In Situ Spectroscopy of Calcium Fluoride Anchored Metal–Organic Framework Thin Films during Gas Sorption

Abstract Surface‐mounted metal–organic frameworks (SURMOFs) show promising behavior for a manifold of applications. As MOF thin films are often unsuitable for conventional characterization techniques, understanding their advantageous properties over their bulk counterparts presents a great analytical challenge. In this work, we demonstrate that MOFs can be grown on calcium fluoride (CaF2) windows after proper functionalization. As CaF2 is optically (in the IR and UV/Vis range of the spectrum) transparent, this makes it possible to study SURMOFs using conventional spectroscopic tools typically used during catalysis or gas sorption. Hence, we have measured HKUST‐1 during the adsorption of CO and NO. We show that no copper oxide impurities are observed and also confirm that SURMOFs grown by a layer‐by‐layer (LbL) approach possess Cu+ species in paddlewheel confirmation, but 1.9 times less than in bulk HKUST‐1. The developed methodology paves the way for studying the interaction of any adsorbed gases with thin films, not limited to MOFs, low temperatures, or these specific probe molecules, pushing the boundaries of our current understanding of functional porous materials.


S1. Additional Figures
. a) Optical images and atomic force microscopy (AFM) images of a pristine calcium fluoride (CaF2) window. The material shows larger (macro) and smaller (micro) trenches, and a small surface roughness (RMS = 1.22 nm) on the planes in between. White scale bar = 100 µm, cyan scale bar = 10 µm, green scale bar 1 µm. b) X-ray diffraction (XRD) pattern of the pristine CaF2, measured in the same range as the XRD patterns in the main manuscript, showing that CaF2 possesses no dominant reflections there (only a minor around 33°). Figure S2. Electron density maps of the organic anchoring molecules, a) FBCOOH, b) TFMBCOOH and c) TFMBOH, showing relative electrostatic potential for each atom. The Voronoi deformation density (VDD) for the FBCOOH fluoride is slightly lower (-0.07) than for the TFMBCOOH (-0.083, -0.085 and -0.09) and TFMBOH (-0.093, -0.097 and -0.086) fluorides. Figure S3. Atomic force microscopy (AFM) micrographs of 100 cycles of HKUST-1 synthesized in layer-by-layer (LbL) fashion on a) pristine CaF2 and b) UV-ozone treated CaF2, demonstrating that without proper functionalization with an anchoring self-assembled monolayer (SAM), HKUST-1 is not anchored on these windows. Figure S4. Contact angle measurements on CaF2 windows, which have been pretreated by UVozone and functionalized with self-assembled monolayers (SAM) for 17 h. The angle was measured between the window and a water droplet (a) directly after formation of the SAM and (b) after the same sample was stored for 2 weeks. The contact angle is significantly increased, from 67° (σ = 1.8°) to 74° with an unusual large error (σ = 7.8°), showing the SAM to degrade over these storage durations. Corresponding X-ray diffraction (XRD) pattern of the sample, highlighting the strong (100) orientation from the utilized -COOH terminated self-assembled monolayer (SAM), as is not obvious anymore from the AFM micrograph due to the large amount of material. Figure S6. FT-IR spectra recorded during the adsorption of NO (10 -5 -100 mbar) on HKUST-1. The νasym and νsym of the linkers C-O stretch are decreasing in intensity as the NO pressure increases. Nevertheless, the position of the bands remains identical (after the initial exposure). Figure S7. FT-IR spectra recorded during the adsorption of CO (10 -5 -100 mbar) on 50 layers of HKUST-1/CaF2. The broad, split band represents the rotovibrational CO being adsorbed on a surface. Note that the intensity is similar to the deconvoluted band (pink) in Figure 6c in the main manuscript. Figure S8. Generated FT-IR spectra for FBCOOH and TFMBCOOH molecules, either free or interacting (with their respective -F or -O groups) with a single Ca 2+ ion.

S2. Discussion on Density Functional Theory Calculations
Density functional theory (DFT) calculations were performed as an estimation on the binding energy of the different organic terminated groups (-COOH, -F, -CF3) on a single Ca 2+ ion in vacuo, as we believe the organics to bind on the Ca 2+ sites in the CaF2. To keep the calculations within a reasonable time frame we focussed on just the interaction between the different fluoriated organics and one Ca 2+ ion. We are aware of the fact that surface effects, fluorine ions, solvent molecules and the large number of neighboring ions, will also influence the calculated binding energy between the Ca 2+ site and the organic molecules. These simplifications are the same for all Ca 2+organics pairs and can thus be considered a systematic error. Nevertheless, the different energies can still be compared with each other as all calculations feature the aforementioned simplifications. As discussed in the main text our calculations show that the coordination of a Ca 2+ ion towards -CF3 groups is stronger than to -F groups. This is in line with our experimental observations form the SURMOF formation. This confirms that -CF3 groups are needed to properly form the SAM and to subsequently anchor the MOF in a controlled way.

