Structural insights into hybrid immiscible blends of metal–organic framework and sodium ultraphosphate glasses

Recently, increased attention has been focused on amorphous metal–organic frameworks (MOFs) and, more specifically, MOF glasses, the first new glass category discovered since the 1970s. In this work, we explore the fabrication of a compositional series of hybrid blends, the first example of blending a MOF and inorganic glass. We combine ZIF-62(Zn) glass and an inorganic glass, 30Na2O–70P2O5, to combine the chemical versatility of the MOF glass with the mechanical properties of the inorganic glass. We investigate the interfacial interactions between the two components using pair distribution function analysis and solid state NMR spectroscopy, and suggest potential interactions between the two phases. Thermal analysis of the blend samples indicated that they were less thermally stable than the starting materials and had a Tg shifted relative to the pristine materials. Annular dark field scanning transmission electron microscopy tomography, X-ray energy dispersive spectroscopy (EDS), nanoindentation and 31P NMR all indicated close mixing of the two phases, suggesting the formation of immiscible blends.


Controls of a g ZIF-62
Figure S4.a. TGA profiles and b.PXRD patterns of the pristine a g ZIF-62 starting material and ball-milled and heat treated a g ZIF-62 control.Small Bragg peaks corresponding to ZnO appear in the heat treated control.

Thermal analysis of a g ZIF-62
Figure S5.Full DSC up and down scans of a g ZIF-62 starting material using a heating and cooling rate of 10 C °min -1 .

Inorganic glass characterisation 2.1 Chemical analysis and density measurements
The density of the 30Na 2 O-70P 2 O 5 glass was 2.38 0.0002 g cm -3 , in close agreement to ± literature values for glasses of similar composition.The peaks present in the IG heated (cyan curve) and ball milled, pelletised and heated control (dark blue curve) likely correspond to recrystallisation to a sodium phosphate or sodium hydrogen phosphate phase which could not be identified from the small number of weak peaks observed.These peaks are absent in the hybrid blends, which suggests a g ZIF-62, and the potential interactions between the glasses in the physical mixtures, stabilises the inorganic glass from recrystallisation.

Thermal analysis
The T g s were assigned using the DSC second upscans instead of the first because after heating and cooling, surface water has been driven off and the thermal history of the glasses has been reset.The T g obtained for the 30Na 2 O-70P 2 O 5 is lower than the value reported in the literature.This is most likely the result of the hygroscopic nature of the glass, which makes precisely determining the T g challenging as it is affected by atmospheric water. 3In line with inorganic glass nomenclature, water itself can be considered as a network modifier in the same sense as sodium oxide and depolymerises the glass network, reducing the T g of ultraphosphate glasses. 4e lack of an endothermic peak in the first DSC upscan suggests that this water uptake is not a surface effect, but rather a structural one.The glass transition is reproducible in both upscans and on cooling, suggesting the integration of water within the structure.This is consistent with the small peak at 1644.80 cm -1 in the glass FTIR (Figure S17a). 5,6  .09(1H, s, H 1 ), 7.66 (5H, m, aromatic), 2.67 (DMSO), 0.00 (TMS).Peak at 7.92 ppm is likely a solvent peak, consistent with literature spectra on a g ZIF-62 in the same solvent. 1Peak at 8.18 ppm likely a solvent peak and the TMS peak is very small but present at 0 ppm.

FTIR spectroscopy
Key bands in the 30Na 2 O-70P 2 O 5 FTIR spectrum are at 1318.48 (P=O bond in Q 3 unit), 912.63 (P-O-P asymmetric stretch), 744.19 (P-O-P symmetric stretch) and 528.72 and 586.55 cm -1 (both associated with P-O-P bending in and Q 2 and Q 3 units). 5,6A small band at 1644.80 cm - 1 is also evident, showing that P-OH bonds exist within the glass structure from the incorporation of water. 5,6milarly, FTIR spectrum of a g ZIF-62 agrees with literature spectra.A sharp peak at 669 cm -1 is associated with C-H stretching in benzimidazole, 7 1086 and 1320 cm -1 C-N stretching 8 and C-H stretch 7 at 3110 cm -1 .

Raman spectroscopy
Two distinct bands are present in the Raman spectrum of 30Na 2 O-70P 2 O 5 at 669 and 1153 cm -1 corresponding to symmetrical P-O-P stretching and symmetric PO 2 stretching in the Q 2 tetrahedra respectively. 3,9The P=O stretching frequency would be expected above 1200 cm -1 .

