Introduction

The global shift towards renewable and clean energy sources has intensified the importance of hydrogen (H) and its isotopes, deuterium (D) and tritium (T)1. However, separating these isotopes remains a significant challenge due to their similar physicochemical properties2. Recently, porous materials, particularly metal-organic frame-works (MOFs), have garnered attention for their potential in gas separation3.

MOFs are ordered three-dimensional structures comprising metal ions and organic linkers4. They offer an energy-efficient alternative to traditional separation methods, potentially reducing consumption by up to 90%5. For hydrogen isotopes, MOFs facilitate separation through two quantum sieving mechanisms: kinetic quantum sieving (KQS) and chemical affinity quantum sieving (CAQS)6. KQS occurs in MOFs with small pore apertures, where deuterium’s faster diffusion relative to hydrogen under cryogenic conditions is attributed to its shorter de Broglie wavelength7,8. CAQS, on the other hand, occurs at strong adsorption sites within the MOFs, where heavier isotopes exhibit stronger interactions due to lower zero-point energy (ZPE)9,10. This has led to promising developments in hydrogen isotope separation using MOFs, achieved by finely tuning the pore size11,12, binding strength13,14, and framework flexibility15, which include phenomena such as breathing transitions and gating effects16.

Growing attention is directed towards the strategic development of triazolate-based frameworks with targeted pore size, particularly given their structural tuneability17,18. The three-dimensional, 4-connected dia-type coordination framework [M(ta)2] (where M represents MnII, FeII, CoII, CuII, ZnII, CdII, and CrII/III), is constructed using 1H-1,2,3-triazole (H-ta) ligands and metal ions19,20. This structure embeds a dia topology pore system, comprising small cavities and narrower connecting channels. Extra pore-surface pockets formed by six triazole ligands connect to these channels with smaller entrance diameters21,22, These isostructural frameworks have undergone extensive study in which diverse properties and phenomena have been explored such as Jahn-Teller and spin-crossover phase transitions23,24, magnetic properties19,25, and electrical conductivity26,27. It has been revealed that the [Fe(ta)2] framework, when exposed to inert gases like Ar or CO2, exhibits conductivity changes, primarily due to alterations in the MOF’s deformation potential caused by guest loading at specific sites28,29. Furthermore, the modular nature of [M(ta)2] allows for accurate pore size adjustments by substitution of metal ions with different radii, including MnII and ZnII, resulting in the size control of features within 0.01 nm and influencing the interaction with gas molecules. Therefore, the tuneable pore size and quadrupole moments within [M(ta)2] make it particularly suitable for H2/D2 separation22.

Extensive research has been conducted on the adsorption of various gases, including Ar, CO2, and N2, in these frameworks30,31. These studies have yielded insights into the mechanisms of adsorption, successfully pinpointing different adsorption sites and their occupancy rates, and capacity. However, despite extensive studies, identifying H2 adsorption sites within this particular MOF remains challenging. Computational potential energy surface (PES) analysis suggests that hydrogen molecules may localize both inside and outside the framework’s pockets22. Despite extensive characterization of gas adsorption behavior in triazolate-based frameworks, a critical knowledge gap exists regarding the molecular-level understanding of hydrogen isotope interactions and the framework’s structural response to these interactions.

Herein, we investigated the penetration of hydrogen molecules into these pore-surface pockets by employing neutron powder diffraction (NPD) using deuterated [Mn(ta)2]. Our in situ diffraction experiments during H2 and D2 dosing at various temperatures revealed two adsorption sites in the void space of the material: one at the side pockets and the other within the channels. These experiments demonstrated that gas loading induces expansion of the unit cell volume, with a more pronounced effect observed for H2 compared to D2 for the same adsorbed amount. We further investigated the direct separation of hydrogen isotopes by cryogenic thermal desorption spectroscopy (TDS). In this study, we have gained detailed molecular-level information on H2 and D2 binding in [Mn(ta)2] and studied the structural dynamics around adsorption sites and their properties, allowing us to reconcile these details with TDS and NPD results. These characterization efforts provide clear insights into the molecular aspects of H2/D2 adsorption and selectivity in this Mn-triazole MOF, which will help in designing materials with the desired properties for efficient isotope separation.

