Synergistic binding sites in a metal-organic framework for the optical sensing of nitrogen dioxide

Luminescent metal-organic frameworks are an emerging class of optical sensors, able to capture and detect toxic gases. Herein, we report the incorporation of synergistic binding sites in MOF-808 through post-synthetic modification with copper for optical sensing of NO2 at remarkably low concentrations. Computational modelling and advanced synchrotron characterization tools are applied to elucidate the atomic structure of the copper sites. The excellent performance of Cu-MOF-808 is explained by the synergistic effect between the hydroxo/aquo-terminated Zr6O8 clusters and the copper-hydroxo single sites, where NO2 is adsorbed through combined dispersive- and metal-bonding interactions.


Supplementary Methods
All reagents were used as received from commercial suppliers unless otherwise stated.
Powder X-ray diffraction (PXRD) patterns were measured with a Bruker D8 diffractometer with a copper source operated at 1600 W, with step size = 0.02° and exposure time = 0.5 s/step. Samples were placed on a borosilicate sample holder and then the sample surface was levelled with a clean microscope slide. All the samples were ground prior to analysis unless otherwise stated. Data were measured using a continuous 2θ scan from 3.0-45º θ. For all samples, PXRD patterns are presented from 0-30º θ for visual clarity.
Scanning electron microscopy (SEM) images were collected with a JEOL JSM 7600F microscope, with field emission gun and electron detector in lens. And for the energy dispersive X-ray spectra (EDS) a S-3000N microscope equipped with an ESED and an INCAx sight of Oxford Instruments was used. All samples were prepared for SEM and EDS by dispersing the material onto a double sided adhesive conductive carbon tape that was attached to a flat aluminum sample holder and they were sputtered with carbon or gold (12 nm).
Textural analyses. Nitrogen adsorption and desorption isotherms were measured at 77 K using a Micromeritics ASAP 2020 system. The samples were outgassed at 100 ºC for 16 h before the measurements. The specific surface areas (BET) were calculated by application of the Brunauer-Emmett-Teller equation taking the area of the nitrogen molecule as 0.162 nm 2 . The linear range of the BET equation was located between 0.05-0.35 P/P0, however, for all other materials studied due to their microporous natures this linear range was much narrower and displaced to lower relative pressures: P/P0= 0.04-0.07. The micropore volume and external surface area, i.e. the area not associated with the micropores, were calculated using a t-plot analysis. Taking the thickness of an adsorbed layer of nitrogen as 0.354 nm and assuming that the arrangement of nitrogen molecules in the film was hexagonal close packed. The mesopore volumes of the materials were calculated from the volume of gas adsorbed at a relative pressure of 0.6 on the desorption branch of the isotherms, equivalent to the filling of all pores below 50 nm, minus the microporosity calculated from the corresponding t-plot. The total pore volume was calculated from the volume of gas adsorbed at a relative pressure of 0.95 on the adsorption branch of the isotherms. The pore-sizedistribution (PSD) curves were obtained from the adsorption branches using non-local density functional theory (NLDFT) method for a cylinder pore in pillared clays, using a regularization of 0.100. MicroActive software was used to perform these analyses. Thermogravimetric analyses and differential thermal analyses (TGA-DTA) were performed using a SDT Q600 from TA Instruments equipment in a temperature range between 20 °C and 800 °C in air (100 mL/min flow) atmosphere and heating rate of 10 ºC/min. Infrared spectra (FTIR) were recorded on a PerkinElmer 100 spectrophotometer using a PIKE Technologies MIRacle Single Reflection Horizontal ATR Accessory from 4000-450 cm -

.
Elemental analyses were performed with a LECO CHNS-932 analyser, with dry samples.

