The increasing demand for improved materials with complex functionalities requires new approaches that can expand the range of accessible phases through rational incorporation of metastability into materials design.1 Metastable materials have become one of the main pillars in advancing novel concepts and transformative technologies in batteries,2 supercapacitors,3 piezoelectrics,4 optoelectronics,5 and heterogeneous catalysis.6 In contrast, in the area of materials-based hydrogen storage, metastability has yet to be fully leveraged and in fact has been a major barrier towards practical applications. Metastable hydrogen storage materials, such as NH3BH3,7 metal amidoboranes,7 LiAlH4,8 and AlH39 offer high volumetric and gravimetric energy densities,10 but have unfavorable thermodynamics of hydrogen release, and can only be rendered reversible using solution chemistries that require complex and energy intensive off-board regeneration.7,9,11-15 On top of disadvantageous thermodynamics, these materials often suffer from sluggish hydrogen release kinetics, which has discouraged their consideration as viable on-board vehicular storage candidates.16
Nanoscaling is a well-known strategy for improving the properties of metal hydrides, but it has been almost exclusively used to destabilize materials that are too thermodynamically stable, with the most dramatic effects being kinetic rather than thermodynamic in nature.16 Most efforts to date have been focused on nanoconfinement within nanoporous hosts to limit the hydride particle size.16,17,18-22 Carbons and carbon-based materials are by far the most common host materials, as they are lightweight and display remarkable chemical and thermal stability under conditions required to achieve reversible H2 release.16,17,23 Apart from utilizing the pores for maintaining nanoparticle size, recent studies by Carr et al. show that the incorporation of heteroatoms in carbon affects the kinetics and desorption pathway of the hydride, making the host material an active participant in the reaction process.24,25 An important advantage of metastable hydrides is that, unlike most binary and complex metal hydrides, very little thermal input is required for hydrogen release. We hypothesize that nanoconfinement can provide a chemical environment similar to the one found in solid adducts of metastable metal hydrides that have shown stabilizing success (for example LiAlH4-dimethylether adduct26). By analogy, we reason that functionalities capable of inducing charge redistribution between the host and the metal hydride can thermodynamically stabilize metal hydrides and render them reversible.
Here, we report a new thermodynamic stabilization concept in which metastable LiAlH4 is nanoconfined within the pores of nitrogen-doped CMK-type carbons. We selected LiAlH4 for this investigation because of its high gravimetric hydrogen capacity (10.6 wt. %) and pure hydrogen release upon mild heating.27,28 In bulk, the thermally induced hydrogen release from LiAlH4 comprises three decomposition steps (Table 1), the first of which is exothermic (-10 kJ mol-1 H2), which makes rehydriding infeasible.29 In this work, we demonstrate that reversibility in LiAlH4 can be enabled by nanoconfinement within a N-doped carbon scaffold. The resulting nanoconfined LiAlH4 material can be regenerated without resorting to complex solution-phase chemistry, opening the door to reconsideration of other high-capacity light metal hydrides that were previously considered unusable for practical applications. These results and the accompanying foundational understanding indicate that “thermodynamic stabilization” of metastable metal hydrides is a viable approach to hydrogen storage, presenting a new strategy for enabling reversibility through guest-host interactions.
Infiltration of LiAlH4 into functionalized scaffolds
Preparation of nanoconfined LiAlH4 samples inside pores of carbon-based scaffolds was achieved via solution infiltration of LiAlH4 freshly recrystallized from diethylether. Prior to the infiltration, the mesoporous scaffolds CMK-3 and NCMK-3 were thermally treated at 350-380 °C under a reducing H2 environment. The hydride nanocomposites are referred to as LiAlH4@CMK-3 and LiAlH4@NCMK-3 depending on the scaffold type (Fig. 1a). LiAlH4 was mixed with the appropriate amount of the chosen scaffold in order to achieve a loading of 20wt% in the final nanocomposite. This loading ensures an efficient and homogenous infiltration of the hydride inside the pores with minimum pore blockage.