Figure S19
. Raman spectra of the three blend samples and starting materials.

Scanning transmission electron microscopy (STEM) and STEM-EDS
Initial image processing was carried out using the HyperSpy 10 open-source Python package.Tilt-series images were first aligned using cross-correlation routines in the Numpy and SciKit-Image Python libraries.Tilt-axis alignment was carried out by manual refinement together with filtered back-projection reconstructions implemented in SciKit-Image.Tomographic reconstructions were performed using a compressed sensing (CS) algorithm implemented in Python. 11,12Briefly, this code implements a non-negative projector using the ASTRA Toolbox 13 together with total variation (TV) based regularisation in a primal-dual hybrid gradient algorithm. 14First order TV regularisation was used in all reconstructions, accounting for material phases with homogeneous density between interfaces.Reconstructions were carried out on ARC3 nodes equipped with Nvidia P100 GPU cards, part of the High-Performance Computing facilities at the University of Leeds.
ADF-STEM tomography enables three-dimensional atomic number Z contrast. 15Given information about the multi-component formation process and EDS spectra showing P and O signals (attributed uniquely to the inorganic glass) and Zn, C, and N signals (attributed uniquely to a g ZIF-62), tomographic reconstruction volumes were inspected for intensity distributions showing two distinct average atomic number densities within the particle volume.An edge spread function (ESF) approach implemented in HyperSpy was used to identify distinct intensity phases, following the methods outlined in work by Yuan 16 and Collins. 17riefly, the volume retained after intensity-based thresholding was calculated for a series of threshold values spanning the intensity range.Sigmoid inflections in the resulting curves, modelled by the ESF, 18,19 indicate phase boundaries.Here, we used the ESF approach to identify candidate threshold ranges for particle/exterior and inorganic glass/a g ZIF-62 phase boundaries.The exact thresholds were then further refined by manual adjustment in ImageJ to ensure the particle/exterior interfaces did not exclude sample material.After the application of thresholds, additional extraneous intensity outside the particle due to residual reconstruction artefacts were removed from the volume ImageJ.
The final segmented volumes were visualised in Paraview (Kitware) as isosurface renderings.EDS maps were generated in Esprit software (Bruker) for P, O, Zn, C and N K lines.Due to the overlap of the Na K and Zn L lines no maps were generated for Na.EDS maps provided validation for the 3D segmentation results from ADF-STEM, with unique single-phase features visible in EDS maps matched to those visualised in 3D by segmentation of ADF-STEM tomography.

Thermal analysis
Very minor mass loss was observed for all three physical mixtures at the working temperature for blend synthesis (400 C).The T w selected was a compromise between promoting liquid °phase mixing at a temperature sufficiently higher than the T g of a g ZIF-62, given its high viscosity at T g , and avoiding minor decomposition of the physical mixtures during blend formation.
Results from FTIR, 1 H NMR, PXRD and Raman indicate negligible decomposition of the blends and thus this T w was selected.The difference in the peaks observed on the same pure inorganic glass sample on direct and CP spectra can be attributed to different intensity gains because of cross-polarisation phenomena in the P-OH groups in the glass.These intensity gains are caused by hydrolysis of the glass network from air exposure prior to sample measurement.Their absence in the blends' 31 P { 1 H} MAS NMR spectra suggests that hybridisation with a g ZIF-62 stabilises the glass against this hydrolysis processes.This is also evident in the PXRD analysis, where the inorganic control displays Bragg peaks corresponding to recrystallisation after ball milling, pelletisation and heating.These are absent in the blend PXRDs.Such evolution of pure inorganic glass sample (before and after heating treatment) is very dependent on the manipulation and storage conditions and therefore, it is not always reproducible.As a proof of this concept, a different batch of 30Na 2 O-70P 2 O 5 and its heat-treated control (under the experimental conditions) was measured, in which lower degree of chemical evolution was observed (Figure S64).In this case, only a limited hydrolysis has been observed after heating treatment.This suggests that differences in the inorganic glass' direct and CP spectra arise from hydrolysis of the glass, instead of the heat treatment itself.