Supplementary Note 1. Synthesis of the materials
MOF-808. Trimesic acid (210 mg, 1.0 mmol) and ZrOCl2·8H2O (970 mg, 3.0 mmol) were added to a mixture of 90 ml of formic acid (45 mL) and DMF (45 mL) in a screw cup glass bottle. The reaction was heated at 130 °C for 48 h in the oven. After cooling to room temperature, white powder was collected by centrifugation (12000 rpm, 2 min), and the solid was washed with DMF, distilled water and acetone (50 mL x 3 each) The solid was dried in the oven at 60ºC overnight. Then, the product (800 mg) was put in a solution 1 M of HCl (100 ml) stirring at room temperature for 24h, with the aim of exchanging some of the formates ligands for OH2. The resulting mixture was centrifuged and the solid was washed with distilled water and acetone (50 mL x 3). The solid was dried in the oven at 60 ºC overnight yielding MOF-808 as a white powder (750 mg). Cu-MOF-808. Different salts of Cu (II) and Cu (I) were tested as precursors to introduce the copper in the MOF structure. They were characterized by Powder X-ray diffraction (Fig S4.1) and ICP (Table S2.1) obtaining as the best option the copper (II) acetate. The explanation relies on the basicity of the conjugate pair of the salts, as acetate is the most basic pair it is the best one to equilibrate the positive charges of the structure formed in the removal of formates at the activation. Water molecules are removed from the pores drying in the oven. However, over time these molecules can be re-adsorbed in the pores. Results of Pawley refinements are shown in Table S4.1 and Fig S4.3.

Supplementary Note 5. Nitrogen adsorption-desorption analyses
The BET analysis of the materials show a diminution of the specific surface area of metalated materials. Regarding the NLDFT pore size distribution calculation, the Cu-loaded MOF seems to maintain the micropore window (ca. 12 Å and ca.18 Å) but the micropore contribution (taken as the pore width under 24 Å) to the total pore volume rapidly decreases with the metal loading.
Supplementary Table 3: Data collected from N2 isotherms at 77 K, BET and t-plot analysis.

Supplementary Note 6. Thermal gravimetric analysis
TGA data collected on the materials showed first a mass loss associated with loss of water molecules with values of 32.5% and 31.1% for MOF-808 and Cu-MOF-808, respectively. This first curve is more pronounced in MOF-808, than Cu-MOF-808 due to the amount of water which this material possesses ( Fig

Supplementary Note 7. ATR-FTIR analysis
The FTIR analysis shows the signals of principal functional groups of the structure. The Cu-MOF-808 FTIR spectra before and after NO2 sensing process are compared. No significant difference is remarked between them. The vibrational band assignment was made based on previous work. 2

Supplementary Note 8. Pair Distribution Function
Synchrotron X-ray total scattering data suitable for PDF analyses were collected at the P02.1 beamline at PETRA III (Deutsches Elektronen-Synchrotron) using 60 keV (0.207 Å) X-rays. Samples were loaded in polyamide (kapton) capillaries (0.8 mm Ø) and sealed using epoxy. Data were collected using an amorphous silicon-based Varex XRD 4343CT (150×150 µm 2 pixel size, 2880 x 2880 pixel area, CsI scintillator directly deposited on amorphous Si photodiodes) area detector. Geometric corrections and reduction to 1D data used DAWN Science software. 3 PDFs were obtained from the data within PDFgetX3 within xPDFsuite to a Qmax = 22 Å −1 . 4 Differential PDFs were obtained by subtraction of a reference PDF (pristine MOF-808) from Cu-MOF-808 in real space after applying a normalization factor to the data.
Supplementary Figure 15. PDF data for Cu-MOF-808 after sensing measurement. In red is the dPDF data for Cu-MOF-808.
Supplementary Figure 16. PDF data for Cu-MOF-808 materials before and after sensing measurement. In red the differential PDF data.
The PDF characterization after sensing does not show any significant new distance in Cu-MOF-808. This indicates that the material is stable to the sensing process and also corroborates the theory that the interaction with NO2 is reversible.

Supplementary Note 9. X-Ray Absorption Spectroscopy
Transmission and fluorescence geometry XAS measurements were performed at the P65 beamline of PETRA III. Cu K-edge XAS spectra were acquired from 8975 to 9010 eV, resulting in a k-range up to 12 Å -1 . The data analysis and background removal were performed within ATHENA and ARTEMIS. 5 Cu(CH3CO2)2 and CuO were employed as references. The data were collected at 10 K using a He cryostat and at 298 K. Zr K-edge XAS spectra were acquired from 17990 to 18030 eV, resulting in a k-range up to 12 Å -1 .
Supplementary Figure 17. Cu K-edge XANES spectrum of Cu-MOF-808 and first derivate analysis at 298 K.
The presence of a pre-edge signal (8986 eV) is consistent with a 1s to 4p transition + LMCT "shakedown" 3p of L to 3d of Cu 2+ . Presence of a pre-edge signal (8979 eV) is consistent with a quadrupole allowed 1s to 3d transition or a distortion from a pure D4h structure, square-planar to a D2d, twisted square-planar. 6 Supplementary Figure 18. Cu K-edge XANES spectrum of Cu-MOF-808 and first derivate analysis at 10 K. For a higher resolution in the high energy EXAFS region, the experiments were conducted at 10 K. At this temperature the pre-edge region decreased in intensity dramatically, being therefore indicative of the proposed metal(3d)-Ligand(4p) orbital mixing. This mixing is directly associated with the vibronic structure of the system, which rapidly decreases with temperature. 7 Supplementary Figure 19.

