging conditions [pressure (P) and temperature (T)], as well as on the microscopic binding characteristics (such as available binding sites and energies). For safety reasons, the maximum delivery pressure is recommended to be lower than 100 bar (4). It is also strongly desirable that the discharging temperature (the condition for hydrogen desorption) is not more than 100 °C. From equilibrium thermodynamics, the expression for g (the average number of adsorbed hydrogen molecules per adsorption site) is (38): X ∂ ðμ−ε Þ= = n n kT ; [1] g kT ∂μ ln gne n=0 Fig. 1. Atomic configurations of 4H2 molecules attached to the backbone structure. The backbone is (A) Fe-decorated, OH-functionalized terphenyl dicarboxylate (TPDC), and (B) V-decorated, OH-functionalized TPDC, re- where μ is the chemical potential for H2 (common to both the gas and the adsorbed phase at the thermal equilibrium) at given P and spectively, where TPDC is the linker part of the IRMOF16. The blue, gray, red, SCIENCES k ε green, and orange balls stand for hydrogen, carbon, oxygen, iron, and va- T, is the Boltzmann constant, n is the adsorption energy per H2 nadium atoms, respectively. APPLIED PHYSICAL molecule when the number of adsorbed molecules is n,andgn is the configurational degeneracy factor for given n. (In practice, the

sum in the grand partition function is usually dominated by a single namely, εn = −0.523, −0.517, −0.500, and −0.435 eV for n = 1, 2, 3, term. In such cases, the gn in the numerator and the denominator is and 4, respectively. We have g = 3.861 and 2.930 for charging and fi effectively cancelled, and the effect of con gurational degeneracy discharging conditions, and the delivery amount is only 0.931 per turns out to be minor.) Suppose that we use the isoreticular MOF site, mere 23% of Fe. 16 (IRMOF16) as the storage system. To increase the storage The above results are the contribution from the relatively strong capacity, the linker is functionalized with OH and decorated with adsorption to transition metal sites. To include the contribution transitional metal atoms such as Fe or V. At the fuel charging (298 from the vdW interaction between H and the MOF, as well as that K and 100 atm) and discharging (373 K and 3 atm) conditions, μ = 2 − μ = − from the gas-phase storage in the void (pore) volume of the MOF, 0.196 eV and 0.386 eV, respectively (39). If we choose Fe we have performed GCMC simulations for the Fe-decorated, OH- (II) for the decorating metal, the calculated binding energies of H2 ε = − − − − functionalized IRMOF16 with four Fe sites per linker (33). In- to the Fe site are n 0.127, 0.246, 0.238, and 0.259eV for cluding all of the contributions, we get the reversibly usable storage n = 1, 2, 3, and 4, respectively. Fig. 1A shows the configuration of capacity of 6.0 wt% (4.3 wt% from the strong adsorption at Fe sites). the linker (terphenyl dicarboxylate) part of our system when the Some authors considered the delivery amount under different maximum number (n = 4) of H molecules are attached. (In max 2 charging/discharging conditions. Han et al. (36) performed GCMC reality, we considered four metal atom sites per linker to increase simulations using the force parameters obtained from ab initio the storage capacity, but the case of only one metal site is shown works and considered the charging and discharging assuming hy- here for visual clarity.) The expected adsorption number g from Eq. 1 is 3.987 and 0.001 at charging and discharging conditions, drogen pressure as the single control parameter. They showed that respectively, and the actual delivery amount is the difference the existence of a substantial free volume could improve the de- (3.986 per adsorption site). In this case, the delivery amount livery amount (36). They suggested that not only the size but the shape of the pore should be important, and some ZIF structures (3.986) is close to nmax (4) because the binding energy is appropriate, namely, it falls within the range of two μ values for charging could provide better adsorption isotherms than MOFs in terms of and discharging conditions. It should be as far below as possible the delivery amount. On the other hand, we suggested that a mix- ture of H and a purging gas, such as NH , would provide a control from μ of charging (−0.196 eV) for efficient H2 adsorption and as 2 3 far above as possible from μ of discharging (−0.386 eV) for efficient mechanism that could be an alternative or a complement to temperature as part of the discharging parameters (40). H2 release. Parenthetically, we note that εn is the binding energy level that is negative, whereas the binding energy is usually presented only with the magnitude (without the negative sign). Fol- Electrostatic Interactions and Alkaline Cations lowing this convention, a representative binding energy value for In the present and following two sections, we elaborate on un- a maximized delivery