Crystal shape controlled H2 storage rate in nanoporous carbon composite with ultra-fine Pt nanoparticle

This study demonstrates that the hydrogen storage rate (HSR) of nanoporous carbon supported platinum nanocatalysts (NC) is determined by their heterojunction and geometric configurations. The present NC is synthesized in an average particle size of ~1.5 nm by incipient wetness impregnation of Pt4+ at carbon support followed by annealing in H2 ambient at 102–105 °C. Among the steps in hydrogen storage, decomposition of H2 molecule into 2 H atoms on Pt NC surface is the deciding factor in HSR that is controlled by the thickness of Pt NC. For the best condition, HSR of Pt NC in 1~2 atomic layers thick (4.7 μg/g min) is 2.6 times faster than that (1.3 μg/g min) of Pt NC with higher than 3 atomic layers thick.


Results
From theoretical geometrical calculation, surface-to-bulk ratio of NC with a known size is proportional to the crystal shape. The step-by-step strategy for determining the impacts of crystal shape on HSR of Pt NC is (1) XRD and SAXS analyses on the coherent length and geometrical asymmetry of NC in carbon support, (2) surface-to-bulk ratio determination by cross-referencing results of XAS and geometrical calculation, and finally (3) adsorption energy calculation on Pt-H bondings at Pt NC with thickness determined by structure characterizations.
Crystal structure and nanostructure of active carbon supported Pt nanocatalyst (Pt NC). Crystal structure of Pt/AC and carbon support are revealed by using XRD analysis. Figure 1 demonstrates the XRD pattern of INER_A (1a) and INER_B (1b) before and after conducting the H 2 purged High-pressure Thermogravimetric analyzer (HP-TGA). As indicated in Fig. 1, peak D is the characteristic line of (002) facet for active carbon (AC) while peak X represents the lines of (111)/(220) facets for metallic Pt and (110) facet for PtO (or (002) facet for PtO 2 ). For freshly prepared NC, the average coherent length (h avg ) of metallic phase at Pt (111)/ (200) facets is 9.2/10.2 Å for INER_A and 11.1/11.2 Å for INER_B. After HP_TGA treatment, h avg of Pt(111)/ Pt(200) in INER_A is decreased/increased by 17.4%/11.8%. In INER_B, h avg of Pt(111)/Pt(200) is increased by 8.1/9.8% after exposed to the same treatment. In this condition, h avg of Pt metal oxide phase is decreased by 9.0% in INER_A and is increased by 10.7% in INER_B.
Average particle size (D avg ), particle size distribution (P R ), and particle shape of Pt NC are characterized by SAXS analysis. In general, features of SAXS spectra result from the interference between incident X-ray and both inter-particle (Q < 0.07 Å −1 ) and intra-particle (Q > 0.07 Å −1 ) structures. Inter-particle structure that is also called structure factor (S(Q)) include inter-particle distance as well as the size and packing density of inter-particle agglomerates. Known as form factor (P(Q)), intra-particle structure consists of the shape, dimension, and configuration of NC. This study only focuses on the discussions related to P(Q) of SAXS spectra because of the dominant role of intra-particle structures in H 2 decomposition activity of NC. Figure 2a compares the SAXS spectra of the experimental Pt/AC with least-square fitting curves calculated by the model of polydispersed spheres with bimodal Schulz distribution (ESI); where background scattering of pure AC powder was subtracted for extracting the contribution of metallic NCs. In the low q range, in comparison with INER_B, the SAXS spectrum of INER_A shows a higher scattering intensity gradient (denoted by slope of S A ), which indicates a larger surface roughness (or asymmetry) of NC in INER_A. In the high q region, the scattering hump at q i with a width (W i ) and a height (H i ) results from scattering interferences of among X-ray and intra-particle structures of NC. The ratio of W i /H i is proportional to surface roughness and polydispersity (P i ) of NC; subscripts A and B denote respectively parameters of INER_A and INER_B.