Supplementary Note 10. X-ray photoelectron spectroscopy
X-ray photoelectron spectra were recorded with a lab-based spectrometer (SPECS GmbH, Berlin) using monochromated Al source (Al Kα1 h = 1486.6 eV) operated at 50W as excitation source. In the spectrometer, the X-ray is focused with a µ-FOCUS 600 monochromator onto a 300 μm spot on the sample, and the data is recorded with a PHOIBOS 150 NAP 1D-DLD analyser in fixed analyser transmission (FAT) mode. The pass energy was set to 40 eV for survey scans and 20 eV for high-resolution regions. The binding energy scale was calibrated using Au 4f7/2 (84.01 eV) and Ag 3d5/2 (368.20 eV). Charge compensation was required for data collection. Recorded spectra were additionally calibrated against the C 1s internal reference. Data interpretation was done with Casa XPS. Shirley or two-point linear background were used depending on the spectrum shape. Surface chemical analysis was done based on the peak area of high-resolution spectra and the CasaXPS sensitivity factors (where RSF of C 1s = 1.000). Photoelectron spectra of Cu-MOF-808 in the Cu 2p3/2 region (Fig. S11.2). The spectrum is characteristic of Cu 2+ (a d 9 cation), featuring a broad satellite next to the main peak due to localized final states (a), in which the valence hole remains mainly on the core-hole site, and a main peak (b), with most of the valence-hole density concentrated on the ligand sites surrounding the core-hole site. Photoelectron spectra of Cu-MOF-808 in the C 1s region (Fig. S11.3). Spectra before and after NO2 exposure consists in two peaks that can be associated with aliphatic C species (at 285 eV) and C=O species (at 289 eV) in a ratio aliphatic:C=O 3:1. The absence of changes in the C 1s region demonstrates the robustness of Cu-MOF-808 towards the adsorption of NO2.

Supplementary Note 11. Computational methodology
Density Functional Theory (DFT) calculations were performed in order to elucidate the possible configurations of Cu-MOF-808. We modelled the structural and energetic properties of several mono-and bi-nuclear copper-oxo & copper-hydroxo clusters deposited on the nodes of the MOF-808. As a starting model for the pristine MOF-808, we choose a molecular cluster that is composed by two Zr6O8 octahedra bridged by 2 ligands. This model has been previously used in our previous work, where the deposition of iron-oxo clusters on the MOF-808 was investigated. 8 The coordinates of the starting model are carved from the experimentally determined crystal structure. The benzene-tricarboxylate ligands, which are bridging the two Zr6O8 octahedra, are cropped to benzene-dicarboxylate. The remaining four ligands are also cropped to formate. Based on the experimental observations, six (6) formate molecules are further added as capping ligands. For charge balancing, four (4) protons have to be added to the μ3-O atoms of each Zr6O8 octahedron. As a next step, four formate capping ligands are removed, with each one being replaced by a hydroxo and a water molecule, giving rise to the MOF-808 model, where the copper-hydroxide species will be deposited.
Subsequently, we investigated the structural and energetic characteristics of the deposition of two Cu(II) atoms on the nodes of the MOF-808 model. Two possible ways of deposition have been considered: i) two mono-nuclear, and ii) one bi-nuclear copper-oxo & copperhydroxo clusters. To reduce the complexity of the system, due to many possible combinations to couple the unpaired electrons of the Cu(II)/Cu(II) atoms, we decided to study only the deposition of the high-spin ferromagnetically coupled Cu(II)-Cu(II) pairs with a spin multiplicity of 3. Because of the different stoichiometries of the resulting structures, the comparison is done by computing the formation energies of the Cu2Ox(OH)y(H2O)z with the equation: ΔE form = E(Cu-MOF-808) -E(MOF-808) +mE(H2O) -E(MOF-808) -nE(precursor), where E are the energies of the Cu-MOF-808, MOF-808, H2O, and the copper precursor molecules, and m, n are the number of water and precursor molecules in the formation reaction respectively. As Cu(II) precursor, a molecule with the stoichiometry Cu(OH)2(H2O)2 is considered.
As a final step, after obtaining the most stable configurations for the deposited Cu(II) atoms, the interactions with NO2 have been computed. Several configurations have been considered and the interaction energies with NO2 are computed with the equation: I.E = E(NO2-Cu-MOF-808) -E(Cu-MOF-808) -E(NO2), where E are the energies of the NO2 complex with Cu-MOF-808, Cu-MOF-808 and NO2, respectively. The total spin multiplicity of the complexes is considered to be a quartet.
During all geometry optimizations, some restrictions have to be applied in order to mimic the crystal environment. Two (2) of the zirconium atoms at the edges of the molecular cluster, and twenty-four (24) oxygen atoms that belong to the ligands are kept frozen. The r2-SCAN-3c functional in combination with the def2-mTZVPP basis set have been used for all geometry optimizations. This low-cost density functional has been shown to perform very well for open-shell transition metal reactions. 9, 10 Finally, single point energies with the M06-L functional in combination with def2-TZVPP have been performed at the r2-SCAN-3c optimized geometries. 11 All calculations have been performed using the ORCA 5.0.3 program. 12

