Metal-organic frameworks (MOFs) are porous materials, built by connecting metallic atoms by organic linkers through coordinative bonding (O'Keeffe and Yaghi, 2012). MOFs are highly tunable due to the numerous combinations of metal and organic constituents. Each set of metal and organic constituents exhibits different coordination geometries and topologies, and involves different chemical functionalities (Chughtai et al., 2015). This versatility, together with the development of reticular chemistry, have the potential to design nanomaterials for a better performance toward specific applications (Guillerm et al., 2018). Furthermore, MOFs are known for having extremely large surface areas and void volumes, and generally exhibit low densities (Furukawa et al., 2010). These properties render MOFs particularly suitable for gas storage (Ding and Yazaydin, 2013). In addition, these materials have potential in applications including gas-mixtures separation (Yoon et al., 2010; Lin et al., 2019), catalysis (Li et al., 2015), redox chemistry (Dalapati et al., 2016), or as sensors (Czaja et al., 2009).

For some specific structures, metals are not completely coordinated by linkers. As such, the coordination spheres are typically occupied by solvent molecules during synthesis (Yaghi et al., 2001). These solvent molecules can be subsequently removed by heating or by washing with volatile solvents (Bae et al., 2017). This leads to formation of so-called exposed metal sites (EMS). Recent contributions have demonstrated the usefulness of EMS as active sites for preferential gas adsorption and catalysis (Schlichtenmayer and Hirscher, 2012; Calleja et al., 2014). Amongst porous materials, one of the highest concentrations of EMS can be found in the family M-MOF-74, also known as CPO-27(M), where M is a divalent metal Zn, Mg, Mn, Co, Ni, Cu, Cd, or Fe (Liu et al., 2008; Zhou et al., 2008; Bhattacharjee et al., 2010; Dietzel et al., 2010; Calleja et al., 2014; Diaz-Garcia and Sanchez-Sanchez, 2014). These MOF-74 materials are traditionally made by linking unidimensional rod-like metal-oxide clusters by tetratopic 2,5-dihydroxiterephtalic acid (dhbdc) molecules, thus generating a hexagonal network shown in Figure 1 (Rosi et al., 2005; Xiao and Liu, 2019). Each metal in the cluster is octahedrally coordinated by five oxygen atoms from surrounding linkers, and the sixth position in each octahedron is occupied by a solvent molecule. Removal of this solvent molecule drives formation of EMS (Liu et al., 2008). The EMS in these materials have found application in gas adsorption (Zhou et al., 2008; Dietzel et al., 2010; Xiao and Liu, 2019) and catalysis (Bhattacharjee et al., 2010; Calleja et al., 2014; Xiao and Liu, 2019). The properties of MOF-74 materials have also been tuned by isoreticular modifications, including varying pore size by using larger linkers (Deng et al., 2012), mixing up to ten different metals in the structure (Wang et al., 2014), increasing EMS volumetric density by reducing linker size (Kapelewski et al., 2014), or partitioning the pore size to enhance the confinement of guest molecules (Suh et al., 2017), among others.

Moreover, the surface activity of MOFs has been enhanced by increasing the number of structural defects in their frameworks. Structural defects can improve mass transport, as well as catalytic activity. This owes largely to the aperture of void spaces in the solid frameworks as well as to the increased accessibility of the active sites. Furthermore, structural defects can improve electronic, magnetic and optical properties (Ren et al., 2017). The creation of structural defects has also been studied in MOF-74 structures by mixing the original linker with some terephthalic or benzoic acid derivatives. These derivative linkers exhibit different topologies and are proposed to reduce the coordination degree of metal atoms in the structure (El-Gamel, 2015; Wu et al., 2017). In these works, MOF-74-like structures with induced structural defects were synthesized by using Co, Ni, Mg, or Mn as metals, and their structural properties and CO₂ adsorption behavior were investigated. Therein, it was found that CO₂ adsorption was proportional to the resulting textural properties of the materials, not demonstrating any direct influence of the structural defects in the gas-solid interaction. Here, we investigate the creation of structural defects in a Co-MOF-74 structure by partially mixing the linker dhbdc with two modifications of this linker: 2,5-dihydroxybenzoic acid (dhbz) and hydroxyterephthalic acid (hbdc), as shown in Figure 1. The linker dhbz was used as a reference since this does not incorporate in the MOF-74 structure (El-Gamel, 2015; Wu et al., 2017). The structural properties of the synthesized materials were compared by powder X-ray diffraction, after different drying-treatments. The presence of EMS-coordinated solvent molecules (N,N-dimethylformamide -DMF- and methanol) was followed by infrared spectroscopy, and the temperature required to remove these solvent molecules was investigated by thermogravimetry. Our results indicate an improvement in the accessibility and activation of the EMS in defective MOF-74 structures, enhancing the potential of this material in applications where EMS are the active sites, such as gas adsorption or catalysis.

