Controlling Interpenetration through Linker Conformation in the Modulated Synthesis of Sc Metal-Organic Frameworks

Interpenetration in metal–organic frameworks (MOFs), where multiple nets of metal ions or clusters linked by organic ligands are nested within each other's pore spaces, affects important physical properties such as stability and gas uptake, and can be controlled through ligand sterics and modifying synthetic conditions. Herein, we extend the use of coordination modulation – deliberate addition of competing monotopic ligands to syntheses – to prepare Sc MOFs containing related biphenyl-4,4′-dicarboxylate (bpdc) and 2,2′-bipyridine-5,5′-dicarboxylate (bpydc) linkers. The Sc-bpdc MOF adopts a two-fold interpenetrated structure, however, the Sc-bpydc MOF is non-interpenetrated, despite only minor electronic modifications to the ligand. A comprehensive experimental and theoretical examination reveals that ligand twisting (energetically favourable for bpdc but not bpydc) and associated π-stacking interactions are a prerequisite for interpenetration. The more rigid Sc-bpdc is susceptible to modulation, resulting in differences in morphology, thermal stability and the synthesis of a highly defective, acetate-capped mesoporous material, while the large pore volume of Sc-bpydc allows postsynthetic metallation with CuCl2 in a single-crystal to single-crystal manner. Controlling interpenetration through linker conformation could result in design of new materials with desirable properties such as bifunctional solid-state catalysts.

Scanning Electron Microscopy (SEM): Powder samples were deposited onto conductive carbon tabs mounted on aluminium stubs and coated with Pd for 150 seconds using a Polaron SC7640 sputter coater.The coated samples were transferred to and imaged using a Carl Zeiss Sigma Variable Pressure Analytical SEM with Oxford Microanalysis.(University of Glasgow) GCMC simulation details: Grand canonical Monte Carlo (GCMC) simulations were employed to obtain N 2 adsorption isotherms at 77 K.These simulations were based on a model that includes Lennard-Jones (LJ) interactions for the adsorbate-adsorbate and adsorbate-adsorbent interactions.The LJ potential parameters for the framework atoms were adopted from DREIDING S4 force field except for metal atoms, which were taken from UFF. S5 N 2 molecules were modelled using the TraPPE force field.S6 For the MOFs studied here, an atomistic representation was used starting from their crystallographic structures.The simulation box consisted of 4 (2×2×1) unit cells for all structures.A cutoff radius of 12.8 Å was applied to the Lennard-Jones (LJ) interactions, while the long-range electrostatic interactions were handled by the Ewald summation technique.Periodic boundary conditions were applied in all three dimensions.Peng-Robinson equation of state was used to convert the pressure to the corresponding fugacity used in the GCMC simulations.For each state point, GCMC simulations consisted 2×10 4 Monte Carlo cycles to guarantee the equilibration, followed by another 2×10 4 production cycles to calculate the ensemble averages.A cycle consists of n Monte Carlo moves; where n is equal to the number of molecules (which fluctuates during a GCMC simulation).Monte Carlo moves included in the simulations were insertion/deletion, translation and rotation of molecules with equal probabilities.

S2. Synthesis and Modulation of 1
The naming scheme and formulae of the MOFs reported in this study are summarised in Table S1 alongside a structural representation of their bridging organic ligands.The ligands used for construction of the MOFs (4,4′-biphenyldicarboxylic acid (bpdc) and 2,2′bipyridine-5,5′-dicarboxylic acid (bpydc)) were purchased from commercial suppliers and used as received.Elemental analysis results supplied by Alfa Aesar for the scandium(III) nitrate hydrate were consistent with a tetrahydrate.Table S1.Naming scheme and formulae of the MOFs reported during this study.The chemical structures of the ligands are given alongside their abbreviations.Initial attempts to synthesise 1 were performed under a variety of conditions where the modulator was varied.Syntheses were carried out either with no modulator or with the addition of hydrochloric acid, acetic acid (AA) or L-proline.The resulting materials are herein named 1, 1-HCl, 1-AA and 1-L-proline to reflect the modulator (or lack of) added to their respective syntheses.