amount under the aforementioned operation derlying interaction mechanisms between hydrogen molecules conditions can be said to be 0.29 eV (or 28 kJ/mol). It is obvious and adsorption sites. It has been widely perceived that the in- that, if the magnitude of binding energy is too small (such as the teraction mechanism of H2 with adsorption sites can be grouped vdW interaction of 0.05 eV), the room temperature storage is al- into the vdW (dispersion), electrostatic, and orbital interactions, most zero. On the other hand, the magnitude of the binding energy which are often intermixed in real physical systems. The vdW should not be too large, either. The case of V, shown in Fig. 1B, interaction is always present, but the strength by itself is too weak constitutes a striking example. The configuration with the maxi- to stably confine the hydrogen gas molecules near room tempera- mum number of H2 molecules adsorbed onto V (nmax = 4) looks ture (18). For a measurable amount of adsorption in noncryogenic similar to Fig. 1A. However, εns are out of the range of two μ values, conditions, additional treatments of surfaces are required to

Park et al. PNAS | December 4, 2012 | vol. 109 | no. 49 | 19895 Downloaded by guest on September 25, 2021 incorporate the electrostatic and orbital interactions. We focus on structures require a sealing procedure to protect the inside from the former in the present section. oxidation or other contamination. Therefore, the transition- The electrostatic interaction involves the dipole and/or quadru- metal decoration on porous structures seems to require more pole moment of H2 induced by the charged species centered on the critical investigations and harder efforts on the synthetic chem- adsorption sites (41). There is an obvious contrast between the istry side in the future. binding mode of H2 onto positively charged sites and that onto negatively charged ones: H2 molecules bind in the side-on fashion Theoretical Issues in the Interaction Between Ca and H2 onto a positively charged center, whereas they are in the end-on Other than alkaline and transition metals, calcium has attracted shape onto negatively charged atoms. The driving mechanism of the attention as decorating-metal atoms recently (54–56). Calcium has end-on pattern is attributed to the dipole moment of the H2 induced a lower aggregation tendency because of its small cohesive energy by the negative charge, whereas the side-on adsorption is the out- compared with transition metals. Besides the potential use of Ca- fi come of the interaction between the electric eld and quadrupole coordinated surface structures as hydrogen absorbents, the in- moment of H2. Lochan and Head-Gordon (42) published a com- teraction between Ca and hydrogen molecules constitutes an in- prehensive article on these issues, using hierarchical levels of the- 2- teresting academic topic from a theoretical point of view. Although ories. They showed that the negatively charged ligands (e.g., SO4 − a free Ca atom does not have 3d electrons and the outermost and F )canattractH with some enhanced binding strength. 2 electron is in the 4s state, the adsorption of a few H molecules and In terms of the adsorption strength and also the capability of 2 the accompanying orbital hybridization lowers the 3d energy levels gathering multiple H2 molecules onto a single adsorption site, the positively charged alkaline cations are superior to the negatively below the Fermi level (54, 57). Thus a sharp discontinuous transition in the valence configuration of Ca atoms is observed from 4s charged species. It has been shown that 6H2 molecules can be + – bound to a bare Li cation electrostatically (42). As stated in pre- to 3d upon H2 adsorption (58 60). Fig. 2 A and B illustrate such vious sections, however, because free ions by themselves cannot be a change in the Ca valence states. This involves the multireference used for the storage medium, they should be incorporated into character of the electron wavefunction and provides a prototypical porous framework structures in the solid-state form. Even though example of the failure of locally approximated density functionals. the experimental synthesis is yet to be proven, theoretical results for The exchange energy as a local functional of density (and the the alkaline-decorated organic porous structures deliver a promis- gradient of density) underestimates the energy costs associated ing message. We considered the substituted alkaline atoms for H with the transition from an s orbital to a higher angular momentum atoms in the linker molecules before (43), whereas Goddard et al. explored the adsorbed alkaline atoms on the aromatic plane of the linker molecules (34–37). Both cases retained substantial cationic features and achieved the binding of a few H2 molecules with a reasonable strength. Transition Metal and Kubas Interaction Undercoordinated transition metal atoms can