In INER_A spectrum, as compared to that of INER_B, shift of oscillation hump (from q B to q A ) to low q and reduction of W i /H i ratio indicate a larger geometrical asymmetry of NC. Figure 2b shows particle size distribution (Schulz distribution) of experimental Pt NCs. Accordingly, the average radius of both INER_A and INER_B is ~6.9 Å (i.e., D avg ~ 14 Å). Compared to P B , a broadened P A (~35.2%) with an asymmetrical profile (tailing to high R region) implies a high aspect ratio of NC in INER_A. Chemical composition and surface oxidation ratio of Pt NC affected by pre-metal doping treatment on active carbon support. Figure 3a compares Pt L 3 -edge X-ray absorption near-edge spectra (XANES) of experimental Pt NCs and standard samples (CNT supported Pt NC and Pt foil). In a L 3 -edge spectrum, the inflection point position (i.e., 1 st deviation maximum (E 0 ) at arrow X of XANES spectra in Fig. S1) refers to the minimum energy of photoelectron for initiating electron transitions from 2p 3/2 to 5d 5/2 orbitals. Intensity (h B ) and width (W) of near-edge absorption peak (arrow B) denote respectively the 2p 3/2 to 5d 5/2 transition probability and band width of 5d orbital of Pt atoms. As depicted in Fig. 3a     Surface-to-bulk ratio determination by chemical composition of Pt NC. To confirm crystal shape, the surface-to-bulk ratio (η ) of experimental NCs is compared with that of ideal NCs in different geometries (scheme shown in Fig. 4a). Eqn 1 and eqn 2 represent the estimation for η of ideal NCs that changes by varying thickness (disk-like crystal, η d ) and height (octahedron crystal, η Oh ) 10 .
In eqn 1, factors of d, P S , P L , H, and D respectively stand for the diameter of atom, surface atomic packing factor, bulk atomic packing factor, thickness of NC, and diameter of NC. In eqn 2, factors of Δ , A, and h represent respectively area of triangle in outmost surface, area of triangle in horizon plane, and height of octahedron. In Fig. 4b, η d for H = 1 NC is exponentially decreased by 13% (100 to 87%) with D from 10 to 70 Å. In the cases of H = 2 and 3 NCs, η d is decreased by 18% (94 to 72%) and 31% (88 to 57%), respectively. These values are higher than 50%, indicating the lack of bulk atom in disk NC when H ≤ 3 11,12 . For octahedron NC, bulk atom ratio goes up significantly by 68% (i.e., η Oh is decreased from 87 to 19%) with an increases of h from 10 to 70 Å. Taking physiochemical natures into consideration, oxidation occurs only in the first two atomic layers from the outmost surface of Pt NC in an ambient condition below 150 °C. This hypothesis suggests the extent of oxidation for Pt NC is a function of its η . In this study, one can notice that D avg of nanocrystallite is ~14 Å for both INER_A and INER_B, where Pt NC is preferentially grown in 2D clusters and 3D ones, respectively. Consequently, taking results of Table 1 and Fig. 4b into account, the substantial high η (79.8%) of INER_A implies the growth of 2D Pt NC in a H ranging from 1 to 2 atomic layers. In INER_B, η B is 42.9%, indicating the growth of 3D Pt clusters. The value of η B is slightly decreased from that of octahedron crystal with a h = 14 Å. This result could be attributed to the growth of Pt clusters in polymorph whose surface facets are more than those of octahedron crystal.

Density functional theory calculation on impacts of crystal shape to H 2 storage rate on Pt NC.