Supplementary Note 12. Computational results
Due to the rich proton topology of MOF-808, several possibilities exist, how a Cu2Ox(OH)y(H2O)z cluster can be deposited on the nodes of the MOF after reaction of a copper precursor with the available protons from the μ3-ΟΗ, terminal-OH and terminal-Aqua ligands. Moreover, the copper species could be deposited either as two isolated, or as bridged-(hydr)oxo. Initially, we investigate, how the first copper can be attached. This can be done either through the reaction of the precursor with the terminal oxygen atoms or with the μ3. The calculations show that adsorption is more favorable, when the copper atom interacts through the μ3and terminal-oxygen atoms than with the two terminal-oxygen atoms only. The two adsorption schemes are denoted as (μ3-OZr/t-OZr)-Cu(OH)(H2O) and (t-OZr/t-OZr)-Cu(OH)(H2O) respectively and are shown in Supplementary Figure 24. The relative energy difference between these two adsorption modes is ~48 and ~29 kJ/mol according to the r2-SCAN-3C/def2-mTZVPP and M06L/def2-TZVPP methods. In the case of adsorption on the terminal-oxygen atoms, the copper atom has one terminal hydroxo and one aqua ligand, which interact via hydrogen bonding with the terminal-hydroxo and aqua ligands of another zirconia node. Subsequently, based on the observation for the deposition of one copper atom, several possible configurations have been considered for the deposition of two copper atoms. Similarly, the deposition as two   We have also calculated the interaction energies between NO2 and the three most stable models of the Cu-MOF-808. For comparison reasons, we have also considered the adsorption of NO2 on the MOF-808 and the Cu-HKUST-1, which is a MOF containing unsaturated copper metal sites in a similar distorted square planar environment. The results are summarized in Supplementary Table 5 and Supplementary Figure 25 .
Supplementary Table 5: Interaction energies (in kJ/mol) of NO2 with the three most stable models the Cu-MOF-808, the MOF-808 and the Cu-HKUST-1. Interaction energies are computed from the r2-SCAN-3c/def2-mTZVPP method. The stoichiometries of the reaction energies, important distances (Cu-Cu, Cu-Zr in Å) and the absorption manner (via the terminal-or the μ3-O) are also reported. In Cu-MOF-808, the most stable adsorption geometry of NO2 on the copper site is via its nitrogen atom. Interestingly, the most stable adsorption mode is via the oxygen atom of the NO2 with the copper atom in the HKUST-1. This can be justified by the presence of additional dispersive interactions of the NO2 with the hydroxo groups of the MOF-808. This type of interaction is not present in the HKUST-1, where the NO2 interacts only with the copper atom.