show a more sa- lient binding affinity toward H2. This is described by the Dewar– Chatt–Duncanson coordination of dihydrogen to transition metal atoms and is called the Kubas interaction (44). Heavier metal atoms with 4d and 5d electrons can also have this quality, but they are not attractive for hydrogen storage because of the heavy weight. The Kubas interaction involves the orbital hybridization between the partially filled (and so partially empty) 3d orbitals of the transition metal atom and the σ (both bonding and antibonding) orbitals of H2. In another view, it is interpreted as the electron donation from the H2 σ-bonding orbital to empty metal 3d orbitals and the simultaneous back-donation from the filled metal d orbital to the H2 σ-antibonding orbital. Upon coordination, the H–H bond length increases by approximately 10% of the free H2 bond length (45, 46), mainly owing to some occupation of the H2 σ-antibonding orbital. Several H2 molecules can cluster onto a single metal atom through this type of orbital interactions. Numerous theorists explored the transition metal- decorated surface structures as a hydrogen storage medium (47– 49), and experimentalists identified the Kubas-type hydrogen molecular adsorptions onto such sites of adsorption (50–53). Despite positive predictions and anticipations, a practical mani- festation of metal-coordinated open sites on the surfaces of porous materials, for use as hydrogen adsorption sites, still waits for a great breakthrough. One of the most serious obstacles is the strong tendency of under-coordinated metal atoms to aggregate or oxidize. The purpose of introducing functional groups such as OH men- tioned before was to disperse metal atoms in the backbone matrix Fig. 2. Geometries of Ca atom with (A) attached and (B) far separated four and suppress the metal aggregation. To disperse the metal atoms hydrogen molecules. (C) Optimized geometry and (D) molecular orbital levels of the Ca adsorbed anthracene structure. (E) Same structure as in C, with four (or, more realistically, metal ions) using solution chemistry, it may adsorbed hydrogen molecules. The isosurface plot of each HOMO state is also be necessary to use organo-metallic precursors containing the metal illustrated in A, B, and C. Large ivory balls represent Ca. Symbols of other element. The detailed route for the synthesis of such sorbent atoms follow the same convention used in Fig. 1. Figure is adapted from the materials is a subject beyond the scope of the present article. published data in Ref (58), copyrighted by the American Physical Society and Furthermore, even after the desired synthesis, the metal-coordinated Ref (64), with permission from Elsevier.

19896 | www.pnas.org/cgi/doi/10.1073/pnas.1217137109 Park et al. Downloaded by guest on September 25, 2021 state and thus artiﬁcially favors the occupation of 3d orbitals over Carbon materials have been considered as a good candidate for

4s orbitals (61). a cost-effective receptor. We present here the hydrogen migration INAUGURAL ARTICLE We tested various theories, including the DFT, hybrid func- characteristics of pure graphene, which is a generic sp2-bonded tionals, and Hartree–Fock-based theories. We found that the hy- carbon surface. In Fig. 3 we show the hydrogen desorption barrier brid functional could alleviate the errors if a somewhat larger-than- and the migration barrier on graphene. The stability of the hy- usual ratio of the exact exchange was used. Other groups also drogenated domains and the barriers, either for the desorption or performed computations with various theories (62, 63). In partic- for the migration of hydrogen adatoms, largely depends on the ular, Bajdich et al. (62) reported that the exact exchange combined adsorption configuration (73, 74). Fig. 3A, Left, is the case in which with a correlation density functional yielded a quite accurate en- two adsorbed hydrogen atoms constitute the metadimer configu- ergy curve similar to the quantum Monte Carlo calculations. Put- ration. In this case the balance between the bipartite sublattice of ting aside these theoretical issues, the Ca-decorated structures the graphene is violated, and the adsorbed state (Fig. 3A, Left)is might be more useful than the transition metal structures if the thermodynamically unstable relative to the desorbed state (Fig. 3B, adsorbed Ca stabilized the down-shifted 3d-derived orbitals after Right). When the hydrogen adatoms in the hydrogenated domain anchoring to the surface. We observed that the valence state of Ca are well paired by occupying each bipartite sublattice equally (Fig. on defective graphene or other reactive carbon surface structures 3 B, Left and C, Left), the desorption involves a large energy cost, readily possessed the 3d characters, and the abrupt orbital change and the desorption barrier increases substantially. The difference induced by adsorption could be effectively