Synthesis and Modulation of 1
Scandium nitrate hydrate (0.085 g, 0.28 mmol, 1 eq), bpdc-H 2 (0.068 g, 0.28 mmol, 1 eq) and N,N-dimethylformamide (DMF) (6.25 ml) were added to a 25 ml PYREX reagent bottle.If required the modulator was added (Table S2), the jar was sealed and sonicated to aid homogeneous distribution of the reagents.The resulting white suspension was placed in the oven at 120 °C for 24 hours.The bottle was removed from the oven after this period and allowed to cool to room temperature.The powder was collected by centrifugation then left to stand in fresh DMF (10 ml) overnight.The product was collected by centrifugation and the DMF was exchanged for acetone.The acetone was exchanged 3 times over 3 days.The product was collected by centrifugation and placed in a vacuum desiccator to dry.

S3. PXRD Analysis of 1
PXRD patterns were collected at room temperature to determine the crystallinity and phase purity of the bulk microcrystalline MOF samples.Comparing the PXRD patterns of the modulated samples of 1 it is clear that all the materials are crystalline and their close agreement confirms their structural similarity (Figure S1).

S7
Under the conditions investigated it was not possible to isolate single crystals of 1, however the corresponding Fe-bpdc analogue has been reported.S7, S8 1 is expected to be structurally  The close agreement between the patterns, even to high angles of 2θ, suggests that 1 has the overall formula [Sc 3 O(H 2 O) 2 (bpdc) 3 X] n (X = OH or Cl) although this is unsurprising since Sc and Fe are well-known to form structurally identical MOFs.S9-S11 To confirm that 1 is indeed two-fold interpenetrated, a comparison was made between the PXRD patterns of 1-HCl and S8 both the predicted pattern from the crystal structure of MIL-126(Fe) and a pattern predicted from the same crystal structure but with one of the two nets removed (i.e. a noninterpenetrated structure, Figure S3).The comparison indicates that 1-HCl is two-fold interpenetrated and structurally analogous to MIL-126(Fe), as the removal of one of the nets results in additional peaks in the predicted PXRD pattern which are not present in the experimental pattern of 1-HCl; it matched much more closely to MIL-126(Fe).
To confirm the purity of the material, Pawley fitting was applied to 1-HCl (Figure 3, main manuscript), which confirmed that the material adopts the two-fold interpenetrated, tetragonal MIL-126 structure.Therefore, 1 forms a rigid structure, and the two-fold interpenetration in 1 is expected to increase its structural stability as the increased rigidity strictly limits breathing and subsequently prevents pore collapse.

S4. Physical Properties of 1
N 2 adsorption and desorption isotherms were collected at 77 K to determine the porosity of the MOFs.The samples were activated by heating at 170 °C for 20 hours under vacuum.The two-fold interpenetration of 1 imparts suitable structural stability to allow the samples to be activated and varying levels of porosity were observed depending on the modulator added during synthesis (Figure S4). 1, 1-HCl and 1-L-proline all display typical Type I isotherms with 1-HCl being the most microporous of the samples.Surprisingly, 1-AA displays a typical Type IV isotherm with the presence of a mesopore clearly visible.The BET areas were calculated for the four samples of 1 and are summarised in Table S3.The addition of modulators to the synthesis of 1 results in a range of BET areas, with all three modulated materials having higher surface areas than the material synthesised in the absence of a modulator.

S10
Table S3.Comparison of the BET areas of the different samples of 1.

1-L-proline 1405
The mesoporosity of 1-AA is unexpected, however, an extra step is observed in its TGA profile and extra peaks are visible in its PXRD pattern, and together this suggests that the material is defective and likely contains acetate-capped Sc 3 O secondary building units.Poresize distributions were calculated (cylindrical pore, QSDFT, equilibrium model, N 2 on carbon at 77 K) for the different samples of 1 (Figure S5).