The successful incorporation of LiAlH4 into the pores of CMK-3 and NCMK-3 was confirmed by several bulk and surface-sensitive characterization techniques. First, N2 isotherms at 77 K indicate a significant reduction in BET surface area and total pore volume upon nanoconfinement (Fig. 1c). The shape of the isotherms of the mesoporous carbons and their corresponding LiAlH4 loaded composites resemble a type IV(a) isotherm, featuring capillary condensation.30 Calculated pore size distributions reveal 5-6 nm mesopores in all samples (Fig. S1). Both NCMK-3 and LiAlH4@NCMK-3 exhibit a wider pore size distribution than CMK-3, consistent with the higher density of defect sites. LiAlH4 infiltration leads to a reduction of up to 50% in the BET surface area and the total volume for both CMK-3 and NCMK-3 samples (Table S1). FTIR spectroscopic measurements also indicate the presence of characteristic alanate Al-H stretching and bending modes in LiAlH4@CMK-3 and LiAlH4@NCMK-3. In addition, the FTIR spectra confirm that a significant fraction of -OH functional groups, which could possibly react with LiAlH4, are removed by H2 cycling treatment prior to infiltration (Fig. S2).
Powder X-ray diffraction (PXRD) results also support the nanoconfinement of LiAlH4 in these carbon-based composites. Both as-synthesized composites show only broad or low intensity reflections of LiAlH4 when compared to a simple physical mixture of NCMK-3 with an equivalent amount of LiAlH4, implying that LiAlH4 confined inside nanoporous hosts exists as nano-sized crystallites or as an amorphous phase (Fig. S3a).31 The as-synthesized LiAlH4@NCMK-3 sample exhibits reflections of metallic Al peaks, indicating a partial decomposition of LiAlH4 during infiltration (Fig. 1d). Interestingly, 7Li MAS NMR reveals that such decomposition generates LiH rather than the expected stable intermediate Li3AlH6 (Fig. S4). Surface-sensitive X-ray photoelectron spectroscopy (XPS) measurements provide clear evidence that four distinct N functionalities are present in NCMK-3 (Fig. S5 and Table S2). Pyridinic N and graphitic N are the major configurations on the surface of the NCMK-3, with minor contributions from oxidized and pyrrolic N groups. Interestingly, the high population of pyridinic nitrogen sites with electron-donating ability is preserved after hydride infiltration (Table S2 and Supplementary note 1).
HAADF-STEM (high-angle annular dark field-scanning transmission electron microscopy) measurements on nanoconfined samples provide clear evidence of partial decomposition of LiAlH4 upon infiltration (Fig. 1b). In addition, a deposit of small ~1 nm particles was observed on the carbon rods, presumably consisting of LiAlH4. There is a visible structure on the carbon walls in the inset of the infiltrated NCMK-3 shown in Fig 1b. This same surface structure is not observed in the non-infiltrated host material, and is thus assigned to the presence of ~1 nm LiAlH4 particles (Fig. S6a). The HAADF-STEM line profile of LiAlH4@NCMK-3 shows the Al signal is uniformly distributed throughout the C, N, O signals of NCMK-3, and supports a morphology of LiAlH4 uniformly coating the 1D rods of the NCMK-3 host, rather than filling the entire available pore space (Fig. S6b). Metallic Al nanoparticles with sizes between ~50-100 nm are formed outside the carbon pores, as evidenced from TEM images. The metallic character of Al was confirmed by harmonic peaks from EELS (electron energy loss spectroscopy), indicating a partial decomposition of LiAlH4 upon infiltration (Fig. S7). Interestingly, replacing diethylether with THF as a solvent for LiAlH4 infiltration still induces a partial decomposition of metastable LiAlH4 (Supplementary note 2 and Fig. S3b). This is consistent with our preliminary AIMD (ab initio molecular dynamics) calculations which indicate a strong interaction between the N binding sites and LiAlH4, leading to hydride decomposition and spontaneous formation of Li adatoms (vide infra) (Fig. S8).
