*2.5. Crystal Structure Determination*

Single crystals of good quality were selected using optical microscopy under plane-polarized light. Intensity data were recorded on a Bruker KAPPA APEX II DUO diffractometer using graphite monochromated Mo-Kα radiation (λ = 0.71073 Å) at 100 or 173 K. Data were corrected for Lorentz-polarization effects and for absorption (SADABS) [32]. The structures were solved by direct methods in SHELXS and refined by full-matrix least-squares on F2 using SHELXL [33] within the XSEED [34] interface. The non-hydrogen atoms were located in difference electron density maps and were refined anisotropically while hydrogen atoms were placed in calculated positions and refined with isotropic temperature factors. Details of crystal structure refinements are given in Table 2 and Table S2.


**Table 2.** Crystallographic information for compounds **1**, **2**, and **3**.

### **3. Results and Discussion**

The frameworks in **1**, **2**, and **3** are identical in terms of connectivity and geometry, with the asymmetric unit consisting of a metal ion (Co<sup>2</sup>+ in **1** and **2**, Zn2<sup>+</sup> in **3**) bound to one 34pba and one 44pba linker. A centre of inversion generates a dinuclear secondary building unit (SBU) in which the two metal ions are connected by two bridging 34pba linkers through carboxylate groups while each metal ion is also coordinated to one 34pba and one 44pba through the pyridyl-N and to a 44pba through a bidentate carboxylate. The extension of this SBU through space gives rise to a double-walled network of *bcu* topology where each side of the square channels consists of a 34pba and a 44pba linker (Figure 1 and Table 2) [26]. Hour-glass shaped channels running parallel to [100] contain DMF (**1** and **3**) or acetone (**2**) gues<sup>t</sup> molecules. The presence of acetone in **2** was unexpected as a mixture of acetonitrile and water had been used to prepare this compound. Conversion of acetonitrile to acetone is likely to proceed via hydrolysis to acetic acid [35] followed by ketonization to form acetone [36,37]. There are weak hydrogen bonds between the gues<sup>t</sup> oxygens and the aromatic walls of the MOF. While **1** and **3** are isostructural, the structure of **2** is subtly different. Torsion angles indicate that the rings of both linkers are twisted slightly more away from coplanar in **2** than in **1** or **3**, while the orientation of the carboxylate groups is closer to coplanar with the aromatic ring in **2** than in the other compounds (see Figure S1 and Table S1 in ESI). The effect of these small changes is a lengthening of unit cell axes *a* and *c* while axis *b* shortens, but without changing the symmetry or space group. It is likely that the gues<sup>t</sup> influences this change through the flexibility of the bent 34pba and linear 44pba linkers which allow a hinge-like expansion or contraction of the guest-accessible void [26].

**Figure 1.** (**Top**) Coordination geometry and SBU in **1**. (**Bottom**) Packing diagrams of **1** (**left**) and **2** (**right**) showing the interactions between gues<sup>t</sup> molecules and walls of the metal-organic frameworks (MOF).

The measured PXRD patterns in Figure 2 show the similarity of **1**, **2**, and **3** frameworks which matched well to the patterns calculated from single crystal structures. However, compound **2** had a small peak at 8.9◦ instead of 9.4◦ as for **1**. There are subtle differences in the pattern for **2** compared to those for **1** and **3**, for example, the shift in peaks at positions 12◦ and 21◦. This dissimilarity could reflect the difference in the crystallographic data explained above. However, the activated forms of both **1d** and **2d** were the same after the removal of gues<sup>t</sup> solvents. All activated forms **1d**, **2d**, and **3d** (**d**: Activated) retained their crystallinities with a slight shift of peaks (except **3d**) to higher 2θ values which corresponds to a small decrease in interplanar spacing in the frameworks after gues<sup>t</sup> removal. Hence, these compounds were stable after removal of gues<sup>t</sup> molecules which is not observed in all MOFs [27,29,30].

**Figure 2.** PXRD patterns for **1**, **2**, **3**, **1d**, **2d** and **3d** and their corresponding dry forms compared to their calculated patterns.

Carbonyl stretches in the FTIR spectra (Figure 3) confirm the presence of DMF (in **1** and **3**) at 1678 cm<sup>−</sup><sup>1</sup> and acetone (in **2**) at 1713 cm<sup>−</sup>1. The removal of these gues<sup>t</sup> solvents was confirmed by the absence of these bands in the spectra of **1d** and **3d**. The spectra of the activated forms were similar to one another as expected from the PXRD analysis.

**Figure 3.** Infrared spectra of **1**, **2**, **3**, **1d**, and **3d** showing functional groups of gues<sup>t</sup> molecules and coordination modes.

Thermogravimetric analysis (TGA) and DSC are shown in Figure 4. The weight loss of 14.1% between 120 and 216 ◦C in **1** was assigned to the removal of one DMF molecule (calculated 13.8%). This was characterised by a broad endothermic peak from 115–280 ◦C in the DSC. MOF **2** shows a total

complex weight loss of 24.5%. The corresponding DSC trace shows an endothermic peak between 110 and 150 ◦C, followed by a small exotherm and a broad endothermic peak between 160 and 250 ◦C. It is possible that the removal of the acetone gues<sup>t</sup> overlaps with the decomposition of the framework. This is contradictory to the PXRD evidence that the framework is robust. It is more likely therefore that the bulk sample selected for thermal analysis may contain a mixture of crystalline forms, only some of which correspond to the MOF characterised by crystal structure elucidation. An observed weight loss of 12.7% for **3** in the range of 120 and 216 ◦C was attributed to the removal of one DMF molecule (calculated 13.7%). The corresponding DSC curve shows a broad endothermic process between 130 and 260 ◦C. The TGA traces for **1d**, **2d**, and **3d** show no mass loss before 300 ◦C, indicating the solvent has been removed from the framework.

