Metal–organic frameworks (MOFs) stand out amongst coordination polymers, a type of inorganic-organic polymeric material, due to their unique properties such as permanent porosity, a high degree of structural diversity, large internal surface areas, and tunability (Zhou et al., 2012; Furukawa et al., 2013). These properties have enabled MOFs for use in a variety of applications including gas adsorption and separations (Zhao et al., 2018; Ding et al., 2019; Ma et al., 2020), catalysis (Jiao and Jiang, 2019; Xiao and Jiang, 2019), energy storage (Xu et al., 2017; Zhao et al., 2018), chemical sensors (Banerjee et al., 2014; Lustig et al., 2017; Wang et al., 2018; Pamei and Puzari, 2019), cancer therapy (Hu et al., 2023; Zhong et al., 2023), and lighting devices (Lustig and Li, 2018). MOFs are constructed from the self-assembly of multitopic organic ligands and metal ions or clusters called secondary building units, SBUs (Kalmutzki et al., 2018). The ligand selection plays a significant role in controlling the resulting structure and properties of the material. MOFs built from tetraphenylethene (TPE)-based molecules have attracted increased attention in the last decade due to their unique photoluminescent properties (Mei et al., 2015; Lustig et al., 2020; 2016; La et al., 2018; Hurlock et al., 2021a). TPE exhibits aggregation-induced emission (AIE) behavior, which can be mimicked through rigidification into a coordination scaffold, making it an ideal candidate for solid-state luminescent materials (Shustova et al., 2011; Wei et al., 2014; Lustig et al., 2019). Luminescent TPE-based MOFs are most commonly utilized in sensing applications, usually of contaminants or explosives (Liu et al., 2021), as well as solid-state emitters such as LEDs (Ma et al., 2017).

The majority of TPE-based ligands used in fluorescent MOFs contain binding moieties in the para position of the end phenyl rings (Gong et al., 2014; Wei et al., 2014; Zhang et al., 2015; Yang et al., 2016), due in part to the ease of synthesis and purification of these compounds. In contrast, TPE-based ligands of lowered symmetry have seen little research interest, despite their potential for accessing more varied conformations and MOF structures without detriment to their photophysical properties. For example, the ligand 4′,4‴,4‴′,4‴‴′-(ethene-1,1,2,2-tetrayl)tetrakis([1,1′-biphenyl]-3-carboxylic acid) (m-ETTC) has a single binding moiety in the meta-position at the end of each phenyl arm. This results in a wide range of conformations the ligand can adopt due to the rotation of the terminal phenyl rings, seen in Figure 1A. For comparison, the tetratopic para-ETTC (p-ETTC) ligand (Figure 1B) is generally considered a simple rectangular or square node in MOFs. Predicting the conformation that ligands like m-ETTC will adopt a priori in solid-state structures is difficult, as it is highly dependent on the synthetic conditions and the geometry of the metal nodes.

Fortunately, predicting and influencing the geometry of the metal SBU is simpler compared to ligand conformation. For example, divalent transition metals like Zn(II), Cu(II), and Co(II) commonly form paddlewheel clusters when combined with carboxylate ligands (Mori et al., 2005; Carson et al., 2009). Paddlewheel SBUs typically coordinate to four carboxylate groups, whose conformation with respect to the linker influences the resulting dimensionality of the MOF (Furukawa et al., 2008). In literature so far, the combination of paddlewheel clusters and para-coordinated TPE-based linkers (namely, ETTC and TCPE) has only been found to form layered 2D sheet structures (Shustova et al., 2011; King et al., 2016; Zheng et al., 2020). In comparison, m-ETTC has a higher potential to form 3D frameworks with paddlewheel SBUs due to the increased diversity of its conformations.

To examine what synthetic factors may affect the conformation of the m-ETTC ligand and overall framework dimensionality, we have synthesized four distinct paddlewheel-MOF structures using m-ETTC and divalent metal cations. The first MOF, WSU-10 (WSU = Washington State University), possesses a pseudo 2D sheet structure with the formula M₂(m-ETTC) (H₂O)₂ (M = Cu or Zn). Upon addition of pyrazine as a pillar ligand to the reaction mixture, a new 3D structure denoted WSU-11 is formed, or M₂(m-ETTC) (½pyz) (H₂O) (M = Cu or Zn, pyz = pyrazine). A third structure was identified and determined to be intrinsically 3D, named WSU-12 or M₂(m-ETTC) (DMAc)₂ (M = Cu, DMAc = dimethylacetamide), and a final 3D structure named WSU-20 was isolated and found to be M₂(m-ETTC) (DMAc)₂ (M = Co or Zn). The bulk properties of these materials were assessed, and structural variations between the compounds resulting from changes to the synthetic conditions have been characterized. Additionally, broad trends in node geometry, ligand conformation, and net dimensionality are presented.