Hydrogen desorption and absorption
Hydrogen desorption profiles were obtained using the volumetric Sieverts technique, and show a lower onset of H2 desorption temperature for LiAlH4@NCMK-3 compared to LiAlH4@CMK-3, implying that the N-functionalities can promote the decomposition of LiAlH4 (Fig. 2a). In accordance with the observed self-decomposition of metastable LiAlH4 into LiH and Al in NCMK-3, the capacity of as-synthesized LiAlH4@NCMK-3 is slightly lower than that of LiAlH4@CMK-3. The onset dehydrogenation temperature in LiAlH4@NCMK-3 is as low as 126 °C, 45 °C lower than in LiAlH4@CMK-3. Interestingly, the dehydrogenation of the LiAlH4@NCMK-3 exhibits distinct double steps of hydrogen release with a short transition in the middle of the desorption step at around 157 °C. The origin of the double step is likely associated with various kinetic regimes for LiAlH4 decomposition inside functionalized NCMK-3 pores.
In-situ small angle X-ray scattering (SAXS) with simultaneous wide angle scattering (WAXS) H2 desorption experiments were used to probe the structural changes of nanoconfined LiAlH4 function of temperature (20 to 300 °C) (Fig. 2b-e). These experiments indicate significant structural changes upon heating and formation of distinct metallic Al peaks. Interestingly, the in-situ WAXS only exhibits the decomposition of LiAlH4 to LiH/Al without any indications of the stable intermediate phase Li3AlH6 upon heating. The intensity in the low-q region of LiAlH4@CMK-3 exhibits an increase at 75 °C and higher, with a simultaneous decrease in peak area for the three mesostructure peaks (Fig. 2b). In contrast, LiAlH4@NCMK-3 displays essentially no change in the small angle intensity (Fig. 2c). At 75 °C and above, where the low-q SAXS intensity of LiAlH4@CMK-3 begins to increase, the WAXS patterns of both LiAlH4@CMK-3 and LiAlH4@NCMK-3 show the decay of LiAlH4 and the formation of Al metal (Fig. 2d-e).
The reversible uptake of hydrogen in LiAlH4@NCMK-3 was achieved by treating the dehydrogenated material under 100 MPa H2 pressure. The PXRD and MAS NMR results support the notion that LiAlH4@NCMK-3 shows unprecedented reversibility under 100 MPa H2 pressure at 50 °C, whereas the LiAlH4@CMK-3 composite cannot be rehydrogenated under those conditions.29 Prior studies have shown that nanoconfinement is as a viable approach to improve the hydrogen storage properties of LiAlH4; however, full reversibility has not been demonstrated.32-35 Fig. 3a displays the PXRD patterns of desorbed and rehydrogenated samples for LiAlH4@CMK-3 and LiAlH4@NCMK-3. The evolution of metallic Al is observed in both N-functionalized and non-functionalized samples, with no other crystalline byproducts formed during the decomposition. However, a prominent difference is observed in the rehydrogenated samples. In contrast to LiAlH4@CMK-3, the N-functionalized LiAlH4@NCMK-3 shows broad reflections of LiAlH4 after the rehydrogenation, as well as some residual Al metal peaks, demonstrating reversible formation of LiAlH4 from LiH and Al under high hydrogen pressure.
