Each selected crystal was first dehydrated at high temperature (450–500 K) and vacuum (3.3 × 106 mbar on the vacuum gauge). Samples were considered dehydrated when the M—Owater (Ow) oxygen residual occupancy fell below 10%. However, complete activation, i.e., a residual occupancy of Ow ~0% could not be achieved even after treatment under high vacuum and temperature; a similar behaviour was also observed in a previous study.30 There are several possible reasons for this. Some residual water may be trapped in blocked one-dimensional channels on surface contact of the crystal with the glue used to hold it in place. It was also found that if the dehydrated samples were cooled, the amount of bound water increased even under (dynamic) vacuum. This suggests that, because the whole gas delivery system cannot be heated, some water remains bound on the cool surfaces of the gas system even when the crystal itself is heated and this water is readsorbed once the crystal is cooled. To mitigate this to some degree (but not completely), the crystal was exposed to the gases of interest while it was still at the dehydration temperature. However, a small amount of residual water will remain on gas uptake. The samples were then cooled to 300 K to observe how the gas binding changes and to gain better quality data. Presented here are the obtained structures. A list of experimental details with corresponding refinement quality factors (R1) can be found in the supplementary information and the full structure determination details have been deposited with the Cambridge Crystal Structure Database as described below.
2.1 CO adsorption in Ni-CPO-27
In a first gas loading experiment, the uptake of CO in Ni-CPO-27 the crystal was first dehydrated at 450 K for 5 h in vacuo. The dehydrated structure is shown in Fig. 1, where the framework shows the expected space group number 148 and characteristic hexagonal channels seen in Ni-CPO-27.30 Ow shows an occupancy of 5.9(19)%, the framework is therefore considered dehydrated.
CO was introduced to the crystal at 450 K at an absolute pressure of 2.5 bar. Figure 1 shows that CO has chemisorbed onto the vacant metal site, via the C atom with a linear geometry and 40(4)% partial occupancy. The CO bond length is 1.18(5) Å, within error of the 1.13 Å bond length computed for molecular CO.32 The Ni—C bond length is 2.17(2) Å, longer than has been recorded spectroscopically on Ni surfaces,33 but of the same order as that found for Co-CPO-27 at lower temperature.19 The Ni-C-O angle is 177.131° approximately linear binding. As shown and discussed later in more detail (Fig. 3), the CO appears to be disordered. Enlarged atomic displacement parameters (ADPs) will be a result of the atomic displacement at the high temperature at which the dataset was recorded. Additionally, despite the structure being deemed dehydrated any disorder between the CO and residual water bound to the metal site may artificially elongate the modelled Ni-C bond and enlarge obtained ADPs for the guest. However, a sensible refinement of multiple oxygen position to resolve any such disorder was not successful for this dataset.
Reducing the temperature to 300 K with the same CO pressure has two effects visualized in Fig. 1. Firstly, the CO loading increases to 63(4)%. With this, the CO bond at 1.16(4) Å is within the error of that from the 450 K structure, the Ni—C bond is also similar at 2.12(4) Å and the Ni-C-O bond angle changes to 168.6°. Secondly, some competing water also binds to the open metal site with 27(3)% loading and a Ni—Ow bond length of 2.16(7) Å. A similar effect was seen with NO in NiCPO27.30 This is not unexcepted nor undesirable as the competition between water and NO is what allows controlled release.17 A reduction in CO bond length on cooling was expected, as Long and co-workers found spectroscopically that CO in Ni-CPO-27 was adsorbed non-classically and was blue shifted on binding.34 However, the presence of water in the model make such variations hard to uncover unambiguously. In addition, the modelled water is likely the main cause of the deviation in Ni-CO bond angle, as modelling the water and CO from the same region of electronic density causes the C to be slightly displaced. Therefore, it is probable that the CO still prefers the linear geometry seen at higher temperatures. The CO molecule was less disordered at 300 K than at higher temperature and yielded a model with only one site (Fig. 3).
By calculating a mask with a 1.2 Å probe it was possible to estimate the open pore volume and the amount of e- density remaining within the pores. The dehydrated sample had a pore volume of 2097 Å3 / unit cell. Upon loading with CO this reduced to 1782 Å3 / unit cell, and cooling reduced it slightly to 1770 Å3 / unit cell. The e- density in the pore of the dehydrated sample was estimated at 0.147 e-/ Å3. Loading with CO increased this to 0.535 e-/ Å3 and cooling slightly reduced it to 0.531 e-/ Å3. This is the expected pattern as introducing CO into the pores will increase the e- density, cooling increases the amount of CO chemisorbed so there is now less free in the pore.