Figure S2 .
Figure S2.Full DSC up and down scans of the crystalline batch of ZIF-62 used for the glass synthesis, using a heating and cooling rate of 10 C min -1 .A melting endotherm at 380 C was observed in the first upscan, with a T g °°at 338 C observed in the second upscan.Dotted red lines denote temperature during the scan.°1.2

Figure S6 . 5 MicroscopyFigure
Figure S6.Thermomechanical analysis of a g ZIF-62 starting material, showing an inflection at 316 C. A heating °range of 30-450 C was used.°1.5

Figure S8 .
Figure S8.Volume of 30Na 2 O-70P 2 O 5 measured over 10 cycles using He pycnometry, where the sample mass was 0.4015 g.

Figure
Figure S9.a. PXRD patterns and b.TGA profiles of the pristine 30Na 2 O-70P 2 O 5 glass and ball-milled and heat treated 30Na 2 O-70P 2 O 5 control.The peaks present in the IG heated (cyan curve) and ball milled, pelletised and heated control (dark blue curve) likely correspond to recrystallisation to a sodium phosphate or sodium hydrogen phosphate phase which could not be identified from the small number of weak peaks observed.These peaks are absent in the hybrid blends, which suggests a g ZIF-62, and the potential interactions between the glasses in the physical mixtures, stabilises the inorganic glass from recrystallisation.

Figure S10 .
Figure S10.Full DSC up and down scans of 30Na 2 O-70P 2 O 5 glass using a heating cycle of 30-450 C. A third °upscan was taken to confirm the reproducibility of the T g.

Figure
Figure S17.a. ATR-FTIR spectroscopy of all three blends and both starting materials and b.Zoomed in range of blends and a g ZIF-62.

Figure S20 .
Figure S20.Secondary electron SEM images of 1:1 physical mixture, indicating a lack of flow prior to heat treatment.

Figure S21 .
Figure S21.SEM-EDS elemental mapping of the 1:1 Zn:P blend, where scale bar is 100 m in all images.Yellow, μ pink, green, red and blue maps represent zinc, phosphorous, oxygen, carbon and nitrogen respectively.

Figure S22 .
Figure S22.SEM-EDS elemental mapping of the 1:3 Zn:P blend.Pink, blue, dark blue, red and green represent phosphorous, zinc, oxygen, carbon and nitrogen respectively.

Figure
Figure S24.a. Cross section of the 1:3 blend particle, b.Inner structure/phase after separation/thresholding and c.Outer structure/phase after separation/thresholding, showing how the two phases were separated.

Figure S26 .
Figure S26.2D EDS elemental maps of oxygen and phosphorous that comprise the inorganic phase of a particle of the intermediate 1:3 blend sample.

Figure S27 .
Figure S27.2D EDS elemental maps of carbon, nitrogen and zinc that comprise the a g ZIF-62 phase of a particle of the intermediate 1:3 blend sample.The Zn K line was used to produce the Zn elemental map.

Figure
Figure S28.a. Cross section of the 1:1 blend particle showing a difference in intensity, b.Outer structure after separation/thresholding and c.Inner structure after separation/thresholding, showing how the two phases were separated.

Figure S29 .
Figure S29.2D EDS elemental maps of oxygen and phosphorous that comprise the inorganic phase of a particle of the 1:1 blend sample.

Figure S30 .
Figure S30.2D EDS elemental maps of carbon, nitrogen and zinc that comprise the a g ZIF-62 phase of a particle of the 1:1 blend sample.The Zn K line was used to produce the Zn elemental map.

Figure
Figure S31.a. Annular dark field STEM (ADF-STEM) image of the 1:1 blend, b.ADF-STEM tomography of the inorganic glass phase of the grain, c.ADF-STEM tomography of the corresponding a g ZIF-62 phase of the particle and d.Combined phases of the studied particle, showing close mixing of the individual phases.

Figure S33 .
Figure S33.TGA trace of the 1:1 physical mixture, showing the change in weight and derivative of the weight curve with respect to temperature.
Figure S34.TGA trace of the 1:3 physical mixture, showing the change in weight and derivative of the weight curve with respect to temperature.

Figure S35 .
Figure S35.TGA trace of the 1:6 physical mixture, showing the change in weight and derivative of the weight curve with respect to temperature.

Figure
Figure S36.a. TGA profiles of all three blends and the pristine starting materials using a heating rate of 10 C min -1 °and b.TMA curve of the 1:6 sample to confirm the T g value obtained from DSC.A heating range of 30-350 C was °used for the TMA analysis, c.PXRD of the 1:3 blend and 1:3 blend after 10 days air exposure with inset showing the blend after air exposure and d.PXRD of the pristine inorganic glass after 10 days air exposure with inset showing the gel-like sample formed from the air exposure.Scale bar on both insets is 1 mm.