eliminated (57, 64). As of desorption barriers between Fig. 3 A and B is remarkable. Of an example, adsorption configurations of Ca on anthracene and greater significance is the even larger migration barrier of two the electronic structure are shown in Fig. 2 C and D, respectively. right-most hydrogen adatoms in Fig. 3C. This suggests that, on the The highest occupied molecular orbital (HOMO), as shown in Fig. pure graphene surface, the unpaired hydrogen adatoms tend to 2C, much resembles the 3d orbital centered at Ca. In this case, the desorb to become molecular hydrogen gas rather than migrate adsorption of hydrogen molecules, as depicted in Fig. 2E, does not within the plane. Note that the gaseous phase has obvious advan- induce a sharp orbital change in the valence states. tage in the entropy (75). On the other hand, once hydrogen adatoms are paired on the graphene plane, they are likely neither to Search for Kubas-Type Orbital Interactions in Light Elements desorb nor to migrate. In previous sections we discussed the mechanisms (electrostatics The most important necessary condition for an efficient hydro-

or orbital interactions) that can bind a few hydrogen molecules gen spillover is probably the existence of migrations channel along SCIENCES onto alkaline cations, transition metal atoms, and (the 3d orbital

which hydrogen adatoms can travel in a deep inner surface of the APPLIED PHYSICAL state of) Ca. In a broad sense, Ca can be included in the family of material. Aforementioned features of hydrogen adatoms on gra- transition metals as far as the Kubas interaction via 3d orbitals is phene suggest that the pure sp2 carbon surfaces are not relevant concerned. Although under-coordinated transition metal, Ca, as a receptor because of the poor kinetics. There must be some and alkaline metals are all very attractive for hydrogen molecular extrinsic sources so as to boost the hydrogen migration within adsorptions, the synthesis of the desired structure and its post- the carbon surface. In this respect, we have explored shuttling synthetic maintenance are not well established. Therefore, it is natural that the search for alternative methods continues. Re- cent DFT calculations revealed that Be 2p orbitals can interact with H2 molecular orbitals in a similar way (65, 66). Because of the light weight and the weaker tendency of aggregation, Be may bear a greater potential from the scientific viewpoint, but its toxicity brings about another problem for the realization. Others previously reported that B-doped fullerene could exhibit a similar Kubas-like H2 adsorption property (66). However, our re- fined calculations revealed that the suggested states were only locally stable, and the adsorbed H2 molecules were easily dis- sociated on B sites of the fullerene (67). We also tested a few other light elements but found no comparable orbital interactions. Therefore, the observed orbital interaction between H2 and Be seems quite exceptional among light elements. Spillover of Hydrogen onto Inert Surfaces A mechanism that is quite distinct from metal hydride, chemical hydride, and Kubas-enhanced adsorption mechanisms has been suggested for the enhancement of the hydrogen storage at elevated temperature. It has been proposed that accumulated hydrogen molecules on metal sites may be activated through catalysis and then spill over onto the otherwise inert surface, called the receptor (68–71). Platinum nanoparticles are commonly used as catalysts, and activated carbon materials combined with MOFs have been tested as receptors. Experimentalists report a high wt% hydrogen Fig. 3. Desorption barrier of two hydrogen atoms from (A) the metadimer storage in metal–carbon complexes even at room temperature and configuration and (B) the paired six adatoms configuration. (C) Migration suggest potential contributions from the spillover mechanism (68, barrier of two hydrogen adatoms from the paired 24 adatoms config- 69). Theorists have suggested a variety of atomistic models related urations into a separated paradimer configuration. IS, TS, and FS stand for fi to the hydrogen movement from the catalyst to the receptor, along initial, transition, and nal states, respectively. Blue balls are hydrogen atoms, and the network of carbon atoms is represented by the wireframe. with the migration within the receptor surfaces (72). Despite (C, Right) The two carbon atoms from which hydrogen atoms migrated are a reported increase in the room-temperature storage capacity, the highlighted with gray balls. Dashed lines are only a guide for the eye. (C, desorption kinetics and reversibility must be improved and the Inset) A zoomed-in view depicting the direction of the migration. Parts of movement of hydrogen atoms and binding states on the receptor figure adapted from the published data in Ref (75). Copyrighted by the surfaces must be unambiguously identified. American Physical Society.