The studied materials were synthesized by using different molar proportions (from 10 to 100%) of the organic linkers dihydroxybenzoic acid (dhbz) or hydroxyterephthalic acid (hbdc) respect to dihydroxyterephthalic acid (dhbdc) in the synthesis media. After the detailed synthesis procedure in the experimental section and before washing with methanol, materials were dried in a fume hood overnight, recovered in a glass vial and weighted. The mass of synthesized solid decreased approximately linearly with increasing amounts of either dhbz or hbdc as the second linker. However, solid material was successfully recovered in all cases. The only exception was for syntheses involving more than 90% of dhbz in the initial solution, for which no solid product was obtained. This could be due to the incorporation of hbdc and the lack of dhbz in mixed-linker M-MOF-74 materials (M: Mg, Co, Ni, or Mn) that was reported in previous works (El-Gamel, 2015; Wu et al., 2017). A possible explanation for this fact is that hydroxybenzoic acid is a monocarboxylic acid, so this substance could just chelate single metal cations avoiding polymerization. The presence of 10% of the reported amount of dhbdc seems to be not enough to yield a solid product.

Figure 2 shows the powder X-ray diffraction (PXRD) patterns of the synthesized materials after substituting up to 60% of the main linker in the synthesis media. The corresponding PXRD patterns are compared to the calculated PXRD pattern of the Co-MOF-74 structure (CSD code: ORIWAP) and to the Co-MOF-74 experimentally synthesized without linker mixture. In the reference phase (i.e., pure Co-MOF-74), the dominant Bragg reflections are found at ca. 2θ = 6.8° and 11.7°. Upon inclusion of hbdc into the lattice (Figure 2A), these reflections remain dominant. However, after using increasing amounts of this second linker, the intensity of all reflections decreases as well as a signal broadening is observed together with a higher amorphous contribution. This suggests a reduction of the crystallinity of the MOF-74 phase in those materials. This qualitative assessment is used to consider the relative crystallinity of the synthesized materials with respect to the reference material. On the other hand, Figure 2C confirms the presence of the Co-MOF-74 crystal phase after using any proportion of dhbz as the second linker. Almost the same crystallinity was achieved in all materials, without any indication of additional phases being present in the powder sample. This is consistent with the lack of dhbz incorporation in the structure, so pure Co-MOF-74 phase is being synthetized in all cases, similarly to a previous work (Wu et al., 2017).

After drying the materials at ambient-temperature in vacuum overnight, the intensities of the main diffraction signals are further reduced in samples synthesized from ≥40% concentrations of hbdc (Figure 2B). This indicates that the crystallinity of those materials is notably reduced. The stability of the Co-MOF-74 crystalline structure has been demonstrated in different studies, even after full activation of the EMS in the structure at high temperature (Dietzel et al., 2008, 2010). Therefore, the reduction of crystallinity after drying in vacuum indicates a reduction in the stability of the crystal phase of the material after the partial substitution of its linker. The importance of solvent molecules for maintaining the structural stability of the modified MOF-74 structures can be a direct consequence of the presence of structural defects, that were also assumed in similar works (El-Gamel, 2015; Wu et al., 2017). Figure 1 depicts how defects are supposed to be created, as a consequence of the mixture of linkers with different topology. The presence of MOF-74 crystal structure implies that metals are in octahedral coordination (Rosi et al., 2005), so that oxygen atoms from the tetratopic linkers occupy five of the six positions in those octahedra. The sixth position is normally occupied by a solvent molecule faced to the cavity, and its removal drives formation of the EMS in a stable network. Since the original linker is mixed with tritopic linker, there must be some metal atoms in the structure where at least two solvent molecules would be necessary. Differently to MOF-74 materials, some MOF structures are not stable after removal of solvents in their coordinatively-unsaturated metal sites (Dietzel et al., 2008). Apparently, the inclusion of these defects after inclusion of higher amounts of the tritopic linker may drive to a higher instability of the framework upon solvent removal. Figure 2D shows PXRD patterns that remain unalterable after drying materials where dhbz was used as the second linker because there is not incorporation of the benzoic derivative in the structure, since no defects are supposed to be generated in this case.