The most conclusive evidence for the reversible H2 uptake and release from LiAlH4@NCMK-3 was obtained from MAS NMR, a technique which provides information about all species present in the sample, including the amorphous ones. Typical 27Al MAS NMR peaks for the tetrahedral Al in LiAlH4 are observed at 98 ppm in both as-synthesized LiAlH4@NCMK-3 and LiAlH4@CMK-3 indicating successful hydride incorporation (Fig. 3c-e).36 After H2 desorption, Al peaks from four-coordinated LiAlH4 (and the remaining six-coordinated Li3AlH6) disappear in both samples, indicating complete decomposition into LiH and metallic Al. Upon rehydrogenation, the LiAlH4 peak at 98 ppm is only observed in LiAlH4@NCMK-3, which confirms that the rehydrogenation reaction only occurs in LiAlH4@NCMK-3, in agreement with the PXRD results. These results confirm that metallic Al is the only Al-containing decomposition product upon desorption (Fig. 3b) since peaks for the intermetallic LiAl decomposition products that occur at 380 and 176 ppm are not detected. The corresponding 7Li NMR data (Fig. S9) also indicate that the regeneration proceeds from LiH to LiAlH4 without any indications of Li3AlH6 or intermetallic LiAl phases. Encouragingly, the integration of the 27Al MAS NMR peaks reveals that during rehydrogenation ~80% of the initially infiltrated LiAlH4 is regenerated. In contrast, the 27Al and 7Li MAS NMR results for the LiAlH4@CMK-3 samples reveal the same reaction products (LiH and Al) and impurity peaks (Al2O3), but the LiAlH4 peak is absent after rehydrogenation, confirming the reaction is only reversible when using the NCMK-3 scaffold.
HAADF-STEM images of the rehydrogenated composite show small Al species inside the NCMK-3 scaffold, in addition to the large metallic Al particles. Fig. 3b shows HAADF-STEM images and energy dispersive X-ray spectra (EDS) of the rehydrogenated LiAlH4@NCMK-3. As expected from the PXRD patterns, the large 50-100 nm Al particles are detected outside the pores of the NCMK-3 scaffold. Both Al-rich areas (at the Al particles), and other low-concentration Al areas were investigated for comparison, and the Al EDS signal was also detected from the background area, implying that species other than metallic Al exist inside the pores (Fig. S10), consistent with the 27Al and 7Li MAS NMR results.
Origin of reversibility and altered reaction pathway
Collectively, the experiments suggest three potentially governing physical effects: (1) nanoconfinement of LiAlH4 in CMK-3 and NCMK-3 destabilizes the Li3AlH6 intermediate upon H2 release; (2) Li adatoms form on nitrogen binding sites in NCMK-3 upon infiltration and affect the subsequent chemical reactivity of LiAlH4; and (3) both factors contribute to reversible LiAlH4 regeneration in NCMK-3 upon applying hydrogen pressure. Each of these effects was explored further using carefully controlled ab initio simulations, which confirm the experimental observations and offer additional insights into the origins of the observed behavior.
To explore the effect of confinement on the destabilization of Li3AlH6, we generated cluster models with varying sizes to evaluate the stability of all relevant compounds for each size. AIMD of LiH, Al, Li3AlH6 and LiAlH4 nanoclusters ranging from one to eight formula units were run, with the larger sizes approximating the real dimensions of the nanoconfined system (Fig. S11). As shown in the equilibrium structure in the inset of Fig. 4a, the AlH63- units that are the building blocks of bulk Li3AlH6 become intrinsically unstable at the nanoscale, converting instead into AlH4- units. In contrast, the other tested species at the nanoscale retain their general molecular structure. A closer examination of the integrated radial atomic distributions reveals that the Li3AlH6 clusters of various sizes (1 to 8 formula units) consists of AlH4- units segregated on the surface and a core comprised of an atomic-scale mixture of LiH and Al (Fig. 4a). Spontaneous atomic-scale mixing of LiH and Al was further confirmed by complementary simulations of individual clusters in close proximity (Fig. S12).
