Carbon Capture Chemistry

Carbon dioxide capture and storage is essential to reducing greenhouse gas emissions and meeting net-zero emissions targets^1,2.

A range of technologies are under development to meet the need for more energy efficient carbon capture. One promising strategy to improve on traditional aqueous amine technology is to use solid adsorbent materials for capture^3,4,5. In particular, installation of reactive amine or hydroxide functional groups within a porous scaffold such as a metal-organic framework or a porous silica brings about selective reactivity with CO₂^6,7,8,9,10, with the porous scaffold providing a large surface area for hosting the reactive groups while maintaining channels for CO₂ transport. The increasingly complex adsorbent materials under consideration bring major challenges in the characterisation of new carbon capture chemistry, hindering the design of improved materials⁴.

Existing characterisation tools for understanding CO₂ capture modes include single-crystal diffraction^11,12,13, powder diffraction¹⁴, infra-red spectroscopy^6,10,15, X-ray absorption spectroscopy¹⁶, and NMR spectroscopy^{8,17,18,19,20}, each of which has strengths and limitations in terms of the materials that can be studied and the information that can be obtained. Solid-state NMR spectroscopy is a promising tool for investigating CO₂ binding modes in adsorbents as there is no requirement for long-range ordering and detailed information about the local structure and dynamics of the CO₂ can be obtained. However, different CO₂ adsorption products often give rise to very similar signals in the NMR spectrum and assigning these signals to specific CO₂ binding modes remains very challenging^{8,17,18,19,20}.

The most common experiment with the NMR approach is to dose the candidate adsorbent with ¹³CO₂ gas and perform ¹³C magic angle spinning (MAS) NMR experiments. These experiments are relatively straightforward to perform, but often lead to ambiguous identification of the adsorption products. For amine-functionalised materials, the ¹³C chemical shifts give poor differentiation between closely related ammonium carbamate, carbamic acid, and ammonium bicarbonate adsorption products, with the signals from these species showing very similar ¹³C chemical shifts^17,18. A similar problem arises for bicarbonate and carbonate products in hydroxide-based materials²⁰. The prediction of NMR parameters with density-functional theory (DFT) calculations²¹ can improve confidence in the structural assignments, and more advanced multi-nuclear NMR experiments can give improved differentiation between adsorption products^17,19,22. However, there remains a pressing need for the exploration of new NMR methods for understanding CO₂ capture chemistry²³.

A representative emerging class of CO₂ adsorbents are amine-functionalised metal-organic frameworks. The framework M₂(dobpdc) (dobpdc = 4,4′-dioxidobiphenyl-3,3′-dicarboxylate) (Fig. 1a) can straightforwardly be functionalised with amines to yield a family of (amine)–M₂(dobpdc) adsorbents (Fig. 1b)¹⁴. These adsorbents have large capacities for selective and reversible CO₂ uptake, and the adsorption thermodynamics can be tuned by varying the amine^{11,24,25,26,27}, and the metal^14,28,29. Importantly, these materials generally display steep adsorption isotherms making them promising for a range of energy efficient carbon capture applications^24,25. Initial characterisation of CO₂ adsorption modes in these materials has revealed a rich chemistry, with three CO₂ adsorption products proposed to date: (i) ammonium carbamate chains (Fig. 1c), thought to be the dominant product in a range of variants¹¹, (ii) carbamic acid pairs (Fig. 1d), identified in the Zn-based framework functionalised with the diamine dmpn (dmpn = 2,2-dimethyl-1,3-diaminopropane)^17,24, and (iii) a mixed adsorption product (Fig. 1e) recently proposed for (dmpn)-Mg₂(dobpdc)¹⁷. The adsorption thermodynamics of these three adsorption processes vary, motivating further characterisation to aid the design of metal-organic frameworks with the best CO₂ capture performances.

a Structure of the metal organic framework M2(dobpdc). b Amine functionalisation yields (amine)-M2(dobpdc). Upon exposure to
CO₂ various adsorption products can form. The most dominant adsorption products are: c ammonium carbamate chains, d carbamic acid pairs and e mixed ammonium carbamate–carbamic acid. 