In order to determine the optimal structure, different configurations of monolayer Pt atoms placed at different stacking sites on graphite surface are built. Configuration with Pt atom on atop position is named as 1PtAC, on bridge site as 1PtB, on hollow sites 1PtC, and 1PtAC for pairs at both atop and hollow sites. Average distance between Pt atom and graphene layer is 2.32 Å. The packing structure and energy relaxation profile of two of those models are shown in Fig. S2a and S2b. Among all configurations, 1PtAC model possesses the lowest binding energy (E dep ) of ~0.33 (eV/Pt atom). Therefore, effects of Pt thickness on the H 2 binding energy in Pt NC surface are further investigated by DFT calculation with monolayer (1PtAC in Fig. 5a) and two atomic layers (2PtAC in Fig. 5b) of Pt on graphene. In simulation, position of carbon positions is fixed to reproduce the bonding characteristics between bulk graphite and Pt atoms. As depicted in Fig. 5 Table 4. Accordingly, cohesive energy (E c ) of 1PtAC is 0.582 eV stronger than that of 2PtAC with a H 2 molecule in vacuum. This result accounts for the stronger attraction force to molecular H 2 in 2D Pt cluster as compared to that of 3D Pt ones. After relaxation, H atoms are preferentially located at hollow and bridge sites of 1PtAC and at hollow sites of 2PtAC. The coherence energy difference (Δ E c = E c f − E c ) between initial and final state is resulted from formation of surface Pt-H bond and is − 0.069 (eV Pt atom −1 ) for 2PtAC and − 0.255 for 1PtAC (eV Pt atom −1 ). There are two factors for estimation of E c including (i) number of Pt atom and (ii) distance/bond strength between H and Pt atoms. In the first factor, Δ E c should be doubled by reducing 50% of Pt atoms in DFT model. However, Δ E c of 1PtAC is 2.7 times higher than that of 2PtAC. Such an abnormal Δ E c could be accounted for the formation of strong Pt-H bond due to the coupling effects between adsorbed H on Pt atom and pi-electron in hollow sites of carbon substrate underneath. Adsorption energy (E a ) for H atom on sorption site of 1PtAC (1PtAC-2H model) is -2.043 (eV atom −1 ) and 2PtAC-2H is − 1.117 (eV atom −1 ).

Discussion
Crystal shape of NC is one of the most influential factors in its HSR because different atomic packing densities of crystal facets will change the activation energy for redox reactions. For nanocrytallite with a known size, surface-to-bulk ratio (η ) is proportional to its crystal shape. The ratio can be estimated by oxidation ratio considering that surface chemisorption of oxygen is a general form of oxide in the outmost layer of Pt NC. Therefore, in determining the correlation between crystal shape to HSR of Pt NC, one should confirm their particle size, oxidation state, and the structure asymmetry by structure characterizations. After that, results of theoretical model calculations on η to crystal shape is needed. In this study, X-ray diffraction (XRD), small angle X-ray scattering (SAXS), X-ray absorption spectroscopy (XAS) analyses, and theoretical geometry calculation are employed to reveal the bulk structure parameters, oxidation ratios, as well as η in relation to the shape of Pt NC. To investigate the impacts of crystal shape on HSR of Pt NC, density functional theory (DFT) calculation on proper atomic models is performed.
In XRD analysis, ratios of peak X and peak D intensity (H X /H D ) are substantially increased from 0.28 (INER_A) to 1.67 (INER_B), which indicates the increasing of long-term ordering structure in carbon support   by decreasing Pt content from 0.35 to 0.13 wt%. After HP-TGA test, the further decreased H X /H D implies a disruption of long-term ordering in carbon support. This phenomenon results from intercalation of Pt clusters into carbon support by different extents between INER_A (22.2%) and INER_B (35.4%). In this event, decreasing Δ (H X /H D ) with increasing full width at half maximum (FWHM) of peak D depicts a restructure of Pt atoms into low-dimension Pt clusters on the high dangling bond density surface of INER_A. This hypothesis is further confirmed by uneven changes between average coherent length (h avg ) of Pt facet (111) (decreased by 17.8%, from 9.2 to 7.6 Å) and (200) facet (increased by 11.8%, from 10.2 to 11.4 Å). Such a phenomenon could be attributed to relocation of atop Pt atoms to carbon sites around metal clusters (Fig. 6).