**Figure 4.** (**a**) TGA curves for **1**, **2**, **3**, **1d**, **2d**, and **3d** (**b**) DSC curves showing the process of the removal of gues<sup>t</sup> molecules and decomposition of the framework.

### *3.1. Sorption of VOCs by Activated MOFs*

To test the potential of these MOFs to serve as sorbents for pollutants, we carried out vapour sorption experiments using a series of chlorinated volatile organic compounds (VOCs) and another series of volatile amines. Sorption of water and of ammonia were also studied. Sorption experiments were carried out using activated samples of the Co-MOF (**1d**) and Zn-MOF (**3d**).

Sorption of chlorinated VOCs dichloromethane (DCM), chloroform (CHCl3) and chlorobenzene (ClBenz) were achieved in a single crystal to single crystal manner, which allowed the elucidation of these crystal structures (Table S2 and Figure 5). The guests are stabilized in place by a number of weak interactions, including Cl··· π, and C−H··· π interactions and, in the case of chlorobenzene, through π··· π interactions with the walls of the MOF. Comparable interactions have been observed in similar systems [38,39]. PXRD patterns (Figure S2a,b) of the phases obtained by vapour sorption of all tested chlorinated VOCs into **1d** or **3d** are unchanged from the starting activated phases, thus confirming the robustness of the retained framework structure [40].

**Figure 5.** Inclusion of dichloromethane, chloroform and chlorobenzene into MOF **1d**.

The extent of selectivity in **1d** and **3d** was investigated from binary mixtures of the same chlorinated VOCs. Table 3 presents the solvent ratios obtained from the integration of relevant NMR peaks (Figure S5) from the competition studies. For **1d**, a mixture of DCM and chloroform were taken up without selectivity, while **3d** exposed to the same mixture selectivity absorbed DCM. Both MOFs selected DCM and chloroform over chlorobenzene from these respective binary mixtures. On the other hand, DCM was selectively sorbed 8.3 times over chlorobenzene. It should be noted that no attempt was made to compensate for di fferences in vapour pressure, and that the more volatile solvent was absorbed in each case, in contrast to a previous study carried out in our laboratory [19].


**Table 3.** Selectivity of **1d** and **3d** for chlorinated volatile organic compounds (VOCs).

> aDetermined by NMR (Figure S5).

**1d** and **3d** show similar sorption trends for chlorinated VOCs as well as a series of volatile amines (Figure S3). Table 4 lists the VOC sorption results for **1d** and **3d**. PXRD traces for sorbed complexes are shown in Figure S2. The loading values were calculated from TGA analysis (Figure S4) and compared to theoretical maximum loading capacities. The loading capacity (Lc) is calculated from the crystallographically derived void volume and the liquid density of the respective solvents. The maximum loading capacity (MLc) for the empty networks was estimated from

$$\text{ML}\_{\text{c}^\*} = \text{(solvent accessible void volume)} \text{(\$Z \times molecular volume)}.\tag{1}$$


**Table 4.** Uptake of selected solvents by the activated phases **1d** and **3d**.

The solvent-accessible void volume of **1d** and **3d** were estimated using Mercury with a probe radius of 1.2 Å and a grid step of 0.2 Å and were found to be 549.0 and 571.4 Å3 per unit cell respectively [41].

For the chlorinated solvents, the loading capacity (Lc) in the proposed formula {[M(34pba)(44pba)]·<sup>x</sup> solvent}n for both systems is lower than the maximum loading capacity. For each individual solvent, the sorption is higher for **1d** than for **3d**.

Water is taken up to near full capacity by both **1d** and **3d**, with little disruption of the framework. To test the potential of these compounds as sorbents for amines, the activated MOFs **1d** and **3d** were exposed to the vapours of a series of amines, *viz*. ammonia (NH3), methylamine (MeNH2), propylamine (PropNH2), 1-butylamine (ButNH2), benzylamine (BzNH2) and phenylethylamine (PhEtNH2), Table 4, Figures S2 and S3. The crystal quality of the resultant compounds was too poor to allow full structural characterisation.

For all amines except phenylamine, the loading capacity of **1d** exceeds the maximum calculated from simple molecular volumes. Complexes also become amorphous. To further understand this, we exposed **1d** to benzylamine (BzNH2) and found that the material remained crystalline until a mass loss of 40% was recorded. Subsequent desorption of the BzNH2 from amorphous **1dBzNH2** under vacuum recovered crystalline **1d** (Figure S9). In **3d** on the other hand, while the loading values obtained were again higher than the calculated maximum, the compounds retained their crystallinity but show some di fferences in phase in their PXRD traces. As with the chlorinated solvents, the amount sorbed by **1d** is greater than that for **3d**.

Amines are capable of hydrogen bonding, hence stronger intermolecular interactions, than chlorinated VOCs, which may allow them to pack more compactly into the channels, and to interact strongly with the internal surfaces of the MOFs, leading to higher loading values [39,42] and phase changes [43–45]. For benzylamine in particular, the MOFs took up a large amount, which could be attributed to aromatic stacking between BzNH2 and the aromatic rings in the MOF walls [46]. The lower sorption capacity for PhEtNH2 is the result of steric effects and lower polarity. No tests for selectivity among amine VOCs were performed.