Results and discussion

Materials and characterization

High-quality crystals of [Mn(ta)2] and [Mn(ta-d2)2] were synthesized according to literature procedures, using 1H-1,2,3-triazole (H-ta) and (4,5-D2)−1H-1,2,3-triazole (H-ta-d2) ligands in a 3:1 ligand-to-metal ratio21. The deuterated version of [Mn(ta)2] was used for NPD experiments to suppress the incoherent neutron scattering arising from the H in the framework that obscures the coherent neutron scattering important to obtaining structural information, allowing for accurate determination of hydrogen binding sites. X-ray powder diffraction (XRPD) data of both compounds align well with existing literature (see Supplementary Figs. 1, 2) indicating the material is phase-pure, and scanning electron microscopy (SEM) images reveal the typical octahedral shape of the M(ta)2-type framework crystals (Supplementary Figs. 3, 4). Fourier transform infrared spectroscopy (FT-IR) shows significant differences between the spectrum of [Mn(ta)2] and [Mn(ta-d2)2], attributable to the heavier deuterium isotope (Supplementary Figs. 5–7). These differences align with those in the spectrum of [Zn(ta)2] and [Zn(ta-d2)2] simulated using the code CASTEP, as similar to previous work and confirming successful synthesis of the material with the deuterated ligand32,33. Weak additional bands in the [Mn(ta-d2)2] spectrum associated with the non-deuterated and mixed-isotope ligand are in fair agreement with the purity of the commercially available H-ta-d2 ligand. Successful activation of [Mn(ta-d2)2] was confirmed by thermogravimetric analysis (TGA), see Supplementary Fig. 8, which involved removing solvent molecules from the synthesis to prepare the material for gas adsorption. This analysis revealed no residual solvent, demonstrating the robustness of the material’s framework to desolvation.

Gas adsorption properties

Isothermal gas adsorption by Mn(ta)2 exhibited a two-step process for both N2 and Ar (Supplementary Figs. 9, 10). For N2, saturation was observed at 125 and 246 cm3/g (STP, standard temperature, and pressure), corresponding to 1.0 and 1.9 N2 molecules per Mn, respectively. Similarly, for Ar, saturation occurred at 125 and 265 cm3/g (STP), corresponding to 1.0 and 2.1 Ar molecules per Mn, respectively. These results indicate a consistent two-step uptake process for both gases, aligning with values previously reported. In contrast, H2 and D2 adsorption at various temperature (30 – 77 K) exhibited a single-step process, i.e., a type I adsorption curve that is typical for nanoporous materials. These type I H2 isotherms at 77 K have been previously observed in metal-triazolates22. Saturation was observed at 15.8 and 16.0 mmol/g for H2 and D2, respectively, corresponding to ~3 H2 molecules per Mn and 3 D2 molecules per Mn at 30 K (Fig. 1a, b, Supplementary Fig. 11). At all temperatures, D2 uptake was slightly higher than H2, indicating preferential adsorption of D2. Interestingly, the isotherms showed a similar uptake at 1 bar when the temperature increased from 30 to 60 K, decreasing to only 2 H2(D2)/Mn at 77 K. This behavior of [Mn(ta)2] is uncommon, as most other materials typically show a significant decrease in uptake at 1 bar with increasing temperature. To understand this unusual behavior, the isosteric heat of adsorption was calculated using the Clausius-Clapeyron equation applied to the adsorption isotherms. The obtained adsorption heat ranged from 4.8 to 5.8 kJ/mol for H2, while for D2, a near-constant heat of adsorption of ~6.5 kJ/mol was observed, increasing with surface coverage, which refers to the fraction of the surface occupied by adsorbate molecules (Fig. 1c, d). The gradually increasing adsorption energy suggests that the uptake of H2/D2 becomes more favorable at higher loadings, which is contrary to the usual trend. The nearly constant heat of adsorption across the entire gas loading range indicates possible structural changes upon adsorption, exposing more favorable binding sites.

Fig. 1: H2 and D2 adsorption properties.
figure 1

a Hydrogen and (b) deuterium in [Mn(ta)2] at various temperatures (30–77 K) in the pressure range of 0–1 bar (adsorption and desorption, full and open points, respectively). Lines through the points are a guide to the eye. The isosteric heat of adsorption of (c) hydrogen and (d) deuterium in [Mn(ta)2] as a function of the hydrogen/deuterium gas uptake. Source data are provided as a Source Data file.

Full size image

In situ high-resolution neutron powder diffraction

To elucidate this unusual adsorption behavior, comprehensive in situ high-resolution NPD experiments were performed using the Echidna instrument at the Australian Centre for Neutron Scattering (ACNS) at the Australian Nuclear Science and Technology Organisation (ANSTO). These experiments provided a better understanding of the material’s site-specific interaction with H2 and D2. Rietveld refinement using the diffraction data for the activated material confirms that [M(ta)2] [Mn(ta-d2)2] crystallizes with cubic symmetry with a Fd\(\bar{3}\)m space group (No. 227) (Supplementary Tables 1, 2). N donor atoms octahedrally coordinate each Mn from surrounding 1,2,3-triazole to form a diamond-type framework. Mn(1) occupies the 8b Wyckoff site (symmetry \(\bar{4}3\)m) and is coordinated only by N donor atoms of the μ3-bridging triazolate linker. Mn(2) is situated at the 16d Wyckoff site (symmetry \(\bar{3}m\)), bonded exclusively to N1 or N3 donor atoms (Supplementary Figs. 12, 13). The pore network of the determined structure runs along the [110] direction, aligning well with that reported from XRPD data of non-deuterated samples21,22.

