Chemical Affinity of Ag-Exchanged Zeolites for Efficient Hydrogen Isotope Separation

We report an ion-exchanged zeolite as an excellent candidate for large-scale application in hydrogen isotope separation. Ag(I)-exchanged zeolite Y has been synthesized through a standard ion-exchange procedure. The D2/H2 separation performance has been systematically investigated via thermal desorption spectroscopy (TDS). Undercoordinated Ag+ in zeolite AgY acts as a strong adsorption site and adorbs preferentially the heavier isotopologue even above liquid nitrogen temperature. The highest D2/H2 selectivity of 10 is found at an exposure temperature of 90 K. Furthermore, the high Al content of the zeolite structure leads to a high density of Ag sites, resulting in a high gas uptake. In the framework, approximately one-third of the total physisorbed hydrogen isotopes are adsorbed on the Ag sites, corresponding to 3 mmol/g. A density functional theory (DFT) calculation reveals that the isotopologue-selective adsorption of hydrogen at Ag sites contributes to the outstanding hydrogen isotope separation, which has been directly observed through cryogenic thermal desorption spectroscopy. The overall performance of zeolite AgY, showing good selectivity combined with high gas uptake, is very promising for future technical applications.

TDS Procedure details for hydrogen isotope separation. The sample chamber is in an ultrahigh vacuum (UHV) at RT, and cool down to aimed temperature at 25 K, 40 K, 60 K, 77 K and 90 K, respectively. Then, the sample is exposed to a defined 1:1 D 2 /H 2 mixture atmosphere (3,10,30 and 60 mbar) for 10 min. After 1:1 D 2 /H 2 mixture loading at the given exposure temperature and pressure, the gas molecules that had not been adsorbed were pumped out. Afterwards, the sample is rapidly cooled to the boiling temperature of the adsorbed gas. Finally, a linear heating ramp (0.1 K·s -1 ) is applied in order to thermally activate desorption. The desorbing gas is continuously detected using a quadrupole mass spectrometer, recognizing a pressure increase in the sample chamber when gas desorbs. The area under the desorption peak is proportional to the desorbing amount of gas.
Calibration of the mass spectrometer signal. A solid piece of a diluted Pd alloy Pd 95 Ce 5 (~0.5 g) was used for calibration. Before the calibration, the oxide layer of the alloy was removed by etching with aqua regia. Then the alloy was heated up to 600 K under high vacuum to remove any hydrogen that might be absorbed during the etching procedure. Afterwards, it was exposed to 40 mbar pure H 2 or pure D 2 for 1.5-2.5 h at 350 K after the mass had been collected. As H and D were bound preferentially to the Cerium atoms at low exposure pressures, the alloy could be handled under ambient conditions for a short time. The alloy was weighed after being cooled down to room temperature. The mass difference between unloaded state and loaded state was equal to the mass uptake of hydrogen or deuterium, respectively. After weighing, the alloy was loaded in the chamber again, and then a 0.1 K·s -1 heating ramp (RT to 600 K) was applied for a subsequent desorption spectrum. The obtained mass of gas is directly corresponded to the area under the desorption peak.

Evaluation of the desorption energy
Desorption energy can be determined via Kissinger method 1 by employing desorption spectra recorded with different heating rates. After exposure to a 10mbar H 2 or D 2 pure gas at 60K the thermal desorption spectra have been recorded applying different heating rates of 0.01, 0.05, 0.1 and 0.2 K·s -1 . The desorption maxima for both H 2 and D 2 show a clear shift to lower temperatures for slower heating rates indicating a thermally activated desorption process.
For an adsorbed gas molecule, the desorption process can be described by the Polanyi-Wigner equation for first order with the assumption that the pre-exponential factor and the desorption energy are independent of coverage: where is the referred to as the surface coverage, ν is the frequency factor, is the activation energy for desorption, is referred to the gas constant and the measured temperature.
Considering the assumption that the surface coverage at the maximum temperature is independent of the heating rate, the temperature of the maximum desorption rate = is determined. The activation energy derivated from a plot vs. ,
For , the following condition must be fulfilled, = The equation 1 can be rewritten as For the first-order desorption process ( ), substituting by equation 1, and solving for = 1 Desorption energy can be obtained from the slope by plotting vs. for a series ln ( 2 ) 1 of heating constant.          In the reference model, the remaining vibrational modes are almost identical irrespective of whether H₂ or D₂ is adsorbed at the Ag⁺ site. Overall, these modes slightly favor the adsorption of D₂ over that of H₂, thereby cancelling out some of the above overestimation.
We conclude that the additional error incurred by the minimal vibrational treatment on the quantities relevant for the calculation of the selectivities, namely the differences between the adsorption energies of D₂ and H₂ should be very low and on the order of 0.1 kJ·mol⁻¹.
Computational level for single-point calculations. Previous experience 4,23 shows that PBE0-D3(BJ) combined with typical triple-or quadruple-zeta basis sets is a robust method to estimate H₂ adsorption energies when the interaction is dominated by charge transfer effects. Ag-Y with an adsorption energy in the 30…40 kJ·mol⁻¹ range would be considered to be such a system.
However, the introduction of water leads to a considerable reduction of the H₂ adsorption energy for some structures and dispersion interactions make up a larger share of the total interaction energy. A correlated wavefunction approach, which natively includes dispersion interactions therefore becomes desirable and this is our rationale for resorting to the DLPNO-CCSD(T) method. While TightPNO cutoffs would be desirable for very weak interactions, these cutoffs are prohibitively resource-consuming for the systems at hand and the accuracy afforded by NormalPNO should be sufficient given the strength of the interaction. With NormalPNO cutoffs, the generous def2-QZVPP basis is still within reach for the systems with up to one molecule of water. Where possible, we compare this larger basis set with the more economical def2-TZVPP basis set. Like in prior studies we note that the differences between the reaction energies are small and def2-TZVPP predicts slightly weaker binding with a maximum deviation of +3.0 kJ·mol⁻¹, a minimum deviation of -0.6 kJ·mol⁻¹ and average and root-meansquare deviations being -1.6 and 1.9 kJ·mol⁻¹, respectively (see Figure S10). We explain this effect with a comprehensive error cancellation between basis set superposition error (overestimating the strength of the interaction) and basis set incompleteness error (underestimating the strength of the dispersion interaction), the latter apparently being slightly dominant. From a practical point of view, this justifies the use of def2-TZVPP.  Figure S11 shows the correlation between PBE0-D3(BJ)/def2-QZVP and DLPNO-CCSD(T)/def2-TZVPP single-point energies. It is apparent that the former overestimates the interaction by 5…10 kJ·mol⁻¹ with respect to the latter. The error appears to be rather systematic, so we expect this DFT method to be a robust predictor of trends for the system at hand where DLPNO-CCSD(T) becomes too expensive.