The Li3AlH6 destabilization at the nanoscale could be a response to the breakdown of intermolecular cohesion. To probe this, we computed the formation energy of each compound cluster from an assembly of individual molecular formula units (i.e., cluster size of unity). As shown in Fig. 4b, much larger clusters of Li3AlH6 are required to recover the stabilizing effect of bulk cohesion compared to the other reaction compounds. We can therefore conclude that the Li3AlH6 phase is significantly destabilized upon nanosizing, consistent with the lack of any indication of Li3AlH6 in experimental data. Instead, the system bypasses Li3AlH6, and hydrogenation and dehydrogenation cycles only involve LiAlH4 and LiH phases.
This inhibition of the Li3AlH6 phase is an important key for understanding the hydrogen release from LiAlH4@NCMK-3. We conclude that the two-step decomposition of LiAlH4 is attributable to hydrogen release from nanoconfined LiAlH4 in various chemical environments inside pores of NCMK-3. The overall process can be represented as LiAlH4 → LiH+Al+3/2H2 based on the following observations. Unlike other observed desorption profiles which show gradually increasing desorption rate, the first decomposition step of LiAlH4@NCMK-3 displays fast linear desorption, indicating hydrogen release from LiAlH4 nanoparticles (Fig. 2a). In addition, we rule out the possibility that the second release is the decomposition of Li3AlH6 because there are no indications of a stable intermediate Li3AlH6 phase, including the highly sensitive 27Al MAS NMR technique. This confirms the prediction that the H2 desorption of LiAlH4 nanoparticles in LiAlH4@NCMK-3 bypasses the stable Li3AlH6 phase. Notably, this is similar to the observed decomposition behavior of NaAlH4 at nanoscale.31,37
Another factor potentially in play is the nature of the interaction between LiAlH4 and carbon scaffolds. We ran additional AIMD simulations of LiAlH4 clusters on planar pristine graphene and pyridinic N-doped graphene sheets to approximate scaffold walls in CMK-3 and NCMK-3 (Fig. S13). Although there is evidence for graphitic and pyrrolic defects in the scaffold (Fig. S1), we focused on the pyridinic defects to explore the interaction with LiAlH4 since the higher degree of excess charge localization on pyridinic defects is likely to have a greater overall impact on the reaction. On pristine graphene (broadly representative of CMK-3), the AIMD simulations reveal little change in the structure of the LiAlH4 cluster. However, in the presence of the pyridinic N defect, Li atoms are drawn away from the cluster towards the graphene surface to coordinate with N. Using Li energy references derived from bulk state LiAlH4, LiH, Al and gaseous H2, it is determined that the Li coordination reaction is exergonic and thus irreversible after initial infiltration (Fig. S13). This coordination also anchors the cluster to the surface, suggesting a strong interaction and indicating an impact well beyond the single Li adatom involved in the Li-N bond formation (Fig. 4c-d).
To further investigate the nature of the Li-N interaction, we analyzed the electronic structure of pristine graphene and pyridinic N-doped graphene models in the presence of the (LiAlH4)6 clusters (Fig. 4e). On pristine graphene, the addition of LiAlH4 has little effect on the electronic density of states (DOS). On the other hand, if pyridinic N is present, the localized state associated with the N atom exhibits a pronounced chemical change upon incorporation of LiAlH4, delocalizing its charge. This different behavior of the two model carbon hosts can be rationalized in part based on the low DOS of graphitic derivatives in the vicinity of the Fermi level, which requires a large energetic penalty for charge redistribution. The presence of pyridinic N removes this DOS limitation, permitting the host to take a more direct role in the process. The positive impact of N on the otherwise low graphitic DOS has been invoked to rationalize improved kinetics in other carbon-confined hydrogen storage materials,24 and closely parallels the known role of N in overcoming DOS limitations in carbon materials for electrochemical energy storage.38
Notably, the dominant effect of LiAlH4 incorporation on the DOS of N-doped graphene is well captured by the simple approximation of single adsorbed Li atom on that surface (Fig. S14). This confirms that the Li-N interaction dominates the essential physics, and that Li atoms are indeed drawn away from the primary cluster (see Fig. S14 for a corresponding map of the charge density redistribution). Using this simple Li-adsorbed model, we were also able to reproduce an observed shift in the experimentally measured X-ray absorption spectrum (XAS) of N K-edge in neat NCMK-3, as-synthesized, desorbed, and rehydrogenated LiAlH4@NCMK-3. As shown in Fig. S15, the predicted Li-N chemical interaction leads to an energy shift in the computed XAS results that is quantitatively consistent with the measured spectra. This same XAS energy shift is detected under all measured conditions, supporting the conclusion that Li-N coordination is an irreversible process and confirming that Li adatoms near pyridinic N defects introduced during infiltration remain bound in the following (de)hydrogenation cycles.