S11
Pore-size distribution analysis shows that the samples of 1 all contain their main pore around ~8-9 Å irrespective of the modulator used, while 1 and 1-L-proline also contain small pores around ~6.6 Å.As expected, based on the N 2 adsorption/desorption isotherms 1-AA contains a mesopore around 37 Å in diameter (Figure S5 inset).
To further investigate the gas uptake behaviour, grand canonical Monte Carlo simulations were used to predict N 2 uptake isotherms for 1 (assuming two-fold interpenetration) as well as a non-interpenetrated analogue.Figure S6a and S6b show the Type 1 adsorption isotherms on linear and semi-logarithmic representations, respectively.The simulated isotherm for 1 overpredicts the N 2 saturation loading measured for 1-HCl by approximately 25%, suggesting that the porosity may not be completely activated for 1-HCl, or the existence of amorphous non-porous phases, but crucially shows that the experimental uptake of 1-HCl does not exceed the predicted maximum capacity of a two-fold interpenetrated structure.The simulated isotherm for the non-interpenetrated analogue of 1 shows a much larger pore volume, pore size and adsorption capacity, as would be expected (Figure S6b).It is also clear that the shape of the simulated isotherms differ in the fact that the adsorption starts at lower relative pressures for 1 compared to 1-non-interpenetrated, suggesting that 1-HCl is indeed two-fold interpenetrated.The TGA profiles reveal that the overall thermal stability is not influenced by the choice of modulator with all materials undergoing thermal decomposition at ~500 °C.There are obvious differences in the profiles of the samples below the point of thermal decomposition, although this is heavily influenced by the amount of solvent that is occluded within the pores.
However, it is interesting to note that 1-AA undergoes a significant mass loss between 300-450 °C and this mass loss presumably corresponds to the removal of residual acetate.
1 H NMR spectroscopy of acid digested (DMSO-d 6 /D 2 SO 4 ) samples of 1 was used to investigate the level of modulator incorporation (Figure S8).The 1 H NMR spectra of the different samples of 1 are similar with varying amounts of residual DMF being the most prominent difference.In the spectrum of 1-AA there is an extra signal observed at ~1.8 ppm and this presumably corresponds to residual acetate remaining from acetic acid added to the synthesis. 1H NMR integral ratios confirm that the acetate loading is around 35 mol % compared to bpdc, suggesting that overall ca 15% of the bpdc ligands are replaced by capping acetates, which may explain the anomalies observed by PXRD, TGA and N 2 adsorption/desorption experiments.

S13
1-AA was collected after N 2 sorption experiments, acid digested and analysed by 1 H NMR spectroscopy to determine whether the amount of acetate changed during activation (Figure S9). 1 H NMR spectroscopy unambiguously shows that the level of acetate remains constant within 1-AA during N 2 sorption experiments and associated activation procedures (heating at 170 °C for 12 hours under vacuum).Since the acetate is not removed during activation this suggests it is either physically trapped within the MOF or that it is covalently attached.The effect of the acetic acid during synthesis is not clear; however it is evident that removal of residual acetate is not required for the creation of mesopores.As such the mesoporosity of 1-AA is likely due to the presence of acetate capped defects.

S14
To investigate the presence of defects in more detail, syntheses of 1 were carried out under the same conditions as in Section S2 but with different quantities of the acetic acid (AA) S15 modulator added.The resulting MOFs were analysed by PXRD and N 2 adsorption isotherms as shown in Figure S10.The PXRD patterns (Figure S10a) of the varying samples of 1 show that addition of AA enhances crystallinity, but in quantities greater than 45 equivalents 1 no longer forms, and presumably unreacted bpdc ligand is isolated.The N 2 adsorption isotherms (Figure S10b) show that modulation also enhances uptake, with the sample modulated by 15 equivalents of AA showing the highest uptake in the micropore region, although the mesoporosity of the sample modulated by 30 equivalents of AA leads to a higher pore volume.The mesoporosity is maintained by the sample modulated by 45 equivalents of AA but the porosity is much decreased.

S5. SEM of 1
Scanning electron microscopy (SEM) images were collected to discern the differences observed by PXRD, TGA, 1 H NMR spectroscopy and N 2 uptake experiments for 1-AA compared with the other materials (Figure S11).