Here we leverage the crystalline and tuneable family of (diamine)–M₂(dobpdc) adsorbents to perform a systematic computational exploration of solid-state NMR parameters for different CO₂ adsorption products. We show that ¹⁷O solid-state NMR spectroscopy is a powerful probe of CO₂ capture chemistry, providing unambiguous identification of carbamic acid formation and a detailed picture of the hydrogen-bonding environments.

Methods and Materials

All of the chemicals used in this project were purchased from commercial suppliers and were used without further purification. The ligand 4,4′-dihydroxy-[1,1′-biphenyl]-3,3′-dicarboxylic acid (H₄dobpdc) was purchased from Hangzhou Trylead Chemical Technology. ¹⁷O-enriched CO₂ gas was purchased from ICON/Berry & Associates, Inc, with ~20 at.% ¹⁷O.

Mg₂(dobpdc) synthesis

Mg₂(dobpdc) was synthesised according to a previously reported procedure¹¹. Mg(NO₃)₂·6H₂O (11.5 g, 45.0 mmol, 1.24 equiv), H₄dobpdc (9.90 g, 36.0 mmol, 1.00 equiv), N,N-dimethylformamide (DMF) (90 mL), and methanol (110 mL) were mixed together in a 350 mL glass heavy wall pressure vessel (Chemglass, CG-1880-42). The reaction mixture was sonicated for 15 min until all of the solids had dissolved, and was then sparged with N₂ for 1 h. The reaction vessel was sealed and heated at 120 °C with stirring for 21 h. This resulted in the precipitation of a white solid from solution. The solid was collected via vacuum filtration and quickly returned to the reaction vessel along with fresh DMF (250 mL). The reaction vessel was then heated to 60 °C for 3 h with stirring. Following this, the solid was again collected via vacuum filtration and returned to the reaction vessel with fresh DMF (250 mL) and again heated to 60 °C for 3 h with stirring. This washing process with DMF was repeated a total of three times, after which the solid was washed three more times in methanol (250 mL) at 60 °C to yield the desired product, Mg2(dobpdc). A small portion of the product (ca 0.1 g) was collected via filtration and activated for characterisation by powder diffraction (Supplementary Fig. 10) by heating to 60 °C in N₂ for 15 h. The remaining Mg₂(dobpdc) was stored in methanol.

Diamine-functionalised Mg₂ (dobpdc) synthesis

Diamine-functionalised Mg₂(dobpdc) materials were synthesised according to a procedure previously reported in literature¹¹. Methanol-solvated Mg₂(dobpdc) was filtered and washed with toluene (50 mL). The filtered MOF (ca 0.1–0.4 g) was then added to a toluene (4 mL) and diamine (1 mL) solution and left to soak for at least 12 h. The solid was then collected via vacuum filtration and washed with toluene (50 mL). e-2, dmpn, ee-2, i-2 and ii-2, functionalised Mg₂(dobpdc) materials were activated by heating in an aluminium bead bath under N₂ to 125 °C for 1 h, 150 °C for 1 h, 125 °C for 1 h, 130 °C for 1 h, and for 130 °C for 1.5 h, respectively. This activation step removes solvent as well as excess diamine, and ensures the samples are dry prior NMR sample preparation (see below). A portion (10–20 mg) was taken for powder X-ray diffraction analysis (Supplementary Fig. 10). To determine sample stoichiometries by ¹H NMR (Supplementary Fig. 9, Supplementary Table 5), ~5 mg of the activated amine-functionalised MOFs was digested in a mixture of dimethyl sulfoxide (DMSO-d₆) (1 mL) and two Pasteur pipette drops of deuterated hydrochloric acid (DCl) (35 wt% in D₂O, ≥ 99 at.% D).