bvAfter Rietveld refinement of the host structure, the sample was dosed with 15 mmol/g of D2 and H2 gas at 40 K and further NPD data were collected at 15 K. Rietveld refinement of the structure in conjunction with the examination of the residual nuclear density arising from the long-range ordering of adsorbed H2 or D2 using Fourier-difference methods enabled the determination of the position of adsorbed H2 or D2 in the framework structure. This revealed two adsorption sites, as depicted in Fig. 2 and Supplementary Figs. 14–16. Site 1 is located inside the pockets formed by the ta-ligands, with the closest framework atoms being the D atoms of the ligand. Site 2 is situated in the pore channel along the [110] direction. Each site, with a multiplicity of 192, can accommodate a maximum of 32 D2/H2 molecules per unit cell, leading to a structure with an orientational disorder of the linkers when saturated with D2 or H2 (Supplementary Fig. 17). The occupation of D2 or H2 at both Site 1 and Site 2 in the saturated material is 1.4 H2 (D2) per Mn, consistent with the 3 H2(D2) per Mn at saturation reached in the adsorption isotherms. The closest distance between adsorbed H2 and the framework at Site 2 is ~3.48 Å, while that for H2 at Site 1 is ~2.48 Å (the distance between the nearest D atom of the ta ligand and the center of mass of the adsorbed H2) (Supplementary Table 3). Relative to Site 1, the larger distance between the host and H2 at Site 2 relative to Site 1 is typical of a weaker interaction. Moreover, the distance between hydrogen molecules adsorbed at Site 1 and 2 lies in the range 3.63 – 3.84 Å. In the case of D2, the adsorption sites are identical to those for H2.

Fig. 2: High-resolution neutron powder diffraction.
figure 2

a Rietveld refinement profile using NPD data, collected at 15 K for [Mn(ta-d2)2], loaded with 15 mmol/g hydrogen at 40 K, where contributions from the vanadium sample holder were excluded from the refinement. Figures of merit shown include the goodness of fit (GOF) and the weighted profile R-factor (Rwp). a Entire data range; b Low-d section of data. c Schematic of the refined [Mn(ta)2] structure showing molecular hydrogen at the two adsorption sites in the framework, viewed along the [110]. Mn atoms are shown as magenta spheres, the Mn-N polyhedra as light magenta, the ta-ligand as gray five-membered aromatic rings with H excluded for clarity, and adsorbed H2 molecules as blue spheres where Site 1 is denoted Hads1 and Site 2 is denoted Hads2, (d) view along the [100], (e) refined short-range average structure of Hads1 viewed along the [111]. Source data are provided as a Source Data file.

Full size image

In addition to the two adsorption sites for H2 and D2 identified, we note an expansion of the [Mn(ta-d2)2] crystal volume, with the material saturated with H2 and D2 being ~1.2% and 1%, respectively, larger than for the empty material. Specifically, the lattice parameter expands from 18.08 to 18.14 Å upon adsorption of D2, which is slightly less than the expansion caused by H2, increasing from 18.08 to 18.15 Å. This smaller lattice expansion for D2 suggests more favorable adsorption of the heavier guest molecules, likely due to the stronger interactions between the D2 molecules and the adsorbent.

The structural dynamics observed in the Mn(ta-d2)2 framework upon hydrogen loading are in good agreement with the structure predicted using density functional theory (DFT). The introduction of H2 molecules into the framework induces a moderate expansion of both lattice parameters and unit-cell volume, detailed data is provided in the Supplementary Section Density Functional Calculations. The refined structure of the empty framework is detailed in Supplementary Table 4 and Supplementary Fig. 19 depicts the progressive increase in volume associated with hydrogen loading. The empty unit-cell volume expands by ~1.3% upon uptake of 18.0 mmol/g H2, in good agreement with the isotherm and NPD data. Though this value is slightly higher than the experimental uptake, it likely reflects the assumption of a perfect-crystal model in the theoretical framework. This expansion primarily originates from an elongation of ~ 0.65% in the manganese-nitrogen bond lengths, while the bond lengths within the triazole rings remain unchanged, as shown in Supplementary Fig. 20.

