Selective capture of carbon dioxide from hydrocarbons using a metal-organic framework

Efficient and sustainable methods for carbon dioxide capture are highly sought after. Mature technologies involve chemical reactions that absorb CO2, but they have many drawbacks. Energy-efficient alternatives may be realised by porous physisorbents with void spaces that are complementary in size and electrostatic potential to molecular CO2. Here, we present a robust, recyclable and inexpensive adsorbent termed MUF-16. This metal-organic framework captures CO2 with a high affinity in its one-dimensional channels, as determined by adsorption isotherms, X-ray crystallography and density-functional theory calculations. Its low affinity for other competing gases delivers high selectivity for the adsorption of CO2 over methane, acetylene, ethylene, ethane, propylene and propane. For equimolar mixtures of CO2/CH4 and CO2/C2H2, the selectivity is 6690 and 510, respectively. Breakthrough gas separations under dynamic conditions benefit from short time lags in the elution of the weakly-adsorbed component to deliver high-purity hydrocarbon products, including pure methane and acetylene.


Supplementary Methods: Thermogravimetric Analysis (TGA)
Freshly prepared MOF samples were washed with MeOH, and then activated at 130 °C under vacuum for 10 hours. Samples were exposed to air for one hour and then transferred to an aluminium sample pan. Measurements were then commenced under an N2 flow with a heating rate of 5 °C /min.

Supplementary Methods: Single crystal X-ray diffraction
A Rigaku Spider diffractometer equipped with a MicroMax MM007 rotating anode generator (Cu radiation, 1.54180 Å), high-flux Osmic multilayer mirror optics, and a curved image plate detector was used to collect SCXRD data.

As-synthesized MUF-16, MUF-16(Ni) and MUF-16(Mn)
General MOF crystals were analysed after removing them from methanol. Room temperature data collections produced better refinement statistics than low temperature data collections. All atoms were found in the electron density difference map. All atoms were refined anisotropically, except hydrogen atoms and certain of the water molecules in the pores (as specified below). The structures of solvated MUF-16 1 and MUF-16(Mn) 2 have been reported previously.

MUF-16
O15 of an occluded H2O molecule was refined isotropically. It does not act as a H-bond donor. Despite numerous data collections, the wR2 value remained high due to an inherent lack of precise ordering in the material. A small (1.95 eA -3 ) electron density peak remained near the Co site due to Fourier series truncation ripples.

MUF-16(Ni)
The crystals diffracted to a resolution of just 1.0 Å thus the calculated sin(max)/wavelength is 0.4999.
This limited the number of data and produced a relatively low data: parameter ratio (7.3) and low precision on the C-C bonds. Despite numerous data collections, the wR2 value remained high due to an inherent lack of precise ordering in the material. A small (1.55 eA -3 ) electron density peak remained near the Ni site due to Fourier series truncation ripples. Occluded water molecule O16 does not act as a Hbond donor. A SHEL command (SHEL 8 1) was used to limit the data used in the refinement to values that were sensibly measured.
A solvent mask was calculated and 124 electrons were found in a volume of 308 Å 3 in one void per unit cell. This is consistent with the presence of three disordered water molecules per asymmetric unit, which account for 120 electrons per unit cell. The top of the 0.3 mm capillary was then covered by glass wool to avoid the elutriation of cobalt chloride crystals during activation.
The capillary assembly was then connected to an adsorption apparatus (Quantachrome-Autosorb-iQ2) using appropriate Swagelok fittings (Supplementary Figure 3) and was kept under vacuum and a temperature of 140 °C for around 5 hours so that the vacuum level reached 0.0008 torr. At this point the cobalt chloride crystals were blue in colour (indicating an anhydrous environment).
The capillary was flame sealed to trap the crystal under vacuum. Alternatively, the capillary was filled with CO2 to a pressure of 1.2 bar and then flame sealed.

General
Certain reflections were omitted from the refinement process since they were mismeasured due to the presence of the glass capillary. All non-hydrogen atoms were found in the Fourier difference map.

MUF-16(Mn) in vacuo
The crystals diffracted to a resolution of just 0.90 Å thus the calculated sin(max)/wavelength is 0.555.

MUF-16(Mn) under CO2
The crystals diffracted to a resolution of just 1.08 Å thus the calculated sin(max)/wavelength is 0.463.
This limited the number of data and produced a relatively low data: parameter ratio (5.3) and low precision on the C-C bonds. A SHEL command (SHEL 8 1.08) was used to define the data range for refinement.
A strong electron density peak was observed in the middle of the pore and two weaker areas of electron density towards the pore surface. The central dense area was assigned to be an oxygen (O15) with a while the other two areas were ascribed to oxygen (O16) and carbon (C17) atoms. The C=O bond lengths were restrained to 1. 16

Supplementary Methods: Powder X-ray diffraction patterns
The data were obtained from freshly prepared MOF samples that had been washed several times with MeOH.
MOF crystals were analysed right after removing them from MeOH. The two-dimensional images of the Debye rings were integrated to give 2 vs I diffractograms. Predicted powder patterns were generated from single crystal structures using Mercury.
For aging experiments on the frameworks, after washing as-synthesized samples several times with MeOH, they were activated and were aged in air at 70-85% relative humidity or water at 20 °C.
To measure PXRD patterns under gas loading or in vacuo, a MOF sample was first loaded into a thinwalled glass capillary in a similar way to the SCXRD experiments. It was then activated under vacuum with mild heating before being flame-sealed directly or after back-filling with the selected gas.
Supplementary Figure 6. PXRD patterns of MUF-16 showing that its structure remains unchanged after activation at 130 °C under vacuum, after isotherm measurements, after breakthrough experiments, after exposure to an air with relative humidity of >80% for at least 12 months, after immersion in water for two weeks, after heating to 330 °C (under a N2 flow in a TGA), loaded with CO2 in capillary, and under vacuum in capillary. Figure 7. PXRD patterns of MUF-16(Mn) showing that its structure remains unchanged after activation at 130 °C under vacuum, after isotherm measurements, after exposure to an air with relative humidity of >80% for at least 12 months and after immersion in water for 2 weeks.