S17
SEM images of the different samples of 1 reveal that different particle morphologies are obtained in the presence of different modulators.In the absence of a modulator, lozenge shaped crystals of varying sizes result (Figure S11a), while when HCl is used larger blocky aggregates of crystals with smoother surfaces are obtained (Figure S11b).The images of 1-AA show the most pronounced differences, with smooth plate like entities intertwining to form larger spherical aggregates (Figure S11c).Similarly, L-proline appears to induce different particle morphologies with aggregates of crystals observed to form larger hexagonal columns that further aggregate into clumps (Figure S11d).All four samples of 1 display similar PXRD patterns (Figure S1), confirming retention of the overall framework structure despite morphological changes.

S6. Synthesis of 2
Identical modulation conditions to those employed for the synthesis of 1 were used for the synthesis of 2, however, only the HCl modulated sample (2-HCl) was crystalline and so HCl modulation was used for the synthesis of both bulk and single crystal samples.

Bulk 2-HCl
Scandium nitrate hydrate (0.085 g, 0.28 mmol, 1 eq), bpydc-H 2 (0.068 g, 0.28 mmol, 1 eq) and DMF (6.25 ml) were added to a 25 ml PYREX reagent bottle.HCl (0.025 ml) was added, the jar was sealed and sonicated to aid homogeneous distribution of the reagents.The resulting white suspension was placed in the oven at 120 °C for 24 hours.The bottle was removed from the oven after this period and allowed to cool to room temperature.The product was collected by centrifugation and left to stand in fresh DMF (10 ml) overnight.The product was collected by centrifugation and the DMF was exchanged for acetone.The acetone was exchanged 3 times over 3 days.The product was collected by centrifugation and placed in a vacuum desiccator to dry.

Single Crystals of 2-HCl
Single crystals of 2-HCl were synthesised according to bulk synthesis conditions except upon cooling to room temperature the reaction DMF was exchanged for fresh DMF and the crystals were left to stand until they were analysed by SCXRD.

S23
The thermal stability of the non-interpenetrated flexible 2-HCl was also examined (Figure S16).The TGA profile of 2-HCl reveals that after a mass loss around 300 °C it is thermally stable to ~500 °C, similar to samples of 1.Therefore, it can be concluded that the level of interpenetration does not significantly influence the MOFs thermal stability.
N 2 uptake experiments performed on bulk samples of 2-HCl revealed that the MOF is nonporous, further suggesting that the activated bulk powder exists in a closed pore state.

S8. Crystal Structure of 3
A single crystal of 3 was serendipitously isolated from a batch of crystals containing 2-HCl.
Crystal data for 3.

Figure S1 .
Figure S1.Stacked PXRD patterns of the samples of 1.Additional peaks observed for 1-AA are marked with an asterisk.

Figure S2 .
Figure S2.Comparison of the predicted PXRD pattern of MIL-126(Fe) (1-pred, pattern predicted from Cambridge Structural Database (CSD) Refcode MIBMER) S7 and the experimental pattern of 1-HCl.The inset is an expanded view of the high angle data.

Figure S3 .
Figure S3.Comparison of the experimental PXRD pattern of 1-HCl with patterns predicted from the crystal structure of MIL-126(Fe) and the structure generated by removing one of the interpenetrating nets (CSD Refcode MIBMER).S7

Figure S4 .
Figure S4.Comparison of the N 2 adsorption (closed circles) and desorption (open circles) isotherms (77 K) of samples of 1 synthesised in the presence of different modulators.

Figure S5 .
Figure S5.Comparison of the pore-size distributions of the samples of 1.The inset is an expanded view of the mesorpore that is observed for 1-AA.

Figure S6 .
Figure S6.a) Comparison of the experimental N 2 adsorption isotherm of 1-HCl with the GCMC simulated isotherm for 1 at 77 K. b) Comparison of the experimental N 2 adsorption isotherm of 1-HCl with the GCMC simulated isotherm for 1, and the structure with one net removed (i.e.non-interpenetrated) on a logarithmic scale at 77 K.

Figure S7 .
Figure S7.Comparison of the TGA profiles of samples of 1 synthesised in the presence of different modulators.

Figure S9 .
Figure S9.Stacked 1 H NMR spectra of 1-AA before and after N 2 sorption experiments.The ligand protons have been integrated relative to the acetate signal to show that the acetate content remains constant.