All MOFs reported herein are composed of the m-ETTC ligand and paddlewheel nodes, but vary in distinct ways that may give insight into the relationship between ligand/node geometry and overall framework structure. Note that conformations between isostructural samples of different elements generally varied by less than 1°. The orientations of m-ETTC in WSU-10 and -11 are essentially identical and will be treated together, demonstrating a symmetric conformation with all four carboxyl groups relatively in-plane with the ethene bond and pointing outwards (Figure 7A). This conformation has been previously observed in the only other two reported paddlewheel-m-ETTC MOFs, which share the WSU-10 structure (Wang et al., 2019; Xu et al., 2022). The pseudo-2D layered framework seems to be favored by this symmetric m-ETTC conformation and the square planar coordination motif of paddlewheel nodes. However, this ligand orientation is also found in WSU-20, with slightly different torsion angles due to the imprecise nature of the paddlewheel distortion (Figure 7B). WSU-20 represents the first time a 3D MOF has resulted from paddlewheel nodes and m-ETTC, and it is surprising that this occurred without a change in ligand orientation (vide infra). Additionally, a unique m-ETTC conformation is observed in WSU-12(Cu) which has not been reported in a MOF to date (Figure 7C). Two carboxyl groups on the same side of the central double bond have twisted out-of-plane and sit approximately perpendicularly to the plane of the ethene core, allowing subsequent paddlewheel coordination that extends in all axes and forms a 3D framework.

Turning to the configuration of the paddlewheel nodes, standard geometries are observed in WSU-10, -11, and −12 where the metal-metal distance is perpendicular to the coordination plane around the metals. The two square pyramids of the cluster are twisted from each other by 16.6°, 9.3°, and 0° respectively, with WSU-12(Cu) possessing an ideal paddlewheel node. In contrast, the WSU-20 structure features a more distorted cluster; here, the metal-metal line has angled 12° away from the normal of the metal coordination planes, and the individual polyhedra are twisted 8.7° from each other. It is not DMAc coordination that directly causes the distortion; the nodes of WSU-12(Cu) were also resolved with two capping DMAc molecules and still display an ideal geometry. Instead, it is the proximity of adjacent DMAc-capped nodes that sterically induces the distortion, since ideal paddlewheel geometry would have resulted in direct DMAc overlap. The emerging relationships between node capping, ligand geometry, and framework dimensionality in this series led us to speculate a mechanistic connection between the WSU-10 and -20 phases.

Consider a paddlewheel node that has formed in solution and is dynamically capped with solvent molecules (Figure 8). To form WSU-10 or -20, an m-ETTC ligand first equatorially chelates one cluster, and then must chelate a second cluster ‘below’ the first one in a way that axially aligns the two SBUs. However, the space between the axial tips of the clusters is not sufficient to allow two sets of solvent caps; the DMAc molecule of the top cluster would overlap directly with the DMAc cap of the lower cluster. Instead, there are two distinct routes for framework propagation: 1) retention of the solvent caps but distortion of the paddlewheel, allowing the two adjacent capped clusters to reside (as in WSU-20), or 2) loss of the bulky solvent caps, maintaining ideal geometry and spacing between all clusters (as in WSU-10). If route one occurs, the distorted clusters occupy positions that prevent direct bridging with a single m-ETTC molecule on the opposite side, as the 6.70 Å space is too small (Figure 8A, left). Instead, two different ligands are required to coordinate these nodes, forcing the framework to form the 3D, solvent-capped WSU-20 phase. However, if propagation occurs through route 2, then no distortion is necessary and the nodes are in ideal geometries and positions for a single m-ETTC ligand to coordinate the 9.99 Å space on the opposite side (Figure 8A, right), forming the pseudo-2D uncapped WSU-10 structure.