RPt-Hh (Å) RPt-Hb (Å) H 2 in vacuum H 2 in sorption site Pt-H at sorption site
In the meantime, h avg of Pt oxide phase is decreased by ~9.0% revealing weak Pt-O bonding on Pt atoms directly adjacent to electron rich support (i.e., shielding effects of graphite or graphene). The growth of 2D Pt cluster is further confirmed by the asymmetry SAXS scattering profile of INER_A. On the other hand, h avg of Pt metal and Pt oxide phases in INER_B is increased by 8.1~10.7% after HP-TGA. The growth of both metal and oxide phases is attributed to Pt restructure into 3D cluster by inter-particle coalesce as a result of weak attachment of NC at carbon support. In this event, Pt-O phase is formed in 3D Pt NC surface.
From SAXS analysis, D avg (~1.4 nm) of Pt NCs in INER_A is similar to that of INER_B. Based on XRD analysis, h avg of Pt NCs in the two samples is highly dependent on surface-to-bulk ratio. These observations validate the correlation between η and shape of NC with a known size. To identify the crystal shape of experimental NCs, correlations between their oxidation state and η are investigated by cross-referencing the results of atomic arrangement, oxidation states, and geometry configurations (VEDO method) 10 . In INER_A, the content of Pt metal is 20.3% and that of Pt oxide (PtO and PtO 2 ) is 79.7%. Given that surface oxygen chemisorption is a general form in Pt NCs, their oxidation ratio also refers to surface-to-bulk ratio. Accordingly, η of INER_A (η A ) is 79.7%. This value is higher than η Oh of a 1.4 nm octahedron (62%), suggesting the growth of disk-like Pt NCs in a thickness between 1-2 atomic layers. As for INER_B, η B is 42.9% indicating the formation of 3D Pt NCs.
Based on the results of structural analyses and H 2 sorption test, our study shows that Pt atoms tend to stack into 2D NC on active carbon by annealing at 102 °C under H 2 for 2 h (hydrogen reduction) after a metal ion impregnation process. In this condition HSR of carbon supported Pt NC is 4.7 μ g/g min. By increasing annealing temperature to 130 °C under H 2 , Pt atoms are shaped into 3D polymorph NC on active carbon; the NC HSR becomes 1.3 μ g/g min. To clarify the impacts of crystal shape on HSR of Pt NC, H 2 sorption energy of proper atomic stacking scheme is simulated by ab initio DFT calculation. Our hypothesis is further confirmed by the results of binding energy of H atom (E a ) in current models. As indicated in Table 4, the endothermic E a indicates that the spillover decomposition of H 2 molecule into 2 H atoms on Pt surface and positioning of the H atoms to neighboring sites are thermodynamically spontaneous reactions in 1PtAC-2H (E a = − 2.043 eV atom −1 ). In 2PtAC-2H, E a is − 1.117 (eV atom −1 ). The smaller E a suggests a weaker Pt-H bonding in 2PtAC as compared to that of 1PtAC. It also means a higher diffusion coefficient of H atoms, which indicates a strong recombination tendency of adsorbing H atoms into H 2 molecule and thus the suppressed HSR in 2PtAC. Taking together, our results of DFT calculation and structure characterizations explain that geometrical configurations of Pt NC is a determining factor in both HSR and H 2 storage capacity of metal doped nanoporous carbon materials [6][7][8] .
In short, our findings illustrate that Pt atoms tend to aggregate into disk-like NCs with a thickness of 1~2 atomic layers in properly selected carbon support (SSA = 1886 m 2 g −1 ). By increasing the hydrogen-reduced temperature to 130 °C and the use of carbon support without acid (function group) treatment, nanoscaled polymorph Pt clusters in 3D are formed on carbon support. We adopt structural characterizations (XRD, SAXS, XAS) and theoretical model calculations to illuminate that HSR for Pt doped nanoporous carbon materials is dominated Scientific RepoRts | 7:42438 | DOI: 10.1038/srep42438 by the crystal shape of supported Pt NC. In a 2D cluster with a thin atomic layer, H 2 molecule will decompose into 2 H atoms. After that, H atoms will be adsorbed in atop sites in the rough surface. In contrast, in a 3D cluster, H atoms tend to locate in hollow sites of Pt surface with a substantially reduced E a as compared to that of a 2D cluster. Therefore, the HSR of a 3D Pt cluster is drastically lower than that of a 2D cluster.