Preparation of amine-grafted SBA15

12 g of Pluronic P123 triblock copolymer, 90 g of distilled water, and 360 g of 2 M HCl aqueous solution were mixed in a Teflon-lined container. The mixture was stirred at 35 °C for about 2 h, until complete dissolution of P123. Then, 25.5 g of tetraethyl orthosilicate was added to this solution under vigorous stirring. Stirring was stopped after 5 min, and the mixture was kept under static conditions at 35 °C for 20 h, followed by 48 h at 100 °C in an autoclave. The solid product was collected by filtration, washed with distilled water, dried at ambient condition, and calcined at 550 °C in flowing air for 6 h.

Amine grafting was carried out as described elsewhere⁵⁷. The SBA-15 support was dried at 120 °C for 4 h to remove residual moisture. Then, 1.0 g of the dried support was transferred to a multineck flask, to which 30 mL of toluene was added. The mixture was stirred at ambient temperature for 2 h, and 0.3 mL of water was added dropwise. The mixing continued for an additional 2.5 h. The temperature was then raised to 110 °C, followed by addition of 1 mL of propylamine silane or 3-[2-(2-aminoethylamino)ethylamino]propyl trimethoxysilane for propylamine-SBA15 (referred to as Pr-Si) and triamine-SBA15 (referred to as Tri-Si), respectively. The mixture was kept under reflux overnight. The grafted materials were filtered and washed with toluene followed by pentane, then dried at room temperature overnight, and archived in sealed vials. Activation before gas dosing for NMR studies was carried out by heating at 120 °C under flowing nitrogen for at least 1 h.

C¹⁷O₂ dosing of amine functionalised adsorbents

The activated amine functionalised adsorbents were packed into 4 mm or 3.2 mm NMR rotors inside a nitrogen-filled glovebag, thereby excluding water as far as possible. Each sample was then evacuated for a minimum of 10 min in a home-built gas manifold, as described previously¹⁷. ¹⁷O-enriched CO₂ gas (20 at.% ¹⁷O) was then used to dose the samples with gas at room temperature, before sealing the rotors inside the gas manifold with a mechanical plunger. (ee-2)-Mg₂(dobpdc) was dosed for 0.5 h with a final gas pressure of 896 mbar. The (dmpn)-Mg₂(dobpdc) sample for measurements at 20.0 T was dosed for 15 h with a final gas pressure of 1253 mbar, and the second independent (dmpn)-Mg₂(dobpdc) sample for measurements at 23.5 T was dosed with 17O-enriched CO₂ for 15 h with a final gas pressure of 1113 mbar. (e-2)-Mg₂(dobpdc) was dosed for 1 h, and the final gas pressure was 448 mbar. (i-2)-Mg₂(dobpdc) was dosed for 0.5 h, and the final gas pressure was 1116 mbar. The (ii-2)-Mg2(dobpdc) sample for measurements at 20.0 T was dosed for 0.75 h with a final gas pressure of 1015 mbar, and the second independent (ii-2)-Mg₂(dobpdc) sample for measurements at 23.5 T was dosed with 17O-enriched CO₂ for 0.5 h with a final gas pressure of 1039 mbar. For activated Mg₂(dobpdc) (i.e., with no amines), activation was first carried out by heating in flowing nitrogen gas at 180 °C for 15 h. This sample was then packed in an NMR rotor (as above) and dosed with gas for 0.5 h, and the final gas pressure was 1075 mbar. For silica grafted amines, samples were dosed inside 3.2 mm NMR rotors for 0.5 h with final gas pressures for Pr-Si and Tri-Si of 1083 mBar and 993 mBar, respectively.

NMR spectroscopy

¹⁷O MAS and MQMAS experiments were performed using spectrometers equipped with a 20.0 T wide-bore and 23.5 T standard bore magnets, corresponding to a 1H Larmor frequencies, ν0, of 850 MHz and 1 GHz. For experiments at 20.0 T, experiments were performed with an MAS frequency, νR, of 14 kHz. 17O MAS spectra were acquired using a spin-echo pulse sequence with radiofrequency field strength, ν1, of ~50 kHz and a recycle delay of 0.05 s. NMR parameters were optimised experimentally and therefore the spin-echo experiments cannot be considered quantitative. MQMAS experiments were acquired using a z-filter pulse sequence^58,59,60,61 with triple-quantum excitation/conversion pulses with ν1 ≈ 50 kHz and a central-transition selective π/2 pulse at ν1 ≈ 11 kHz. All MQMAS spectra are shown after shearing using the convention described in ref.⁵⁸. For experiments at 23.5 T, a spectrometer equipped with a 3.2 mm HX double resonance probe was used with a MAS rate of 20 kHz. 17O MAS spectra were acquired using a spin-echo pulse sequence, with ν1 = 25 kHz, and with a recycle delay of 0.05 s. Chemical shifts are given in ppm, and are referenced relative to liquid H₂O at 0 ppm.