Figure 3 shows SEM images of mixed-linker materials synthesized by using either dhbz or hbdc as the second linkers. A general reduction of crystal size is observed when the proportion of hbdc as the second linker increases from 0 to 30% (pictures from a to d), although the crystal morphology remains needle-like as expected for Co-MOF-74 (Botas et al., 2011). For larger concentrations of hbdc the morphology instead becomes spherical. Samples prepared with dhbz as the second linker (pictures from g to l) do not exhibit any significant change in crystal morphology or size, irrespective of dhbz concentration. Generally, when using dhbz, crystals are larger than those resulting from using hbdc. They also exhibit well-defined needle-like morphology in all cases. The latter observation could be related to the lack of incorporation of dhbz in the structure. Previous reports have suggested dhbz to act as a capping agent or modulator (Sindoro et al., 2014). In that case, the presence of dhbz is therefore expected to reduce the free concentration of Co²⁺ in solution. Since these samples are synthesized with a lower concentration of dhbdc, this general reduction of reagents presumably slows down the nucleation step, yielding larger crystals. Considering the incorporation of hbdc in the structure and the lower synthesis yield for greater proportions of this second linker in the synthesis media, the smaller crystals could indicate that hbdc is reducing the rate of crystal growing (Sindoro et al., 2014).

Figure 4 shows the normalized FTIR spectra for the synthesized materials from 0 to 30% of hbdc after washing with methanol and drying. Only materials with similar crystallinity were compared in order to avoid potential changes in spectra due to relative crystallinity. The band located at 1,410 cm⁻¹ could correspond to the stretching vibration of the C = C bond of the aromatic ring of the linker, and was previously assigned around 1,485 cm⁻¹ for material Mg-MOF-74 (Tan et al., 2014). This chemical bond is always present in the same proportion for all the linkers used in this study. All spectra were therefore normalized to this vibrational band. We noted that normalization with different bands led to the same overall trend, and thus slight intrinsic variation of intensity of the 1,410 cm⁻¹ between samples can be neglected. The signal located at 1,662 cm⁻¹ corresponds to DMF, and its absorbance signal is lower for higher incorporation of hbdc as the second linker. The band at 1,026 cm⁻¹ corresponds to methanol, and becomes more pronounced with degree of linker substitution. Considering that the boiling temperature of DMF (153°C) is higher than the temperature used for drying the materials (80°C), the evolution of these two signals indicates a facilitated solvent exchange during the washing with methanol in the defective structures, because of a greater accessibility of DMF in the pores. The presence of methanol in the materials despite heating samples above the boiling temperature of methanol (60°C) indicates the strong interaction of the solvent molecules with the structure, specifically with the EMS as was previously reported (Fu et al., 2012). This interaction makes it necessary to outgas the sample at 150°C and high vacuum for 18 h to remove the methanol molecules (material 0%-hbdc_AC). Previous work demonstrated the absence of this signal in fully-activated MOF-74 materials (Villajos et al., 2015).

Figure 5 compares the thermogravimetric (TG) analysis of Co-MOF-74 and the mixed-linker 30%-hbdc material, after washing with methanol. The TG curves show two main weight losses that correspond to solvent removal between 50 and 150°C, and linker combustion above 280°C. The latter leads to formation of the metal oxide. These temperatures are in accordance with those previously reported for methanol-washed MOF-74 materials (El-Gamel, 2015). The temperature for maximum mass loss rate (minimum of dTG-curves) in the solvent loss is higher than the boiling point of methanol due to strong interactions of solvent with the EMS, as commented above, but also some kinetic effects like diffusion could be involved. By comparing curves from both materials, this temperature is lower for the substituted-linker structure. This indicates that defective MOF-74 materials are more readily activated than the non-defective structures. This is consistent with the FTIR results presented above, where DMF molecules are partially exchanged by methanol molecules. This substitution makes the solvent-removal temperature to decrease. The presence of structural defects and the reduction in the crystal size are other factors that could enhance the mass transport rate.