Time-resolved neutron powder diffraction

To further elucidate structural dynamics in the framework as a function of H2 and D2 loading and temperature, time-resolved NPD data were collected using the high-intensity neutron diffractometer Wombat at the ACNS. Data were collected every minute consecutively during isothermal gas adsorption and desorption. Datasets for H2 and D2 adsorption and desorption were recorded at 30 and 60 K and confirmed the two adsorption sites previously identified using NPD data collected under equilibrium conditions. The rapid collection of diffraction data during adsorption and desorption processes revealed distinct adsorptive behaviors for H2 and D2 at the specified sites, as shown in Fig. 3. At a loading temperature of 30 K, as shown in Fig. 3a, b, H2 molecules were observed to simultaneously occupy both adsorption sites following a 2 mmol/g loading. Similar adsorption behavior was noted for D2, with occupancy at both Site 1 and Site 2. The unit cell volume of [Mn(ta)2] at 30 K is plotted as a function of H2 and D2 loading in Fig. 3c. The data reveal an expansion in the unit cell with increasing gas loading, with the framework unit cell volume expanding by ~1.3% upon adsorption of 13 mmol/g of H2, with a slightly lesser expansion for D2 (1.1%).

Fig. 3: Gas induced site occupancy and framework expansion.
figure 3

(a–f) Data collected at 1-min intervals during isothermal gas adsorption and desorption. Top row (a–c): Results at 30 K; bottom row (d–f): Results at 60 K. Left column (a, d): Occupation of H2 at Site 1 (blue, pockets) and Site 2 (red, channels) as a function of total H2 uptake. Middle column (b, e): Occupation of D2 at Site 1 (blue, pockets) and Site 2 (red, channels) as a function of total D2 uptake. Right column (c, f): Unit cell volume expansion of [Mn(ta)2] as a function of total gas uptake for H2 (purple) and D2 (light blue). Error bars are smaller than the symbol size. Source data are provided as a Source Data file.

Full size image

At 60 K, [Mn(ta)2] exhibits distinct adsorption mechanisms for H2 and D2. For H2, a sequential pore-filling process is observed: initially, H2 molecules exclusively occupy Site 1 (smaller pockets) at low loadings. Occupation of Site 2 (channels) commences only after reaching 8 mmol/g (~1.5 H2/Mn), with adsorption continuing until near-saturation at ~3 H2 molecules per Mn. In contrast, D2 adsorption occurs simultaneously at both Site 1 and Site 2, with Site 2 reaching 50% occupancy when Site 1 is fully saturated at 8 mmol D2/g (Fig. 3d, e).

The unit cell volume evolution of [Mn(ta)2] at 60 K (Fig. 3f) displays a similar profile to that observed at 30 K, despite the different adsorption mechanisms. The framework undergoes a significant expansion for both isotopes, with the unit cell volume increasing by ~1.35% upon adsorption of 11 mmol/g of H2. D2 induces a slightly smaller expansion of 1.2% at an equivalent loading. Analysis of the adsorption data reveals that this expansion is predominantly driven by H2/D2 occupation at Site 1, with minimal volume changes occurring during the occupation of isotope gases at Site 2. This site-specific expansion behavior elucidates the slightly larger overall expansion observed at 60 K compared to 30 K. The differential impact of Site 1 and Site 2 occupation on framework expansion provides crucial insights into the structural dynamics underlying the adsorption process and may contribute to the material’s isotope separation capabilities. However, the exact correlation between site occupation and lattice expansion remains to be fully elucidated. As evidenced in Fig. 2, the adsorption sites within the framework exhibit distinct geometries and short-range environments, likely contributing to the complex interplay between gas loading, site occupation, and structural deformation.

These findings highlight the intricate nature of gas-framework interactions in [Mn(ta)2] and their potential implications for hydrogen isotope separation. The subtle difference in adsorption behavior of H2 and D2, coupled with the framework’s structural response, may be the key to understand the material’s high isotope separation efficiency. Further studies employing advanced in situ characterization techniques and theoretical modeling could provide deeper insights into the molecular-level mechanisms underlying these observations and their impact on separation efficiency.

Isotope separation performance

Direct separation of isotopic H2/D2 mixtures was measured using a custom-built cryogenic thermal desorption spectroscopy (TDS) system. We exposed the sample to 100 mbar of 1:1 mixture of H2:D2 for 10 min at various temperatures (Texp) ranging from 30 to 60 K and TDS data was collected between 20 and 110 K (Fig. 4 and Supplementary Fig. 21). In TDS data, the desorption curve area is directly proportional to the number of adsorbed gas molecules, allowing quantification of the amount of adsorbed gas with the D2/H2 selectivity (SD2/H2) derived from the peak area ratio. Figure 4c displays the D2/H2 selectivity and corresponding D2 uptake as a function of Texp, revealing a decreasing uptake but increasing selectivity with temperature, peaking at 60 K (SD2/H2 = 32.5). This value represents one of the highest selectivities reported for porous materials under similar conditions, as confirmed by a comparison with previously reported D2/H2 separation materials (Supplementary Table 5). We also examined the effect of gas pressure on isotope separation at 60 K using a 1:1 mixture of H2:D2 pressures from 10 to 300 mbar (Supplementary Fig. 22), observing consistent preferential adsorption of D2. Exposure of the sample to a 5:95 mixture of D2:H2, reflecting the natural abundance of D2, at 100 mbar at 60 K for 10 min resulted in a selectivity of 74 (Fig. 5), with D2 concentration reaching nearly 75% in just one cycle, highlighting Mn-triazolate’s potential as a highly effective material for industrial deuterium separation.