Park et al. PNAS | December 4, 2012 | vol. 109 | no. 49 | 19897 Downloaded by guest on September 25, 2021 molecular species that can exchange hydrogen atoms with carbon storage system for such wide spectra of applications, the volumetric surfaces (76). In a sense, the spillover mechanism can be regarded storage density and reversibility can be a target of prime consider- as an intermediate storage mechanism between the chemical hy- ation, which imposes different limitations on the choice of storage dride and the surface adsorption. We stated in the previous sec- materials than the vehicular applications. These features should be tions that poor reversibility and slow kinetics are considered to be considered more seriously in future investigations. In any case, al- the limitation of chemical hydrides. It was also pointed out that though we have searched for many different absorbent materials under-coordinated transition metals and alkaline metal cations on based on the idea of metal-decorated porous structures, we must the surfaces of porous materials have not been shown to be syn- concede that much more effort is still required for the synthesis of thesizable yet. Taking all these limitations into consideration, an intended materials. The study should be further extended to post- appropriate choice of a catalyst and a receptor to facilitate spillover-type hydrogen storage would be highly desirable and deserves synthetic management to avoid the problem of oxidation or other more focused study in the future. A recent study that demonstrated contaminations of the adsorption sites. Regarding the spillover hydrogen dissociation and ensuing migration on Pd-doped Cu mechanism, despite numerous suggestions and debates, the micro- surfaces suggests a clue for a choice of catalyst and receptor (77). scopic details of the hydrogen adatoms on receptors have not been fully explored and deserve intensive study. In particular, when the Summary and Future Directions sp2 carbon materials are considered as the receptor, methods for an In this article we have reviewed recent progress on sorption-based efficient hydrogen migration along the carbon surfaces need to be hydrogen storage, emphasizing the materials design aspects. Ex- developed to establish the hydrogen spillover onto carbon materials. perimental studies concerning the development of large surface Up to now the capacity of sorption-based hydrogen storage is porous framework structures were briefly sketched. Computation- acceptable only at very low temperatures, and room temperature based studies that mainly concentrated on enhancing the binding reversible storage is yet to be achieved. However, in view of a very affinity of hydrogen molecules onto surfaces were closely examined short history of this (sorbent-based storage) research area com- fi and summarized. We classi ed the mechanisms of hydrogen mo- pared with others, the developments in the last decade may be lecular adsorption (beyond the vdW interaction) into three groups: regarded as impressive. We expect that a more thorough study of electrostatic interactions based on alkaline cations, Kubas inter- porous structures and surface adsorption mechanisms in the fu- actions with open transition metals, and orbital interactions in- ture will eventually lead to a successful development of efficient volving Ca and other nontransitional metals. Various efforts were and safe storage materials usable under ambient conditions. undertaken to take advantage of such metal elements to increase the H2 binding strength onto porous materials. Theoretical issues ACKNOWLEDGMENTS. This work was supported by the National Research and methods of computations were also summed up. Foundation of Korea (NRF) through the Ministry of Education, Science and Besides the recent focus on the automotive applications, the Technology (MEST) Grant 2006-0093853, the 2010 Korea-Sweden Research objectives of hydrogen storage can be defined in a wide perspective. Cooperation Program of NRF, and the Core Competence Enhancement Pro- gram (2E22790) of the Korea Institute of Science and Technology. N.P. was As the global capacity for harvesting renewable (such as solar and supported by the Hydrogen Energy R&D Center, a 21st Century Frontier R&D wind) energy increases, large-scale energy back-up systems are Program, funded by MEST, Korea. Computations were performed through attracting greater attention than before. To develop a hydrogen the support of the Korea Institute of Science and Technology Information.