Figure S10 .
Figure S10.a) Stacked PXRD patterns of 1 modulated with differing quantities of acetic acid (AA).b) N 2 adsorption (closed circles) and desorption (open circles) isotherms (77 K) of samples of 1 modulated with differing quantities of acetic acid.

S7.
data for 2-HCl.C 36 H 23 N 6 O 16 Sc 3 , M r = 930.48,crystal dimensions 0.13 x 0.04 x 0.04 mm, Hexagonal, a = b = 17.1503 (11) Å, c = 25.6245(19) Å, V = 6527.2(10) Å 3 , T = 100 K, space group P6 3 /mmc (no.194), Z = 2, 34005 measured reflections, 2192 unique (R int = 0.172), which were used in all calculations.The final R 1 = 0.118 for 972 observed data R[F 2> 2σ(F 2 )] and wR(F 2 ) = 0.378 (all data).There is only one crystallographically independent terminal oxygen atom (O2) which corresponds to a H 2 O for two of the three equivalent positions and an OH -for the third.The hydrogen atoms were not located for O2 or placed in calculated positions but are included in the unit cell contents and all values derived from them.Approximately 77% of the cell volume is not occupied by the framework and contains diffuse and disordered solvent molecules.This electron density was accounted for using SQUEEZE within PLATONS12  which calculated a solvent accessible volume of 4994 Å 3 S19 containing 772 electrons (the equivalent of ~19 molecules of DMF) per unit cell.Crystal structure data are available from the CCDC, deposition number 1559284.S20 Characterisation of 2 PXRD analysis of the modulated samples of 2 revealed that only the HCl modulated material (2-HCl) was crystalline.The PXRD patterns of 1-HCl and 2-HCl are compared to evaluate their structural similarities (FigureS12).

Figure S13 .
Figure S13.Comparison of the predicted and experimental PXRD patterns of 2-HCl.

Figure S14 .
Figure S14.a) Attempted Pawley fit of the PXRD pattern for 2-HCl.b) Zoom in of the low angle region.

Figure S15 .
Figure S15.SEM images of a) 2-HCl, and b) a zoomed in region with blocky crystals of a second phase distinct from 2, which may represent 3.
C 18 H 9 N 3 O 6 Sc, M r = 408.24,crystal dimensions 0.13 x 0.10 x 0.05 mm, Monoclinic, a = 8.8618 (4) Å, b = 48.6247(16) Å, c = 14.5573 (6) Å,  = 102.581(4) °, V = 6122.1 (4) Å 3 , T = 100 K, space group C 2 /c (no.15), Z = 8, 54506 measured reflections, 13433 unique (R int = 0.072), which were used in all calculations.The final R 1 = 0.090 for 11018 observed data R[F 2 > 2σ(F 2 )] and wR(F 2 ) = 0.270 (all data).The data for 3 were integrated as two equal (0.494(2)/0.586(2))twin components corresponding to 180° about the direct lattice direction (100).The atomic displacement parameters of all the atoms of the linkers are elongated perpendicular to the plane of the rings, including the potential pivot atoms (C2/C6, C89/C12 and C14/C18) suggesting a bowing of the linker rather than twisting.There is only one crystallographically independent terminal oxygen atom (O1) which corresponds to a H 2 O for two of the three equivalent positions and an OH -for the third.The hydrogen atoms were not located for O1 or placed in calculated positions but are included in the unit cell contents and all values derived from them.Approximately 55% of the cell volume is not occupied by the framework and contains diffuse and disordered solvent molecules.This electron density was accounted for using SQUEEZE within PLATONS12    which calculated a solvent accessible volume of 3349 Å 3 containing 1401 electrons (the equivalent of ~35 molecules of DMF) per unit cell.Crystal structure data are available from the CCDC, deposition number 1559285.The structure is shown in FigureS17.

Figure S17 .
Figure S17.Portions of the solid-state structure of 3 viewed down a) the crystallographic a axis, and b) the crystallographic c axis.H atoms removed for clarity; C atoms grey, O atoms red, N atoms blue, Sc atoms silver spheres.

Table S2 .
Summary of the modulator quantities added to the synthesis of 1.