Fig. 4: Isotope separation TDS measurements.
figure 4

H2 (black) and D2 (red) thermal desorption data during exposure of [Mn(ta)2] to 100 mbar of a 1 : 1 mixture of H2 : D2 for 10 min at various temperatures (Texp): (a) 30 K and (b) 60 K. c The corresponding amount of adsorbed D2 (black) and selectivity (blue) as function of Texp. Lines through the points are a guide to the eye. Source data are provided as a Source Data file.

Full size image
Fig. 5: Separation of diluted isotope mixture.
figure 5

TDS spectra of H2 (black)/D2 (red) after exposure of [Mn(ta)2] to a 100 mbar mixture of 95 : 5 H2 : D2 for 10 min at 60 K. Source data are provided as a Source Data file.

Full size image

Structural dynamics and separation mechanism

To understand the exceptional separation performance of our framework, we conducted additional in situ NPD experiments, using deuterated [Mn(ta)2], at POWGEN, a high-resolution NPD instrument located at the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory (ORNL), focusing on isotope exchange at 40 K. To validate the reliability of the NPD data, we first compared the isotope separation performance of the deuterated sample with that of the non-deuterated sample under identical conditions. The TDS data, illustrated in Supplementary Fig. 23, demonstrate comparable separation performance for both the deuterated and non-deuterated ligand samples. Initially, the framework was loaded with 15 mmol/g H2 at 40 K, showing an average occupancy of 1.4 H2 molecules per adsorption site. We introduced D2 to the framework at the same temperature to observe the exchange behavior between H2 and D2 within the adsorption sites. NPD data were collected continuously during this process allowing for real-time monitoring of the changes in the scattering length density of atoms residing at the two adsorption sites. Hydrogen and deuterium atoms have scattering lengths of opposite sign, and therefore having deuterium substituting hydrogen at the same site would lead to a reduction in the refined nominal hydrogen occupancy. The refinement revealed a significant reduction in the H occupancy, suggesting an increase in scattering length, at both Site 1 and Site 2. Particularly, Site 1 exhibited a more pronounced decrease in H2 occupancy compared to Site 2. This indicates that H2 molecules at Site 1 are more susceptible to replacement by D2 molecules, suggesting a higher propensity for isotope exchange at this site.

This change is also clearly observed in the residual nuclear density map, where a gradual reduction in negative nuclear density indicates the replacement of H2 by D2 (data shown at 30, 60, and 90 min, Supplementary Figs. 24, 25), providing a clear representation of the isotope exchange process. This exchange is significant because strong adsorption sites are typically associated with open metal sites in other materials13. Unlike many other frameworks that rely on undersaturated metal sites for strong adsorption, our framework features fully-coordinated metal sites. Our findings reveal that the confinement within the structure itself, rather than the presence of open metal sites, acts as a potent framework feature for gas adsorption.

In summary, our results reveal a complete and detailed experimental study of both the structural and hydrogen isotope adsorption behavior of the triazolate-based MOF [Mn(ta)2], possessing pores with a diameter of ~6.1 Å. Hydrogen and deuterium adsorption by the material is characterized by a single-step type I isotherm with a saturation uptake of 3 H2(D2)/Mn, while N2 and Ar exhibit a lower uptake and a two-step isotherm. The isosteric heat of adsorption for H2 and D2 remains relatively constant during increasing adsorption. High-resolution and high-intensity neutron powder diffraction identified two distinct adsorption sites for H2 and D2, with Site 1 located at the entrance of the pocket formed by triazole ligands and Site 2 located in the pore channels. Gas loading induces expansion of the unit cell which can be attributed to Mn-N bond elongation in line with theoretical predictions. Time-resolved neutron powder diffraction during in situ gas dosing revealed occupation of both sites by D2 and H2 simultaneously at 30 K. At the higher loading temperature of 60 K, population of Site 1 and 2 by D2 occurs more quickly, with a faster population at Site 1 and further population of Site 2 occurring after saturation of Site 1. In contrast, at 60 K, H2 occupation at Site 2 begins only after Site 1 is fully occupied. Based on this subtle isotopic difference in structural dynamics, our work highlights the potential for developing novel porous frameworks with high selectivity for hydrogen isotopes. Thermal desorption spectroscopy experiments following exposure of the sample to H2/D2 mixtures at 60 K confirm the very high D2/H2 selectivity of 32.5 and separation from a 95:5 mixture of H2:D2 with a 75% recovery of D2 in one cycle. The observed high D2/H2 selectivity coincides with differences in the isotopologue-induced structural dynamics of the framework upon adsorption of these isotopes. While our data indicate a correlation between the slightly lower volume expansion for D2 and its preferential adsorption, attributing the high selectivity solely to the structural dynamics would be an oversimplification. Multiple phenomena, including quantum sieving effects, differences in zero-point energy, and subtle variations in gas-framework interactions (gas-driven unit cell expansion), likely contribute to the observed selectivity. Further studies, such as detailed computational modeling of isotope-framework interactions or isotope-specific vibrational spectroscopy, could provide more direct evidence of the mechanisms underlying this high selectivity. Understanding the interplay between framework confinement, gas-solid interactions, and structural responses remains an important area for future research in isotope separation using metal-organic frameworks. The commercial availability of the ligand and ease of synthesizing dia-topology MOFs with various divalent metals make industrial-scale production and application of these materials for isotope separation feasible.