1. Jacobson MZ, Delucchi MA (2009) A path to sustainable energy by 2030. Sci Am 301 19. Møller C, Plesset MS (1934) Note on an approximation treatment for many-electron (5):58–65. systems. Phys Rev 46(7):618–622. 2. George Thomas (2000) Overview of Storage Development DOE Hydrogen Program. 20. Adams DJ (1974) Chemical potential of hard-sphere fluids by Monte Carlo methods. Available at http://www1.eere.energy.gov/hydrogenandfuelcells/pdfs/storage.pdf. Mol Phys 28(5):1241–1252. Accessed October 30, 2012. 21. Han SS, Mendoza-Cortés JL, Goddard WA, 3rd (2009) Recent advances on simulation 3. van den Berg AWC, Areán CO (2008) Materials for hydrogen storage: Current research and theory of hydrogen storage in metal-organic frameworks and covalent organic trends and perspectives. Chem Commun (Camb) (6):668–681. frameworks. Chem Soc Rev 38(5):1460–1476. 4. US Department of Energy (2011) DOE Targets for Onboard Hydrogen Storage Systems for 22. Venna SR, Carreon MA (2010) Highly permeable zeolite imidazolate framework-8 Light-Duty Vehicles. Available at http://www1.eere.energy.gov/hydrogenandfuelcells/ membranes for CO2/CH4 separation. J Am Chem Soc 132(1):76–78. storage/current_technology.html. Accessed October 30, 2012. 23. Rosi NL, et al. (2003) Hydrogen storage in microporous metal-organic frameworks. 5. Lim KL, Kazemian H, Yaakob Z, Daud WRW (2010) Solid-state materials and methods Science 300(5622):1127–1129. – for hydrogen storage: A critical review. Chem Eng Technol 33(2):213 226. 24. Jung DH, et al. (2006) Grand canonical Monte Carlo simulation study on the cate- 6. Schlapbach L, Züttel A (2001) Hydrogen-storage materials for mobile applications. nation effect on hydrogen adsorption onto the interpenetrating metal-organic – Nature 414(6861):353 358. frameworks. J Phys Chem B 110(46):22987–22990. 7. Xiong Z, et al. (2008) High-capacity hydrogen storage in lithium and sodium amido- 25. Dinca M, Long JR (2008) Hydrogen storage in microporous metal-organic frameworks boranes. Nat Mater 7(2):138–141. with exposed metal sites. Angew Chem Int Ed Engl 47(36):6766–6779. 8. Zhao D, Yuan D, Zhou H-C (2008) The current status of hydrogen storage in metal- 26. Wang Z, Cohen SM (2009) Postsynthetic modification of metal-organic frameworks. organic frameworks. Energy Environ Sci 1(2):222–235. Chem Soc Rev 38(5):1315–1329. 9. Yaghi OM, et al. (2003) Reticular synthesis and the design of new materials. Nature 27. Mulfort KL, Farha OK, Stern CL, Sarjeant AA, Hupp JT (2009) Post-synthesis alkoxide 423(6941):705–714. formation within metal-organic framework materials: A strategy for incorporating 10. Furukawa H, et al. (2010) Ultrahigh porosity in metal-organic frameworks. Science highly coordinatively unsaturated metal ions. J Am Chem Soc 131(11):3866–3868. 329(5990):424–428. 28. Park N, Hong S, Kim G, Jhi SH (2007) Computational study of hydrogen storage 11. Suh MP, Park HJ, Prasad TK, Lim D-W (2012) Hydrogen storage in metal-organic characteristics of covalent-bonded graphenes. J Am Chem Soc 129(29):8999–9003. frameworks. Chem Rev 112(2):782–835. 29. Cheon YE, Suh MP (2009) Enhanced hydrogen storage by palladium nanoparticles 12. Hohenberg P, Kohn W (1964) Inhomogeneous electron gas. Phys Rev 136(3B): fabricated in a redox-active metal-organic framework. Angew Chem Int Ed Engl 48 B864–B871. – 13. Kohn W, Sham LJ (1965) Self-consistent equations including exchange and correlation (16):2899 2903. effects. Phys Rev 140(4A):A1133–A1138. 30. Lim DW, Yoon JW, Ryu KY, Suh MP (2012) Magnesium nanocrystals embedded in – 14. Becke AD (1988) Density-functional exchange-energy approximation with correct a metal organic framework: Hybrid hydrogen storage with synergistic effect on – asymptotic behavior. Phys Rev A 38(6):3098–3100. physi- and chemisorption. Angew Chem 124(39):9952 9955. 15. Baker TA, Head-Gordon M (2010) Modeling the charge transfer between alkali metals 31. Jeon K-J, et al. (2011) Air-stable magnesium nanocomposites provide rapid and high- and polycyclic aromatic hydrocarbons using electronic structure methods. J Phys capacity hydrogen storage without using heavy-metal catalysts. Nat Mater 10(4): Chem A 114(37):10326–10333. 286–290. 16. Heyd J, Scuseria GE (2004) Assessment and validation of a screened Coulomb hybrid 32. Zou X, Zhou G, Duan W, Choi K, Ihm J (2010) A chemical modification strategy for density functional. J Chem Phys 120(16):7274–7280. hydrogen storage in covalent organic frameworks. J Phys Chem C 114(31): 17. Anisimov VI, Zaanen J, Andersen OK (1991) Band theory and Mott insulators: Hub- 13402–13407. bard U instead of Stoner I. Phys Rev B Condens Matter 44(3):943–954. 33. Cha MH, Nguyen MC, Lee YL, Im J, Ihm J (2010) Iron-decorated, functionalized metal 18. Okamoto Y, Miyamoto Y (2001) Ab initio investigation of physisorption of molecular organic framework for high-capacity hydrogen storage: First-principles calculations. J hydrogen on planar and curved graphenes. J Phys Chem B 105(17):3470–3474. Phys Chem C 114(33):14276–14280.