The retention of Li at the N sites is expected to further impact the reaction cycle, including reversibility. To explore this effect further, a new set of models containing one Li atom anchored to each N site was generated, with additional (LiH+Al)6 and (LiAlH4)6 incorporated separately as the reactant and product, respectively. From Fig. 4f, the presence of the anchored Li adatoms on the N sites significantly alters the structural configurations of the clusters, enhancing spreading along the surface to maximize the Li-N interactions. Based on these models, the thermodynamic impact was assessed. Note that with the small cluster sizes accessible within our model (chosen to facilitate manageable structural sampling while capturing the essential effect of cohesive interactions on phase stability), all computed ΔEhydriding are exothermic regardless of the presence or nature of hosts. However, a more meaningful comparison can be obtained by instead examining the effect on the hydriding reaction energy, ΔEhydriding, for a given cluster size. For the (LiH+Al)6 to (LiAlH4)6 conversion under H2, the presence of Li adatoms on N defect sites lowers ΔEhydriding by 8 kJ mol-1 H2 compared to pristine graphene (Fig. 4f), and by 4 kJ mol-1 H2 compared to unsupported clusters in vacuo. This trend should also hold for larger cluster sizes, implying that H2 uptake is more energetically favorable in the presence of N dopants. Notably, this is consistent with the observed difference in hydriding behavior between the CMK-3 and NCMK-3 confined materials, confirming the role of the Li adatom formation on N binding sites and resulting charge redistribution on reaction thermodynamics.
Conclusion and outlook. We demonstrate here a conceptually new strategy for altering the reaction pathways and energetics of chemical processes at nanoscale. The thermodynamic properties of metastable LiAlH4 are significantly modified by confinement in nanochannels of N-doped CMK-3. Our experiments and calculations reveal that the N-functionalities on the NCMK-3 scaffold promote low-temperature dehydrogenation of LiAlH4 to form Al particles and LiH, bypassing the stable Li3AlH6 intermediate phase observed in bulk. Remarkably, the material obtained after hydrogen desorption from LiAlH4@NCMK-3 can be regenerated back into LiAlH4 in 80% yield, which has thus far been regarded as not feasible. DFT and AIMD modelling results indicate that the contribution of cluster cohesion energy is far weaker for (Li3AlH6)n than for (LiAlH4)n clusters (n = 1–8), leading to the bypassing of the intermediate Li3AlH6 phase. Furthermore, it was revealed that N defect sites can lower the energy of hydriding (ΔEhydriding) by interacting with LiAlH4 clusters more intimately than in pristine carbon, changing the DOS in the vicinity of the Fermi level and effectively acting as solvation sites for lithium ions.
Strict control over the orientation of hydrides at the nanoscale can enable precise placement of electron acceptor species in the vicinity of electron donor species within the porous host to optimize the promotional effects observed, accelerate the H2 sorption kinetics, and tune the energetics of chemical reactions. As the results in this work indicate, the creation of such a “designed metal hydride-functionalized scaffold composite” and the optimal tuning of its properties can address important materials science challenges in energy generation and storage and can lead to a better understanding of confinement effects in nanoscale volumes.39