Methods

Materials synthesis

N,N-Diethylformamide (DEF; 99%; TCI), N,N-dimethylformamide (DMF; 99.8% analytical grade; VWR), methanol (MeOH; 99.8% analytical grade; VWR), 1H-1,2,3-triazole (1,2,3-triazole; 98%; BLD Pharmatech Ltd.), (4,5-D2)−1H-1,2,3-triazole (1,2,3-triazole-4,5-d2, 95%, CombiPhos Catalysts, Inc.), and manganese(II) nitrate tetrahydrate (Mn(NO3)2·4H2O; 98.5%; Merck) were used as received from the commercial suppliers.

[Mn(ta)2] Mn(NO3)2·4H2O (3.20 g, 12.75 mmol), dissolved in 160 mL of DEF, and 1H-1,2,3-triazole (2.40 mL, 2.86 g, 41.41 mmol) were added into a 400 ml inner volume ACE round-bottom pressure flask and capped with a silicon O-ring at the front seal of the polytetrafluoroethylene (PTFE) bushing. The mixture was heated to 120 °C in an oven for 3 days. The product was filtered and washed successively with DMF (3 × 20 mL) and MeOH (3 × 20 mL). Drying overnight under vacuum at room temperature (RT) afforded the phase pure product as a white powder (950 mg, 39%).

[Mn(ta-d2)2] Mn(NO3)2·4H2O (1.17 g, 4.66 mmol), dissolved in 57 mL of DEF, and 1H-1,2,3-triazole-4,5-d2 (1.00 g, 14.07 mmol) were added into a 200 ml inner volume ACE pressure tube and capped with a silicon O-ring at the front seal of the PTFE bushing. The mixture was heated to 120 °C in an oven for 3 days. The product was filtered and washed successively with DMF (3 × 10 mL) and MeOH (3 × 10 mL). Drying overnight under vacuum at RT afforded the phase pure product as a white powder (0.86 g, 63%).

Material characterizations

Thermogravimetric analysis (TGA)

TGA of a 6.207 mg sample of the as-synthesized [Mn(ta-d2)2] was prepared at ambient conditions and measured at the University of Augsburg with a TA Instruments Q500 device in a temperature range of 25–800 °C under a nitrogen atmosphere with a heating rate of 5 °C min–1 starting after a 5 min isothermal step.

Fourier-transform infrared spectroscopy (FTIR)

FTIR spectra were measured at the University of Augsburg under ambient conditions on a Bruker Equinox 55 FT-IR spectrometer equipped with a PLATINUM ATR unit and a KBr beam splitter in the wavenumber range 4000 – 400 cm–1. Spectra in the range 1800 − 180 cm−1 were recorded on the same instrument equipped with a Si beam splitter. The signals were labeled strong (s), medium (m), weak (w), and very weak (vw).

X-ray powder diffraction (XRPD)

XRPD data of the as-synthesized samples were collected at the University of Augsburg under ambient conditions with an Empyrean (PANalytical) diffractometer equipped with a Bragg–BrentanoHD mirror, PIXcel3D 2 × 2 detector using Cu Kα1 radiation. The Mn(ta)2 and Mn(ta-d2)2 samples were measured with a step size of 0.0016° in the 2θ range 3°–60° and 3°–120° and measurement times of 37882 s (10.52 h) and 58445 s (16.23 h), respectively.

Gas sorption isotherm measurements

A fully automated Sieverts apparatus iQ2 (Quantachrome Instruments) was used to perform gas adsorption experiments. High-purity hydrogen (H2, 99.999%) and deuterium (D2, 99.8%) gases were used throughout the experiments. Corrections to the sample volume and the non-linearity of the adsorbate were made using data for an empty analysis carried out at the same temperature and pressure range as for samples. Approximately 30 mg of sample was heated to 393 K under vacuum overnight to remove solvent molecules. A cryocooler based on the Gifford-McMahon cycle was used to control the sample temperature from 20−300 K with an estimated error of <0.05 K.