19898 | www.pnas.org/cgi/doi/10.1073/pnas.1217137109 Park et al. Downloaded by guest on September 25, 2021 34. Han SS, William A G, 3rd (2007) Lithium-doped metal-organic frameworks for re- 55. Wang Q, Sun Q, Jena P, Kawazoe Y (2009) Theoretical study of hydrogen storage in versible H2 storage at ambient temperature. J Am Chem Soc 129(27):8422–8423. Ca-coated fullerenes. J Chem Theory Comput 5(2):374–379. INAUGURAL ARTICLE 35. Blomqvist A, Araújo CM, Srepusharawoot P, Ahuja R (2007) Li-decorated metal-or- 56. Lee H, Ihm J, Cohen ML, Louie SG (2010) Calcium-decorated graphene-based nano- ganic framework 5: A route to achieving a suitable hydrogen storage medium. Proc structures for hydrogen storage. Nano Lett 10(3):793–798. Natl Acad Sci USA 104(51):20173–20176. 57. Kim G, Jhi SH, Lim S, Park N (2009) Crossover between multipole Coulomb and Kubas 36. Han SS, Choi SH, Goddard WA (2011) Improved H2 storage in zeolitic imidazolate interactions in hydrogen adsorption on metal-graphene complexes. Phys Rev B 79 frameworks using Li+,Na+, and K+ dopants, with an emphasis on delivery H2 uptake. (15):155437. J Phys Chem C 115(8):3507–3512. 58. Cha J, Lim S, Choi CH, Cha MH, Park N (2009) Inaccuracy of density functional theory 37. Mendoza-Cortés JL, Han SS, Goddard WA, 3rd (2012) High H2 uptake in Li-, Na-, K- calculations for dihydrogen binding energetics onto Ca cation centers. Phys Rev Lett metalated covalent organic frameworks and metal organic frameworks at 298 K. J 103(21):216102. “ Phys Chem A 116(6):1621–1631. 59. Ohk Y, Kim Y-H, Jung Y (2010) Comment on Inaccuracy of density functional theory ” 38. Lee H, Choi WI, Ihm J (2006) Combinatorial search for optimal hydrogen-storage calculations for dihydrogen binding energetics onto Ca cation centers Phys Rev Lett, nanomaterials based on polymers. Phys Rev Lett 97(5):056104. 104(17):179601, author reply 179602. 39. Lee H, et al. (2007) Ab initio study of dihydrogen binding in metal-decorated poly- 60. Cha J, Choi CH, Park N (2010) Cha, Choi, and Park reply. Phys Rev Lett 104(17):179602. acetylene for hydrogen storage. Phys Rev B 76(19):195110. 61. Gunnarsson O, Jones RO (1985) Total-energy differences: Sources of error in local- – 40. Lee H, Huang B, Duan W, Ihm J (2010) Releasing H_{2} molecules with a partial density approximations. Phys Rev B Condens Matter 31(12):7588 7602. pressure difference without the use of temperature. Phys Rev B 82(8):085439. 62. Bajdich M, Reboredo FA, Kent PRC (2010) Quantum Monte Carlo calculations of di- 41. Yoon M, Yang S, Wang E, Zhang Z (2007) Charged fullerenes as high-capacity hy- hydrogen binding energetics on Ca cations: An assessment of errors in density drogen storage media. Nano Lett 7(9):2578–2583. functionals for weakly bonded systems. Phys Rev B 82(8):081405. 42. Lochan RC, Head-Gordon M (2006) Computational studies of molecular hydrogen 63. Purwanto W, Krakauer H, Virgus Y, Zhang S (2011) Assessing weak hydrogen binding on Ca+ centers: An accurate many-body study with large basis sets. J Chem Phys 135 binding affinities: The role of dispersion forces, electrostatics, and orbital inter- (16):164105–164111. actions. Phys Chem Chem Phys 8(12):1357–1370. 64. Cha J, Choi CH, Park N (2011) Ab initio study of Kubas-type dihydrogen fixation onto 43. Huang B, Lee H, Duan W, Ihm J (2008) Hydrogen storage in alkali-metal-decorated d-orbitalstatesofCaadatoms.Chem Phys Lett, 513(4–6):256–260. http://www. organic molecules. Appl Phys Lett 93(6):063107–063103. sciencedirect.com/science/journal/00092614 44. Kubas GJ (2001) Metal–dihydrogen and σ-bond coordination: The consummate ex- 65. Lee H, Huang B, Duan W, Ihm J (2010) Beryllium-dihydrogen complexes on nano- tension of the Dewar–Chatt–Duncanson model for metal–olefin π bonding. J Orga- structures. Appl Phys Lett 96(14):143120–143123. nomet Chem 635(1–2):37–68. 