S3. Experimental Details a. Materials Synthesis
4-(Trifluoromethyl)benzoic acid (98%, Sigma-Aldrich), 4-(trifluoromethyl)benzyl alcohol (98%, Sigma-Aldrich) or 4-fluorobenzoic acid (98%, Sigma-Aldrich) were dissolved (20 mM) in ethanol (abs, VWR Chemicals) and stored in a fridge. IR-polished calcium fluoride (13 mm * 0.5 mm, supplied by Crystran Ltd, Poole) windows were first put in an UV-ozone cleaner (175 nm) 30 min per side. Then they were directly put in a glass vial vertically agains the cylindrical wall with 2 mL of SAM solution (described above) and put at 45°, to ensure both sides of the window were accessible. After 17h (for standard synthesis) the windows were washed with ethanol and dried using compressed air.
The HKUST-1 was synthesized using a layer-by-layer approach with a SILAR automated dipping robot. For one cycle, the windows were dipped for 2 min in subsequently M Cu(NO3)2 (99+%, Sigma-Aldrich) ethanolic or trimesic acid (95%, Sigma-Aldrich) ethanolic solution. In between, they were washed at 350 rpm for 60 s in pure ethanol. The samples used for AFM and XRD consisted of 100 cycles, the sample used for CO-FT-IR spectroscopy consisted of 800 cycles. Bulk HKUST-1, or copper benzene-1,3,5-tricarboxylate, was obtained from Sigma-Aldrich (Basolite ® C300).

b. Materials Characterization
Atomic force microscopy (AFM) was performed using a Bruker Multimode 8 and ScanAsyst-Air SiN3 cantilevers (F = 0.4 N/m) in non-contact mode, except for the 800 layers HKUST-1 sample which was measured with a HA_NC cantilever with monocrystal silicon tip (F= 12 N/m) in tapping mode. The data was post-treated using Gwyddion, and open-source SPM software. [1] The images were flattened by substracting a background plane and alligning the rows using a "median-ofdifferences" function. The green and blue color mask, representing the different coordination shapes in Figure 2, were drawn manually using photoshop, to highlight the specific features. The median height of the sample with 800 cycles was determined by setting a region corresponding to CaF2 in Figure S5 to z = 0, and then the median height was taken from Gwyddion's "Statistical Quantities" function.
X-ray diffraction (XRD) was measured using a Bruker Bruker-AXS D2 Phaser powder X-ray diffractometer in Bragg-Brentano geometry, using Co Kα1,2 = 1.79026 Å, operated at 30 kV. The measurements were carried out between 5 and 50° using a step size of 0.05° and a scan speed of 1 s, with a 0.1 mm slit for the source. Simulated XRD patterns were obtained by processing the corresponding .cif files with VESTA ® (λ = 1.79026 Å, FWHM = 0.2).
FT-IR spectroscopy was performed on a PerkinElmer System 2000 instrument (64 scans, 4 cm -1 resolution, MCT detector cooled with liquid N2, cell with KBr windows). The films deposited on the CaF2 were mounted on a stainless-steel cell and activated under the conditions at p < 10 -5 mbar (temperatures) described in the main text. For the bulk HKUST-1, 10 mg was pressed into a selfsupporting pallet and then mounted in similar fashion and activated as indicated in the main manuscript. Thereafter, the cell was cooled with liquid N2 temperature down to 85 K, and a mixture of 10% CO/He v/v (Linde AG, 99.999%) or 10% NO/He v/v (Linde AG, 99.9%) was dosed stepwise into the cell via a stainless-steel manifold and a valve to the pressures indicated in the main text. Deconvolutions of the spectra were performed using Origin 2017, utilizing the "Multiple peak" fit function on the 2300 -2000 cm -1 range fitting Lorentz curves. The deconvoluted peaks were then integrated over the full 2300 -2000 cm -1 range to obtain the utilized integrated areas reported in Table 1 in the main manuscript.
Contact angle measurements were performed dropping UV and milipore-filtered demineralized water on the different substrates. Images were recorded using a Dataphysics OCA 15 optical contact angle measuring setup equipped with a single direct dosing module (SD-DM) and the resulting contact angles were calcultated using SCA20 software.

c. Quantum Chemical Calculations
To model the chemical interaction between TFMBCOOH respectively FBCOOH and the CaF2 surface at a reasonable computational demand only a single Ca 2+ ion was used to simulate the surface. As discussed earlier this drastic simplification is useful to get the binding energies between Ca 2+ and the fluorinated organics.
All quantum chemical simulations were performed using DFT with the global hybrid functional B3LYP [2] and a non-augmented double-ζ-basis set of slater-type orbitals with one polarization function (DZP) [3] implemented in the Amsterdam Density Functional (ADF) Modeling Suite 2016. [4] No electron approximations were made (no frozen core), relativistic effects were included via the scalar zeroth-order regular approximated (ZORA) relativistic equation and the numerical quality was 'Good'.