The isosteric heat of adsorption of H2 and D2 in the samples was independently determined by isothermal gas adsorption and thermal desorption spectroscopy (TDS). H2 and D2 adsorption isotherms were measured in the low-pressure region from 30 K to 77 K. The heat of adsorption was determined by the isosteric method. Data were transferred into Van’t Hoff plots for different coverages, where the isosteric heat of adsorption is defined as the negative of the isosteric enthalpy of adsorption. For a given surface coverage, the isosteric enthalpy of adsorption can be derived from the so-called Van't Hoff equation, analogous to the Clausius-Clapeyron equation. The pressure of equilibrium (Peq) for a fixed amount of adsorbed gas (na) was determined by linear interpolation of the data points of the adsorption isotherms at different temperatures. A linear regression of the slope for the relation ln(Peq) versus 1/T was used to calculate the isosteric heat of adsorption.

Thermal desorption spectroscopy (TDS) studies of hydrogen isotope separation

The selective adsorption after exposure to D2/H2 isotope mixtures was measured by an in-house designed thermal desorption spectroscopy (TDS) system34,35. For a typical process, ~3.5 mg sample was loaded in the sample holder and activated at 393 K under vacuum for 5 h. An equimolar D2/H2 isotope (purity H2 99.999% and D2 99.8%) mixture was exposed to the activated sample at a fixed temperature Texp for an exposure time texp. The free gas was evacuated and the sample cooled to 20 K to preserve the adsorbed state. The sample was then heated to room temperature at 0.1 K/s during which the desorbing gas was continuously detected using a mass spectrometer (QMS), noting a pressure increase in the sample chamber as the gas desorbed. The area under the desorption peak is proportional to the desorbed gas amount, and calibration of the TDS apparatus enabled its determination.

Calibration of the mass spectrometer was carried out using ~0.5 g solid Pd95Ce5 etched with aqua regia to remove any oxide layer. The alloy was heated to 600 K under a high vacuum to remove any hydrogen absorbed during the etching procedure, after which it was exposed to 40 mbar pure H2 or pure D2 over 1.5 – 2.5 h at 350 K, with the mass recorded. As H and D are bound preferentially to Ce atoms at low exposure pressure, the alloy was handled under ambient conditions for a short time. The alloy was weighed after cooling to room temperature, and the mass difference between the unloaded and loaded state being equal to the mass uptake of hydrogen or deuterium, respectively. After weighing, the alloy was loaded in the chamber again and heated at 0.1 K · s−1 from room temperature to 600 K and another desorption spectrum measured in which the area under the desorption peak directly corresponds to the mass uptake of gas.

In situ neutron powder diffraction

In situ neutron powder diffraction (NPD) experiments were conducted using the Echidna and Wombat diffractometers at the Australian Centre for Neutron Scattering (ACNS) at the Australian Nuclear Science and Technology Organisation (ANSTO). Echidna utilized a Ge(335) monochromator at a takeoff angle of 140°, yielding a wavelength of 1.62207(1) Å, with no pre-sample collimation and fixed tertiary 5′ collimation. Wombat used a Ge(115) monochromator at a takeoff angle of 90°, resulting in a wavelength of 1.54298(1) Å, with no pre-sample collimation and tertiary oscillating collimation.

The exact wavelength and instrumental contributions to the data were obtained using data for the National Institute of Standards and Technology (NIST) standard reference material (SRM) 660b (La11B6). Data analysis was carried out using the program TOPAS7, and for both instruments, the peak profile was characterized by a convolution of an axial divergence model with one refinable asymmetry parameter, a constant hat function with a refinable breadth parameter, and a TCHZ-pseudo Voigt function in which the refinable parameters are U, V, W, and Y.

Data were collected for a known mass of polycrystalline desolvated [Mn(ta-d2)2], which was pre-treated at 1 × 10−4 mbar at 110 °C and transferred to a sealed 9 mm diameter vanadium can inside a helium-filled glove box. The sample can with temperature sensors at the top and bottom was attached to a custom-designed gas delivery sample stick, which was positioned within a top-loading cryo-furnace, keeping the sample isolated from air. A Hiden Isochema IMI manometric dosing system was used to control gas delivery/vacuum to the sample, and the top-loading cryostat was used to control the sample temperature. This experimental setup is reported elsewhere36,37,38.

At Echidna, NPD data of [Mn(ta-d2)2] were collected at ~15 K for the evacuated framework and the samples dosed with 2, 5, 7.5, and 15 mmol/g of D2 or H2 gas at all temperatures. During all collections, the top of the sample was ~19 K and the bottom was 15 K. Each collection had a slightly different time length, but most were ~7 h.