Figure S37 .
Figure S37.TGA trace of the 1:1 blend, showing the change in weight and derivative of the weight curve with respect to temperature.

Figure S38 .
Figure S38.TGA trace of the 1:3 blend, showing the change in weight and derivative of the weight curve with respect to temperature.

Figure S39 .
Figure S39.TGA trace of the 1:6 blend, showing the change in weight and derivative of the weight curve with respect to temperature.

Figure S40 .Figure S42 .
Figure S40.PXRD analysis of blend samples post heating for a. 1:1 blend, b. 1:3 blend, c. 1:6 blend compared to the ball milled, pelletised and heat-treated inorganic glass and d.Comparison plot of all blends post heating to 800 C. °S25

Figure S43 .Figure S44 .
Figure S43.Full DSC up and down scans of the 1:3 physical mixture using a heating cycle of 30-300 C. °Created

Figure S47 .
Figure S47.Full DSC up and down scans of the 1:3 blend using a heating and cooling rate of 10 C min -1 .A third °upscan was taken to confirm the reproducibility of the T g.

Figure S48 .
Figure S48.Full DSC up and down scans of the 1:6 blend using a heating and cooling rate of 10 C min -1 .A third °upscan was taken to confirm the reproducibility of the T g .

Figure S50 .
Figure S50.X-ray total scattering structure factor, S(Q), data of the three blend samples.

Figure S51 .
Figure S51.X-ray pair distribution function D(r) of pristine a. a g ZIF-62 and b. 30Na 2 O-70P 2 O 5 with assigned correlations.

Figure
Figure S53.a. D(r) and b.Total scattering structure factor, S(Q), data of the three physical mixtures.

Figure S54 .
Figure S54.Experimental D(r) plotted against the multiple linear regression fits of a. 1:1 blend, b. 1:3 blend, c. 1:6 blend and d. adjusted 1:3 blend.An asterisk denotes the potential interface peak in all four plots.

Figure
Figure S55.a. D(r) comparison of the 1:3 adjusted and unadjusted sample, showing minor differences between the two D(r) and b.Structure factor, S(Q), comparison of the 1:3 adjusted and unadjusted sample.

Figure S56 .
Figure S56.D(r) of the empty borosilicate capillary used for data correction, showing a large peak at r = 1.6 Å.

Figure S58 .
Figure S58.Goodness of fit R 2 values obtained from the multiple linear regression fits for the physical mixtures and blends.The % of a g ZIF-62 (x-axis) corresponds to the weight % of a g ZIF-62 in the blends.

Figure S59 .
Figure S59.C 1 and C 2 across the compositional series with increasing % a g ZIF-62.C 1 and C 2 are linked to the amount of a g ZIF-62 and inorganic glass in the blends respectively.The % of a g ZIF-62 (x-axis) corresponds to the weight % of a g ZIF-62 in the blends.

Figure S60 .
Figure S60.Residuals from the multiple linear regression fits of a. 1:1 physical mixture and blend, b. 1:3 physical mixture and blend, and c. 1:6 physical mixture and blend.

Figure S61 .
Figure S61.Average Zn-P distance via a bridging oxygen atom from three zinc phosphate CIFs, CCDC numbers 1007095, 2310787 and 2310789.

Figure S64 .
Figure S64.Comparison of second batch of inorganic glass and its heat treated control a. Direct 31 P and b. 31 P { 1 H} cross polarisation.

Figure S65 .
Figure S65.Gaussian fitting of the second batch of inorganic glass, a. Direct 31 P and b. 31 P { 1 H} cross polarisation and Gaussian fitting of the heat treated second batch of inorganic glass, c.Direct 31 P and d. 31 P { 1 H} cross polarisation.

Table S2 .
R 2 values from the two-component multiple linear regression fits.

Table S3 .
C 1 and C 2 values from the two-component multiple linear regression fits.

Table S4 .
Chemical shifts obtained from the Gaussian fitting of Direct 31 P and CP 31 P { 1 H} NMR data.

Table S5 .
Peak intensity changes in 31 P NMR vs 31 P { 1 H} NMR data.

Table S6 .
Chemical shifts obtained from the Gaussian fitting of Direct 31 P and CP 31 P { 1 H} NMR data of the second batch of inorganic glass and its heat-treated control.