Metal–organic frameworks (MOFs): structure, synthesis and uses

A metal–organic framework is built from two kinds of piece. The nodes are metal ions or small metal–oxygen clusters, often called secondary building units (SBUs); the linkers are rigid organic molecules, usually carboxylates or azolates, that bind a node at each end. Because each node accepts linkers in fixed directions and each linker has a fixed length, the pieces assemble into a regular, open network with pores of a defined size and shape.

Before 1999, most open coordination frameworks collapsed when their guest solvent was removed (Li et al. 1999). MOF-5, Zn₄O(BDC)₃, changed that: its octahedral Zn₄O(CO₂)₆ clusters, joined by terephthalate (BDC) into a cubic net, stay crystalline when the pores are emptied and heated to 300 °C (Li et al. 1999). Handled without air, MOF-5 reaches a BET surface area of about 3,800 m² g⁻¹ (Kaye et al. 2007).

The same year, Williams and co-workers reported the copper framework HKUST-1 (Chui et al. 1999), and the field grew quickly. The metal node and the linker can be chosen separately, so MOFs span a far wider range of compositions than zeolites, and unlike covalent organic frameworks they have metals in the backbone, which can act as catalytic or binding sites.

A MOF can be reduced to a net: replace each node by a point and each linker by a line. The net depends on how many linkers each node takes and in which directions, and naming it gives a blueprint for making related frameworks. The six MOFs on this site show the main types of node:

• Zn₄O(CO₂)₆ clusters, six-connected octahedra, on the primitive cubic pcu net in MOF-5.
• Cu₂(CO₂)₄ paddlewheels, square four-connected units, joined by triangular trimesate on the tbo net in HKUST-1 (Kong et al. 2013).
• Zr₆O₄(OH)₄ clusters with twelve carboxylates, on the face-centred cubic fcu net in UiO-66 (Valenzano et al. 2011; Wu et al. 2013).
• Single tetrahedral Zn²⁺ ions joined by 2-methylimidazolate on the sodalite sod net in ZIF-8; the zinc takes the place of the tetrahedral Si of a zeolite and the imidazolate bridges two zinc ions as oxygen bridges two silicons, so zeolitic imidazolate frameworks take zeolite nets (Park et al. 2006).
• Cr₃O trimers joined by terephthalate into supertetrahedra, which share corners on the MTN zeotype net in MIL-101(Cr) (Férey et al. 2005).
• Infinite metal–oxygen rods joined by dobdc⁴⁻ into a honeycomb of one-dimensional channels in MOF-74 (Rosi et al. 2005).

Keeping the net and changing the linker is called isoreticular design. In the IRMOF series, MOF-5 is IRMOF-1, and longer or functionalised linkers on the same pcu net gave frameworks with designed pore sizes and functional groups, tested for methane storage (Eddaoudi et al. 2002).

MOF pores range from windows narrower than many small molecules to mesoporous cages. ZIF-8 has cages 11.6 Å across joined by six-ring windows only 3.4 Å wide (Park et al. 2006), although the linkers swing so that the effective sieving aperture is 4.0–4.2 Å (Zhang et al. 2012). MOF-5 has alternating cavities about 12 and 15 Å across (Rowsell et al. 2005), HKUST-1 three kinds of pore about 12.6, 10.6 and 5.0 Å across (Wong-Ng et al. 2015), and MIL-101(Cr), with a giant cubic cell of about 702,000 ų, mesoporous cages of about 30 and 34 Å (Férey et al. 2005).

Surface area is measured from nitrogen adsorption at 77 K, usually with the Brunauer–Emmett–Teller (BET) equation fitted to the low-pressure part of the isotherm (see BET surface area):

where $v$ is the amount adsorbed at relative pressure $p/p_0$, $v_m$ the monolayer capacity and $c$ the BET constant. The value depends on how the sample is made and handled: MOF-5 gave a Langmuir surface area of 2,900 m² g⁻¹ in the first report (Li et al. 1999) and a BET value of about 3,800 m² g⁻¹ when made and kept air-free (Kaye et al. 2007), and UiO-66 ranges from 1,000 to 1,600 m² g⁻¹ with the number of missing linkers (Wu et al. 2013).

In some MOFs, solvent molecules bound to the metal can be removed without collapsing the framework, leaving open metal sites that face the pores. In HKUST-1 the water on each copper is removed by heating under vacuum, and the bare Cu²⁺ sites then bind CO and H₂ at low temperature (Prestipino et al. 2006). MOF-74 has one such site per metal, which is why Mg-MOF-74 captures CO₂ so well (Britt et al. 2009).

Most MOFs are made by solvothermal synthesis: a metal salt and the linker acid are dissolved in a high-boiling solvent such as N,N-dimethylformamide (DMF) or N,N-diethylformamide (DEF), sealed and heated, and crystals grow as the acid is slowly deprotonated. Typical conditions on the pages here:

• MOF-5: zinc nitrate and terephthalic acid in DEF at about 85–105 °C (Eddaoudi et al. 2002).
• HKUST-1: copper(II) nitrate and trimesic acid in water/ethanol at 180 °C for 12 h (Chui et al. 1999).
• UiO-66: ZrCl₄ and terephthalic acid in DMF at 120 °C for 24 h (Cavka et al. 2008).
• Mg-MOF-74: magnesium nitrate and 2,5-dihydroxyterephthalic acid in DMF/ethanol/water at 125 °C (Caskey et al. 2008).
• MIL-101(Cr): chromium(III) nitrate, terephthalic acid, HF and water at 220 °C for 8 h (Férey et al. 2005).