Supplementary
Supplementary Figure 8. PXRD patterns of MUF- 16(Ni) showing that its structure remains unchanged after activation at 130 °C under vacuum, after isotherm measurements, after exposure to an air with relative humidity of >80% for at least 12 months and after immersion in water for 2 weeks.

Supplementary Notes: Calculation of BET surface areas
BET surface areas were calculated from N2 adsorption isotherms at 77 K according to the following procedures 9 : 1) The isotherm region where (1 − 0 ⁄ ) increases versus 0 ⁄ , where is the amount of N2 adsorbed, was identified.
2) Within this isotherm region, sequential data points that led to a positive intercept in the plot of 3) The BET surface area was calculated according to:

Supplementary Notes: Heat of adsorption
Isosteric heat of adsorption (Qst) 10 values were calculated from isotherms measured at 293K, 298K and 303 K for CO2. The isotherms were first fit to this viral equation: Where N is the amount of gas adsorbed at the pressure P, a and b are virial coefficients, m and n are the number of coefficients require to adequately describe the isotherm. To calculate Qst, the fitting parameters from the above equation were input in to this equation: Where q is the uptake of a gas; P is the equilibrium pressure and q1, b1, t1, q2, b2 and t2 are constants.
These parameters were subsequently used for the IAST calculations.          * 2 mLN/min of helium was used as carrier gas in this experiment.

CO2/CH4 and CO2/CH4+C2H6+C3H8 breakthrough separations
Activated MUF-16 (0.9 g) was placed in an adsorption column (6.4 mm in diameter × 11 cm in length) to form a fixed bed. The adsorbent was activated at 130 °C under high vacuum for 7 hours and then the column was left under vacuum for another 3 hours while being cooled to 20 °C. The column was then purged under a 20 mLN/min flow of He gas for 1 hr at 1.1 bar prior to the breakthrough experiment. A gas mixture containing CO2/CH4 or CO2/ CH4+ C2H6+C3H8 was introduced to the column at 1.1 bar and 9 bar for CO2/CH4 and CO2/CH4+C2H6+C3H8) and 20 °C.
A feed flowrate of 6 mLN/min was set. The operating pressure was controlled at 1.1 or 9 bar with a backpressure regulator. The outlet composition was continuously monitored by a SRS UGA200 mass spectrometer. The CO2 was deemed to have broken through from the column when its concentration reached 600 ppmv.

Simulations of CO2/CH4 breakthrough curves
The simulation of breakthrough curves was carried out using a previously reported method. 12

CO2/C2 hydrocarbon separations
In a typical breakthrough experiment, 0.9 g of activated MUF-16 was placed in an adsorption column

Adsorbent regeneration
The desorption behaviour of CO2 and C2H2 from the adsorption column was also investigated. Once the adsorbent was saturated with an equimolar mixture of CO2 and C2H2, the column was purged with a helium flow of 5 mLN/min for 18 mins at 20 °C at 1 bar while monitoring the effluent gas. Then the column was then heated to 80 °C with a ramp of 10 °C/min for 20 mins. Finally, the column was heated to 130 °C with the same ramping for 15 min before cooling to 20 °C. A breakthrough measurement was then performed, which showed that the absorbent had been fully regenerated.

Simulations of CO2/C2H2 breakthrough curves
The simulation of breakthrough curves for CO2/C2 hydrocarbons was carried out using the method reported above. A summary of adsorption column parameters and feed characterizations are presented in Supplementary Table 9.

Supplementary Tables: Reported separation metrics
The CO2/CH4 and CO2/C2H2 separation parameters of MUF-16 in comparison to other MOFs and related materials are presented in Supplementary Tables 10 and 11. Materials with molecular sieving mechanisms are excluded from this analysis. IAST selectivities are presented for a 50/50 CO2/CH4 and CO2/C2H2 at 1 bar, unless otherwise stated. Qst values are reported at low loading, unless otherwise stated. Uptake ratios are calculated by dividing the uptake of CO2 by that of CH4 or C2H2 (all at 1 bar and the specified temperature in the tables). These were taken from either a direct statement of relevant details in the manuscript or were extracted from figures by a digitizer software.

Supplementary Methods: DFT calculations
The lowest energy configuration of the CO2 binding sites in the P1 form of MUF-16(Mn) were calculated using density functional theory (DFT) with the software package VASP 5.4.4. 56 One guest molecule was admitted per cell. We implemented dispersion corrections using the DFT-D3 method, 57 as standard DFT methods based on generalized gradient approximation do not fully account for the long-range dispersion interactions between the framework and the bound adsorbate. Electron exchange and correlation were described using the generalized gradient approximation Perdew, Burke, and Ernzerhof (PBE) 58