Figure 6 shows the N₂ adsorption-desorption isotherms at the temperature of −196°C of the materials 0%- and 30%-hbdc_SC after their activation. Both materials show a type I isotherm corresponding to microporous materials. The isotherm corresponding to material 30%-hbdc_SC shows a higher gas uptake at P/P₀ around 0.2, indicating a higher microporous volume despite a higher crystallinity reduction after analysis (Figure 6-inset). A hysteresis loop in the desorption branch and an important high-pressure adsorption (P/P₀ > 0.95) are also observed in both materials, reaching a similar value at P/P₀ = 0.998. Furthermore, the adsorption and desorption branches do not meet even at lower P/P₀ values than 0.4. These features in the isotherm could indicate lack of equilibrium in the analysis. Similar results were found in mechanochemically synthesized Zn-MOF-74 materials, where a similar effect is a consequence of the presence of non-reacted linker molecules blocking the pores (Julien et al., 2016). Herein, materials were solvothermally synthesized and washed several times, so that the presence of unreacted molecules in the pores can be dismissed. The lack of equilibrium might be a consequence of a short time during equilibration and a high initial gas-dose in the analysis. Therefore, in addition to an enhanced accessibility to porous structure due to an improved activation, results are also showing that N₂ diffusion at −196°C is favored in the defective structures. Calculated BET surface areas for the synthesized materials are 215 m² g⁻¹ for 0%-hbdc_SC and 323 m² g⁻¹ for 30%-hbdc_SC, significantly lower than other values published for Co-MOF-74, usually between 900 and 1,100 m² g⁻¹ (Dietzel et al., 2010; Botas et al., 2011; El-Gamel, 2015; Villajos et al., 2015). However, the calculated surface area values in this work are not directly comparable to those higher values because of the above commented lack of equilibrium. Figure 7 shows that MeOH and, more importantly, DMF signals are reduced up to disappear in the activated 30%-hbdc_SC material. The sample 0%-hbdc_AC still retain a significant amount of DMF after activation in addition to some MeOH molecules. These results confirm a better activation process after the creation of defects, and could be the reason for the improved porous properties that were measured by N₂ adsorption.

In summary, Co-MOF-74 materials were synthesized with different proportions of hydroxyterephthalic acid mixed to the original linker (2,5-dihydroxyterephthalic acid) of the structure. This led to the formation of structural defects in the coordination sphere of the metallic sites. A demonstration of the presence of those structural defects is the reduction of crystallinity when materials are dried, which does not occur for pure-linker MOF-74 materials. Despite this, synthesized materials using up to 30% of hydroxyterephthalic acid as the second linker still exhibit the MOF-74 crystal structure. FTIR spectra indicate a better substitution of coordinated DMF to the EMS for methanol after washing the defective structures, and TG analysis suggests a reduction in the temperatures that are required to remove the solvent from pores and activate defective structures. This demonstrates that structural defects involve the coordination of the EMS in the structures and likely enhance their accessibility and facilitate their activation. SEM images show a general reduction in the crystal size for more defective structures, that also facilitates the mass transfer kinetics for the activation. N₂ adsorption-desorption results are according to a better activation of defective structures, but also confirm a better diffusion of the gas molecules to the inner surface of those materials. This is beneficial for applications where EMS are the active sites, since the activation process is a key step that is required prior to usage in e.g., gas-adsorption or catalysis. The severe conditions that are normally used for their laboratory-scale activation are difficult to reproduce at a scaled-up level, so the generation of structural defects can facilitate usage of these materials at an industrial level.

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

NJ performed the synthesis of materials in the laboratory. JV planned the work, prepared the samples for their characterization, and wrote the manuscript. SR performed the TG analysis. FE helped for the discussion of the XRD patterns and contributed to the manuscript writing.