At Wombat, NPD data of Mn(ta)2 were collected every minute during adsorption at 30 K, 55 K, and 60 K of D2 and H2. H2 and D2 adsorption isotherms were performed at 30, 55, and 60 K, each consisting of 19 pressure steps in the range 0.2 to 700 mbar. Minimum and maximum wait times for equilibration at each pressure point were 3 and 10 min, respectively. NPD data were collected continuously during the isotherm experiments and recorded at 1 min intervals. Timestamps for the NPD data frames were converted to the corresponding guest uptake amount by linear interpolation of the time-dependent uptake data recorded by the manometric system at irregular 1–20 s intervals.

All NPD data were analyzed using the Rietveld method as implemented in TOPAS7. A starting model for the crystal structure of Mn(ta)2 was obtained from an isostructural compound Fe(II)(ta)2 reported by Grzywa et al.23. For all high-resolution data at 15 K, 6 profile parameters (as used to describe data for the La11B6 standard reference material 660b), the cubic lattice parameter, the zero shift, 6 positional parameters, 5 isotropic atomic displacement parameters for all atoms in the asymmetric unit, and 10 Chebyshev background polynomial parameters were refined in the angular range from 6 to 155° 2θ. In the case of the crystal structure of Mn(ta)2 without any adsorbed gas, the refinement converged quickly. To model the disorder in the H2/D2 loaded samples, the ta-d linker group was constrained to be a rigid unit, allowing only rotational degrees of freedom. Complete refinement parameters, atomic coordinates, and thermal factors are provided in Supplementary Information NPD analysis, Tables 1, 2.

Additional NPD data were collected using the POWGEN diffractometer at the Spallation Neutron Source (SNS), Oak Ridge National Laboratory (ORNL). Approximately 0.6 g of sample was loaded into a 6 mm diameter cylindrical vanadium sample can. Data were collected for ~3 h in high-resolution mode using a center wavelength of 1.5 Å, covering a d-spacing range from 0.50 Å to 12.0 Å. The experiment was carried out using a top-loading JANIS cryofurnace. The unloaded pristine sample was first measured at 6 K to obtain the structure of the framework without adsorbed gas. Gas loading was performed using an in-house built manifold at 40 K, with either pure H2 or D2. After loading, the sample was cooled to 6 K and diffraction data was collected to determine the structure of the gas-loaded framework.

The isotopic exchange measurements were conducted using a sequential gas-loading protocol. First, hydrogen was introduced to the sample at 40 K and allowed to equilibrate for 30 min, establishing a baseline of H2 occupancy throughout the framework. The system was then cooled to 6 K, where initial NPD data were collected to accurately determine the H2 adsorption sites and occupancies. Subsequently, the sample was warmed back to 40 K, and an equivalent amount of D2 gas was introduced to the system. To monitor the progressive isotope exchange process, NPD data were collected at three time intervals (30, 60, and 90 min) after D2 introduction, allowing direct visualization of the replacement of H2 by D2 at specific adsorption sites.

Rietveld analysis of the NPD data was performed using the GSAS-II software package39. The peak profile was modeled using a convolution of a Gaussian peak shape function and a back-to-back exponential function to account for peak asymmetry. The instrumental profile parameters were determined by refining the data from a standard silicon sample (NIST SRM 640 d)40. The background was modeled using a shifted Chebyshev polynomial function with 8 terms.

Density Functional Theory (DFT) calculations

All calculations were performed using density functional theory (DFT) with the Perdew-Burke-Ernzerhof (PBE) functional41, employing Goedecker-Teter-Hutter (GTH) pseudopotentials42 with scalar-relativistic core corrections. Long-range dispersion interactions were accounted for using the D3 London dispersion correction scheme43. A triple-zeta valence plus polarization (TZVPP) basis set was used, with a grid cutoff of 400 Ry for geometry optimization, as implemented in the QUICKSTEP module44 of the CP2K program package45. The framework structures were modeled using periodic boundary conditions in all three dimensions.

The empty framework structure was first optimized with respect to atomic positions and lattice parameters. This optimized structure served as the starting configuration for all loaded structures. Hydrogen molecules (H2) were introduced into the accessible volume of the framework, with initial positions and orientations randomly assigned using the Packmol code. Subsequent geometry optimization was performed, allowing both atomic positions and lattice parameters to relax. The resulting total energy of the loaded structure was recorded. A single-point calculation was then carried out on the emptied structure (after H2 removal) to determine its total energy. This procedure is schematically illustrated in Supplementary Fig. 18 and has been previously described15.

To account for potential variations in adsorption sites and molecular orientations, statistical averaging was conducted over 25 independent configurations, each with randomly assigned initial positions and orientations of the H2 molecules. The deviations in adsorption energies across these configurations were found to be negligibly small.