Thus, we hypothesize that the kinetic dependence of WSU-10/-20 formation is related to the lability of the metal-solvent interaction during crystallization. By this logic, a metal with a high affinity for solvent coordination and a low rate of solvent self-exchange (i.e., low lability) would be more likely to form the capped WSU-20 structure, while a metal with lower solvent affinity and high rates of solvent exchange (i.e., high lability) would be more likely to form the uncapped WSU-10 structure. One common measure of metal-solvent lability is the well-established rate constants of H₂O exchange in metal aqua complexes (Lincoln, 2005). Although the precise rate magnitudes would not apply in nonaqueous systems such as this, lability in these aqua species is primarily dictated by factors specific to the metal ion, such as surface charge density and d orbital occupancy. Temperature also has an effect on lability and exchange kinetics, generally increasing rates of exchange with increasing temperatures.

In the WSU-10/-20 system, we observe that there is in fact a phase dependency on the identity of the metal as well as the temperature of the reaction, supporting this hypothesis (Figure 8B). Using Cu(II), the uncapped WSU-10 phase was obtained across all temperature ranges, and the capped WSU-20 phase was never obtained. This indicates that the Cu-DMAc interaction is labile, even at room temperature, and there is never a need for the node to distort in order to propagate a framework. Conversely, using Co(II), the WSU-20 phase is obtained even up to 100°C, and transition to the uncapped WSU-10 phase was not observed. This implies that the nature of the Co-DMAc interaction is quite strong (i.e., nonlabile) and node distortion is preferred, even at high temperatures. Zn(II) was able to form both structures in a temperature-dependent manner. As expected, the uncapped WSU-10(Zn) phase was formed at higher temperatures (where metal-solvent lability is higher), while the distorted WSU-20(Zn) phase was achieved at lower temperatures, where the metal-solvent interaction is less labile. This broad trend of metal-solvent lability across these metal ions, increasing from Co(II) < Zn(II) < Cu(II), is reflected exactly in the H₂O exchange rates of their metal aqua complexes (Lincoln, 2005). It is the combined influence of the metal ion and the reaction temperature that controls the rate of metal-solvent exchange on the paddlewheel SBU, which in turn has a profound impact on the MOF phase that is obtained. The effects of reaction temperature (Yuan et al., 2016a; 2016b) and solvent identity (Zhang B. et al., 2015; Seetharaj et al., 2019) on MOF syntheses have been reported in a number of examples before. These works mainly focus on the solubility and acidity of the solvent; here, the kinetics of the solvent interaction with the metal ion appear to play the largest role.

One might wonder how the WSU-12 structure fits in to this mechanism, since it is a constitutionally isomeric phase to WSU-10 and -20 as well. This MOF was only obtained with Cu and, despite the established lability of Cu(II), was resolved with DMAc molecules coordinated to the nodes. Here, we propose that since the WSU-12 structure allows typical paddlewheel geometry while capped, there is no energetic need for the node to follow one particular route. The metal-solvent capping interaction is not in fact related to the assembly of WSU-12, disqualifying this phase from the same dependence on lability that the WSU-10 and -20 phases experience. This kinetic independence suggests that both the WSU-12(Zn) and WSU-12(Co) phases may also be accessible if ligand conformation can be controlled. Unfortunately, the synthetic factors that encourage specific ligand orientations are far less direct than the kinetic control on node geometry demonstrated here, and the WSU-12 phase appears in this study to be less energetically favorable than alternatives with planar, symmetric m-ETTC conformations.

In summary, a series of coordination polymers (WSU-10, -11, −12, and −20) were synthesized based on Zn, Cu, and Co paddlewheel metal nodes and the TPE-based ligand m-H₄ETTC, featuring a variety of ligand and node conformations that influence their dimensionalities. WSU-20 is a kinetic isomer of WSU-10, and a mechanistic relationship between the two phases has been presented that relies on the lability of the metal-solvent interaction during synthesis. This kinetic influence on phase distribution could lead to further understanding and control over the MOF construction process. Additionally, these new structures provide a better picture of the conformation that the lower-symmetry m-H₄ETTC ligand can adopt with paddlewheel nodes to propagate extended frameworks. Continued synthetic explorations are being performed to further probe the kinetic landscape of the m-ETTC system and expand this isomeric series.