Other routes avoid heat or solvent. ZIF-8 forms within minutes when methanol solutions of zinc nitrate and 2-methylimidazole are mixed at room temperature (Cravillon et al. 2009), and water can replace methanol (Pan et al. 2011). MOFs have been made by ball milling (Beldon et al. 2010; Užarević et al. 2016), by continuous extrusion with little or no solvent (Crawford et al. 2015), at the half-kilogram scale under reflux (Ragon et al. 2014), and as thin films from the vapour phase (Stassen et al. 2016). In zirconium MOFs, a modulator such as acetic or benzoic acid competes with the linker for the clusters and controls crystal size and defects (Schaate et al. 2011).

Activation comes last and matters as much as the synthesis: the solvent in the pores is exchanged for a more volatile one and removed under vacuum, often with heating. Air- and water-sensitive frameworks such as MOF-5 must be activated and kept dry (Kaye et al. 2007).

• X-ray diffraction (XRD). Single-crystal diffraction solves the structure; powder XRD compared with the pattern calculated from that structure confirms phase purity and shows whether the framework survives activation, heating or soaking. HKUST-1’s thermal stability to 240 °C was shown by thermogravimetry and high-temperature single-crystal diffraction (Chui et al. 1999).
• Gas sorption (BET). Nitrogen at 77 K gives surface area and pore volume; uptake of H₂, CH₄ or CO₂ tests a target use (Eddaoudi et al. 2002; Britt et al. 2009).
• Thermal analysis (TGA). Weight loss shows solvent removal and decomposition: above 500 °C for UiO-66 (Cavka et al. 2008), up to 550 °C for ZIF-8 (Park et al. 2006).
• Spectroscopy (FTIR, NMR). Shows the linker, bound guests and defects; infrared spectroscopy followed CO and H₂ binding at the open copper sites of HKUST-1 (Prestipino et al. 2006).
• Defects. Missing linkers change the properties: a typical UiO-66 sample lacks about 10% of its linkers, which raises its pore volume and changes its gas uptake (Wu et al. 2013), and modulators tune this (Shearer et al. 2016).

Stability varies more among MOFs than any other property, and it often decides whether a MOF is useful. A useful rule is that strong, inert metal–linker bonds give stable frameworks: the Cr³⁺–carboxylate bonds of MIL-101 make it far more resistant to water, acid and base than zinc carboxylates such as MOF-5 (as reviewed by Ding et al. 2019).

• MOF-5 degrades in humid air, so it is stored and handled dry (Kaye et al. 2007).
• HKUST-1 is stable to 240 °C (Chui et al. 1999), but water binds to its copper sites.
• ZIF-8 kept its crystallinity for 7 days in boiling water, methanol and benzene and for 24 h in 8 M sodium hydroxide at 100 °C (Park et al. 2006).
• UiO-66 decomposes only above 500 °C, resists most chemicals and stays crystalline under 10 tons/cm² of pressure (Cavka et al. 2008).
• MIL-101(Cr) is stable in water at pH 0–12 for 2 months (as reviewed by Ding et al. 2019).

• Gas storage. MOFs were early candidates for methane and hydrogen storage (Eddaoudi et al. 2002; Kaye et al. 2007).
• CO₂ capture. Mg-MOF-74 has a dynamic CO₂ capacity of 8.9 wt% and releases the CO₂ at 80 °C (Britt et al. 2009); poly(ethylenimine)-loaded MIL-101(Cr) monoliths took up 1.05 mmol g⁻¹ of CO₂ from 400 ppm CO₂ at −20 °C (Wang et al. 2024).
• Separations and membranes. Thin ZIF-8 membranes separate propylene from propane (Kwon and Jeong 2013); Fe-MOF-74 separates olefins from paraffins (Bloch et al. 2012) and binds O₂ over N₂ (Bloch et al. 2011).
• Catalysis and protection. Modulated synthesis raises the catalytic activity of UiO-66 (Vermoortele et al. 2013), and UiO-66-NH₂ on fibres hydrolyses chemical warfare agent simulants (Lee et al. 2017).
• Drug delivery. MIL-100 and MIL-101 take up large amounts of drug molecules in their mesopores (Horcajada et al. 2006).
• Electronics and carbons. HKUST-1 films filled with TCNQ conduct electricity, with a tunable conductivity (Talin et al. 2014), vapour-deposited ZIFs are low-k dielectrics (Krishtab et al. 2019), and carbonised ZIF-8 gives porous carbons for supercapacitors (Chaikittisilp et al. 2012).

• MOF-5 (IRMOF-1): Zn₄O clusters and terephthalate on the pcu net, the prototypical MOF (Li et al. 1999).
• HKUST-1 (Cu-BTC): copper paddlewheels and trimesate, 1 nm channels and open copper sites (Chui et al. 1999).
• UiO-66: twelve-connected Zr₆ clusters on the fcu net, the prototypical stable zirconium MOF (Cavka et al. 2008).
• ZIF-8: zinc and 2-methylimidazolate on the sodalite net, robust in boiling water and base (Park et al. 2006).
• MIL-101(Cr): supertetrahedra on a zeotype net with giant 30–34 Å cages (Férey et al. 2005).
• MOF-74 (CPO-27): metal–oxygen rods around hexagonal channels lined with open metal sites (Rosi et al. 2005; Britt et al. 2009).

For a paper, slide or poster, show a MOF at three levels: the metal node and the linker, the net they form, and the crystal with a pore marked. The reticular design figure above shows these for MOF-5, HKUST-1 and UiO-66, with a library of nodes, linkers and nets, and the synthesis figure shows the common routes; both open as editable LabCanvas figures. The crystal structure drawing is a high-resolution SVG of MOF-5 from its crystal file, and the LabCanvas figure editor’s Library has crystal files for the other MOFs here, where you can switch between atoms and polyhedra and change the cells and view. The 3D model opens in the LabCanvas 3D Studio for a rendered image.