DFT calculations

The candidate structures were first geometry optimised using CASTEP^{21,62,63,64,65,66,67,68,69,70,71,72,73}. This was done with (i) plane-wave basis set with an 80 Ry (1088 eV) cut-off energy, (ii) the on-the-fly generated ultrasoft pseudopotential (C17), (iii) a 1 × 1 × 3 k-point grid, (iv) the Perdew-Burke-Ernzerhol (PBE) functional with a G06 Van de Waals correction.

The NMR parameters were calculated using the same parameters and this gave values of δiso, anisotropy, asymmetry, CQ and ηQ which converged within 0.1 ppm, 0.25 ppm, 0.001, 0.0 MHz and 0.0, respectively, for the investigated oxygen and carbon nuclei at the selected k-point grid and cutoff energy.

For ¹³C and ¹⁷O NMR, the principal components of the chemical shielding tensor (σ₁₁, σ₂₂ and σ₃₃ where σ₃₃ ≥ σ₂₂ ≥ σ₁₁) were obtained directly from the CASTEP calculations, in terms of σ_xx, σ_yy and σ_zz where |σ_zz – σ_iso| ≥ | σ_xx – σ_iso| ≥ |σ_yy – σ_iso|. The principal components of the chemical shielding tensor were converted to chemical shift principal components using δ = –(σ_calc – σ_ref) where the reference values for ¹³C and ¹⁷O were 171.2 and 249.8 ppm, respectively. These values were obtained from CASTEP calculations on cocaine (¹³C)⁷⁴ and the amino acids tyrosine and valine (¹⁷O)⁷⁵, and correlation of the calculated values with the experimental values with a linear fit with a fixed gradient of –1.

Gaussian calculations

The cluster models were created using Avogadro⁷⁶ based off the models given in ref.²³. Dangling silicon bonds at the surface edges were terminated by OH species⁵⁰. All calculations were performed using the Gaussian 09 software⁷⁷. Geometry optimisations and frequency calculations were performed on the model structures prior to the calculation of NMR parameters. Note: no imaginary values were observed in the frequency calculations, and as such the structures were determined to be at the true minima. All calculations were carried out at the CAM-B3LYP/pcS-2 level of theory.

Conclusion

In conclusion, this work shows that ¹⁷O NMR is an excellent probe of different CO₂ adsorption products in amine functionalised adsorbents. In particular, ¹⁷O NMR can differentiate between ammonium carbamate chains and carbamic acids in a wide range of materials. Our measurements provide new support for ammonium carbamate chain formation in a series of (amine)–Mg₂(dobpdc) variants, and also provide strong evidence for a recently proposed mixed ammonium carbamate–carbamic acid mechanism for the material (dmpn)–Mg₂(dobpdc). We reveal carbamic acid formation in a previously poorly studied adsorbent, (ii-2)-Mg₂(dobpdc), highlighting the prevalence of carbamic acid in frameworks with bulky amine groups. Finally, initial measurements on amine-grafted silica materials showcase the excellent versatility of the technique, and support the formation of ammonium carbamates in these materials, while also suggesting a new adsorption mode may be in operation. It is worth noting that care was taken to ensure no water was present during the preparation of samples. It is known that the presence of water would impact the CO₂ adsorption mechanism and ¹⁷O NMR spectroscopy would be a powerful tool to explore additional mechanisms further. In the future, ¹⁷O NMR spectroscopy will be extended to a range of carbon capture technologies and will ultimately enable the design of improved materials that can help tackle the climate crisis.