Methods
Synthesis and the gravimetric hydrogen uptake measurement of Pt doped nanoporous activated carbon composite (Pt NC). The Supporting Material for growing Pt clusters is a nanoporous active carbon (ACB16) with a BET specific surface area (SSA) of 1886 m 2 g −1 . The Pt clusters were synthesized by incipient wetness impregnation of Pt 4+ ions in ACB16 with a vacuum-attraction system 13,14 . The subsequent hydrogen reduction was performed at 102 °C (rising rate = 5 °C min −1 ) for 2 hours. To reinforce NC attachment, the carbon support was washed in an acid base before Pt impregnation. In this treatment, 10 g nanoporous carbon powder was dispersed in the 500 mL acid solution of H 2 SO 4 (95%), HNO 3 (65%) and distilled water The mixed solution was heated from room temperature to 105 °C (rising rate = 0.2 °C min −1 ) and then cooled to room temperature. Details for the experiment procedures are given in our previous works 14 . The samples were named as INER_A (Pt cluster on freshly prepared ACB16) and INER_B (Pt clusters grown on the un-acid-treated carbon ACB17 with SSA = 1857 m 2 g −1 was hydrogen-reduced 130 °C for 2 hours) which respectively contained 0.35 and 0.11 wt% of Pt metal ( Table 5). The gravimetric HSRs for the experimental Pt/AC at room temperature were measured by using HP-TGA (Cahn-Thermax-500) of Institute of Nuclear Energy Research, Taiwan). The parameters for preparing the two samples were decided in resulting their optimum HSR. Details of measurement receipt are similar to those given elsewhere 14 . The small angle X-ray scattering (SAXS) spectra were measured using the Bruker Nanostar instrument 6 . The shape of NC was determined by combing the results of Pt oxidation state and the surface ratios of nanocrystallites that vary according to the changes of their sizes in a 3D atomic packing model (VEDO) 10 .

Density Functional Theory Calculation.
To investigate the H-adsorption on Pt cluster decorated graphite structure, we have carried out first-principles calculations using Vienna Ab initio simulation package (VASP) 15,16 . The calculations were conducted at formosa5 cluster of National Center for High Performance Computing (NCHC), Taiwan. We adopted the projected augmented wave (PAW) 17,18 pseudopotential method with local density approximation, in which plane-wave energy cutoff of 600 eV and Monkhorst-Pack k-point mesh 19 density of 0.03 Å −1 are used for all calculations. To explore the optimized decoration sites of nano-Pt on graphene, the atop (A), bridge (B), hallow (C) and A&C mixing sites, and decoration distance from 2.1 Å to 3.7 Å were tested. For Pt-decoration sites at atop (A), bridge (B), hallow (C), the supercell's lattice size was fixed at a = b = 19.68 Å with the optimized C-C bond length of 1.42 Å (128 C + 16Pt atoms). For the mixing decoration sites of A&C, another supercell was built up with the lattice size fixed at a = 17.0434, and b = 25.5651 Å, which ensures the C-C bond length kept at 1.42 Å (144 C + 16 Pt atoms). All the supercells were built up with at least 24 Å vacuum regions in the z-direction to avoid the interaction between adjacent periodic images. To find the most energetically favorable structures, all the Pt atoms were fully relaxed with the supercell size and graphene atoms fixed (stiff graphene without topological change). The ionic relaxation was stopped when the total energy change is below 1 meV. Table 5. Physical Characteristics and hydrogen storage rate (HSR) of experimental Pt doped nanoporous carbon materials. * SSA: BET determined specific surface area, data quoted from previous studies [6][7][8] .