2.2 CO adsorption in Co-4,6-dhip
Following the CO loading of Ni-CPO-27, CO loading of Co-4,6-dhip was performed to identify any differences between the two systems. Dehydration of a Co-4,6-dhip crystal was more complex than the Ni-CPO-27 sample, involving prolonged treatment at 450 K followed by a slow ramp over 10 h up to 500 K and holding at temperature for 1.5 h in vacuo. This was followed by flushing the sample with CO at 2.5 bar and then removing it in vacuo (Figure S3). Figure 2 shows the resulting dehydrated Co-4,6-dhip with the expected R3m symmetry.10 The Ow occupancy on the Co was 9.0(4)%, higher than that achieved for the dehydration of Ni-CPO-27. The significant difference in dehydration conditions is likely due to the differences in morphology, as one might expect Ni to be harder to dehydrate due to its higher hydration enthalpy.35 Co-4,6-dhip crystallizes in long needles, whereas Ni-CPO-27 forms broader hexagonal rods (Figures S4).30 If the hexagonal channels run down the length of the Co-4,6-dhip needle then the available surface area from which to lose water will be significantly reduced, therefore making dehydration harder.
CO was introduced to the crystal at 500 K and 2.5 bar. This caused CO to be chemisorbed to the open cobalt site, via the C atom; however, it also introduced water, as seen in Fig. 2. This caused the CO loading to be only 10.7(15)% with 15.2(15)% water loading. The CO bond length is 1.13(3) Å, the Co—C bond length is 2.26(4) Å and the Co-C-O bond angle is 151.747°. The Co—Ow bond length is 2.22(2) Å. The increase in Co—CO bond length and deviation in bond angle may be due to a weaker bond but the higher temperature and large proportion of water will affect these values significantly and further spectroscopy is needed to verify any changes. Again, the CO is highly disordered, and it was not possible to model sensibly as highlighted in Fig. 3. This binding behavior corresponds with results obtained by Long and co-workers who found similar bond lengths and geometry at lower temperatures in Co-CPO-27.36 Upon cooling the sample, a black residue formed on the outer wall of the gas cell and no further data could be gathered.
By calculating a mask with a 1.2 Å probe it was possible to estimate the open pore volume and the amount of e- density within the pores. The dehydrated sample had a pore volume of 2061 Å3 / unit cell and e- density of 0.568 e-/ Å3, higher than expected under the conditions. On loading the sample, the pore volume changed to 1830 Å3 / unit cell and the e- density to 0.644 e-/ Å3. The high e- densities may be explained by the glue at one end of the needle-like crystal, interfering with this calculation. Another possible explanation is disorder within the crystal structure, or even framework decomposition, placing apparent e- density within the pores.
2.3 NO adsorption in Co-4,6-dhip
Lastly, NO loading of the Co-4,6-dhip system was studied, allowing a direct comparison of the binding of different gases in this system and any changes of the NO binding in Co-4,6-dhip compared to Ni-CPO-27.30 This crystal dehydrated more easily than the previous, with a max temperature of only 450 K in vacuo required with no need for gas flushing. A possible explanation for this is the position of the glue relative to the crystal. The previous sample had glue at one end of the needle, whereas this sample had the glue in the middle. Since the hexagonal channels run down the length of the needle it would explain the difference in dehydration conditions needed for the same structure as more pores are blocked when the crystal is mounted end on. The residual occupancy of Ow was 7.7(4)% (Fig. 4), and the crystal structure was as expected and discussed above.
Loading NO at 450 K and 2.5 bar, caused NO to become chemisorbed to the open metal site, via the N atom, with 84.5(15)% loading (Fig. 4). As was also observed with Ni-CPO-27,30 the NO was disordered and was modelled with five O environments around the N atom (Fig. 4d). To maintain a sensible geometry the refinement included restraints on the N-O bonds as additional observations which resulted in N-O distances that ranged from 1.12(2) − 1.15(2) Å. The unrestrained Co—N bond length was 1.972(10) Å. The NO binds in a bent geometry; again this was also seen with Ni-CPO-27.30
Cooling the sample to 300 K under the same NO pressure caused an increase in NO loading to 92(3)% (Fig. 4). The O atom was again disordered and was modelled with four O environments around the N. This reduction in disorder may be down to the reduction in temperature, as seen previously.30 Unfortunately, this data set is weaker due to increased background noise, likely caused competition at lower temperature with diffraction from the gas cell capillary (Fig. 4e). The N-O interatomic distance was restrained in the same manner as above and resulted in refined distances that ranged from 1.11(2)-1.14(3) Å. The Co—N bond length was 1.960(11) Å, within error of the higher temperature sample. The NO still prefers the bent geometry.
The mask calculated with a 1.2 Å probe estimated the following parameters. The dehydrated sample had a pore volume of 2061 Å3 / unit cell which decreased to 1434 Å3 / unit cell when NO was added and decreased further to 1425 Å3 / unit cell on cooling. The e- density in the pore was again surprisingly high for the dehydrated sample at 0.578 e-/ Å3, it increased to 0.785 e-/ Å3 when NO was added and dropped to 0.648 e-/ Å3 on cooling. See above for explanation of high e- densities.