66. Kim YH, Zhao Y, Williamson A, Heben MJ, Zhang SB (2006) Nondissociative adsorp- 45. Niu J, Rao BK, Jena P (1992) Binding of hydrogen molecules by a transition-metal ion. tion of H2 molecules in light-element-doped fullerenes. Phys Rev Lett 96(1):016102. Phys Rev Lett 68(15):2277–2280. 67. Lee H, et al. (2008) Room-temperature dissociative hydrogen chemisorption on bo- 46. Gagliardi L, Pyykkö P (2004) How many hydrogen atoms can be bound to a metal? ron-doped fullerenes. Phys Rev B 77(23):235101. Predicted MH12 species. J Am Chem Soc 126(46):15014–15015. 68. Li Y, Yang RT (2006) Hydrogen storage in metal-organic frameworks by bridged hy- 47. Yildirim T, Ciraci S (2005) Titanium-decorated carbon nanotubes as a potential high- drogen spillover. J Am Chem Soc 128(25):8136–8137. capacity hydrogen storage medium. Phys Rev Lett 94(17):175501. 69. Tsao CS, et al. (2010) Effect of catalyst size on hydrogen storage capacity of Pt-im- SCIENCES 48. Sun YY, Kim YH, Zhang SB (2007) Effect of spin state on the dihydrogen binding pregnated active carbon via spillover. J Phys Chem Lett 1(7):1060–1063. strength to transition metal centers in metal-organic frameworks. J Am Chem Soc 129 70. Psofogiannakis GM, Froudakis GE (2009) DFT study of hydrogen storage by spillover APPLIED PHYSICAL – (42):12606 12607. on graphite with oxygen surface groups. J Am Chem Soc 131(42):15133–15135. 49. Bhattacharya A, Bhattacharya S, Majumder C, Das GP (2010) Transition-metal deco- 71. Psofogiannakis GM, Froudakis GE (2011) Fundamental studies and perceptions on the ration enhanced room-temperature hydrogen storage in a defect-modulated gra- spillover mechanism for hydrogen storage. Chem Commun (Camb) 47(28):7933–7943. – phene Sheet. J Phys Chem C 114(22):10297 10301. 72. Lin Y, Ding F, Yakobson BI (2008) Hydrogen storage by spillover on graphene as 50. Hamaed A, Trudeau M, Antonelli DM (2008) H2 storage materials (22 KJ/mol) using a phase nucleation process. Phys Rev B 78(4):041402. organometallic Ti fragments as σ-H2 binding sites. J Am Chem Soc 130(22):6992–6999. 73. Stojkovic D, Zhang P, Lammert PE, Crespi VH (2003) Collective stabilization of hy- 51. Hamaed A, Mai HV, Hoang TKA, Trudeau M, Antonelli D (2010) Functionalized porous drogen chemisorption on graphenic surfaces. Phys Rev B 68(19):195406. silicas with unsaturated early transition metal moieties as hydrogen storage materi- 74. Sofo JO, Chaudhari AS, Barber GD (2007) Graphane: A two-dimensional hydrocarbon. als: Comparison of metal and oxidation state. J Phys Chem C 114(18):8651–8660. Phys Rev B 75(15):153401. 52. Kim TS, Kim KJ, Jo SK, Lee J (2008) Interaction of D2 with Ti-adsorbed polyaniline and 75. Han SS, Jung H, Jung DH, Choi SH, Park N (2012) Stability of hydrogenation states of implication for hydrogen storage. J Phys Chem B 112(51):16431–16436. graphene and conditions for hydrogen spillover. Phys Rev B 85(15):155408. 53. Phillips AB, Shivaram BS (2008) High capacity hydrogen absorption in transition 76. Han SS, Kim H, Park N (2011) Effect of shuttling catalyst on the migration of hydrogen metal-ethylene complexes observed via nanogravimetry. Phys Rev Lett 100(10): adatoms: A strategy for the facile hydrogenation of graphene. J Phys Chem C 115(50): 105505. 24696–24701. 54. Yoon M, et al. (2008) Calcium as the superior coating metal in functionalization of 77. Kyriakou G, et al. (2012) Isolated metal atom geometries as a strategy for selective carbon fullerenes for high-capacity hydrogen storage. Phys Rev Lett 100(20):206806. heterogeneous hydrogenations. Science 335(6073):1209–1212.

Park et al. PNAS | December 4, 2012 | vol. 109 | no. 49 | 19899 Downloaded by guest on September 25, 2021