MOF Model I.E (kJ/mol) Binding mode
The two types of interaction are qualitatively illustrated by performing an Interaction Region Indicator (IRI) analysis and plotting the results in the Supplementary Figure 26. The IRI plots indicate vdW interactions between the NO2 and the μ3-OH groups as illustrated with isosurfaces of green colour in part A of Fig S.12

Competitive adsorption of other common gases in the Cu-MOF-808:
We have also explored computationally the nature of the binding between competing molecules (such as H2O, NO, CO and CO2) with the Cu-MOF-808. The relative strength of the interactions between the three competing molecules (NO2, NO, CO2, CO and H2O) and the copper site is used as a qualitative descriptor to predict, which of the above molecules will be preferentially adsorbed. However, recent experimental and computational studies 13,14 have shown that the binding strength does not always indicate the preferred adsorbed molecule, and that a combination/competition between thermodynamics (binding energies) and kinetics (energy barriers) defines the adsorbed molecule. For example, although NH3 has stronger binding than H2O with the open metal sites of the Ni-and Mg-MOF-74, the presence of multiple H2O molecules will cause the displacement of a preabsorbed NH3 molecule. 13 Moreover, the displacement of CO bound inside Ni-MOF-74 (binding energy of 53 kJ/mol) is readily driven by CO2 exposure, even though CO2 has a noticeably weaker binding energy of 41 kJ/mol. 14 Here, we have attempted to assess the competitive adsorption of NO2, NO, CO2, CO and H2O by comparing their interaction energies with the copper sites. The results are presented in the Supplementary Table 6. The copper sites have distorted square-planar geometries. Therefore, the binding site of the adsorbed molecule is considered to be on the axial position. In all of these three cases, the water molecule moves away from the copper site and interacts with hydrogen-bonds with the neighbouring hydroxo and aqua ligands of the ZrO2 node.

Supplementary
In the cases of NO and CO, the interaction energies with the Cu(II) sites are weaker than the interaction of NO2. This is consistent with previous works by us 15 and others 16 on the CO and NOx binding with the open metal sites of the HKUST-1 MOF. This is not unexpected, considering that CO and NO interact stronger with Cu(I) sites than the Cu(II) due to significant pi-backdonation. 17 The interaction energies of CO2 with the two Cu(II) sites are computed weaker compared to NO2, with values of -50.0 and -41.2 kJ mol -1 . Unexpectedly, the preferred adsorption site of the water molecule is not on the copper site. During the geometry optimization, the water molecule moves from the axial site of the Cu(II) towards the ZrO2 node and prefers to interact via Hydrogen-bonds with the hydroxo and aqua ligands. Same results are obtained, when a second and third water molecule are inserted on top of the copper site. In all cases, the water molecules move away from the axial binding site of the Cu(II). Although, the interaction energies of H2O with the Cu(II) sites are calculated to be stronger than of NO2, the presence of water molecule will not affect the sensing, because the axial site of one copper centre is still available for binding with NO2 in the presence of water molecules. In the second copper site, the water molecule remains adsorbed on the metal axial position with an interaction energy of -77.3 kJ mol -1 that is significantly stronger than the computed interaction energy of NO2 (-48.5 kJ mol -1 ). Thus, at humid conditions one of the two copper sites will be still available for adsorbing and sensing NO2.
In conclusion, the calculations suggest that NO, CO and CO2 are adsorbed less strongly than the NO2, and that at humid conditions one of the two copper sites will be able to bind NO2.

Supplementary Note 13. Optical sensing
Room temperature photoluminescence measurements in the presence of NO2 (50 ppm NO2 in synthetic air) were performed by enclosing the MOF in a home-made gas chamber with optical access. The MOF powder was loaded in the recess of an aluminum plate covered by a hollow copper grid. The gas flow was controlled with flow meters. The MOF powders were photoexcited by a TEEM Photonics Nd:YAG laser (=355 nm), delivering pulses of 300 ps duration at repetition rates from single shot to 1 kHz. Photoluminescence was free-space collected at approximately 20° respect to the incident beam by an Acton Research SP2500 spectrometer (f= 500 mm) equipped with a Princeton Instruments Spec-10 liquid nitrogen cooled back-illuminated deeply depleted CCD for the acquisition of PL spectra. The scattered light arising from the excitation line was appropriately cut by placing a 370 nm long-pass filter in front of the spectrometer.
Supplementary Figure 27 presents the activation process of Cu-MOF-808 by purging N2 for an hour. As it can be seen, the activation takes to exploit in gas sensing measurements. (All the measurements for Cu-MOF-808 were performed after the activation) Supplementary Figure 27. PL spectra of Cu-MOF-808 upon exposure to N2 gas for activation (exc: 355 nm).
In order to know how the results are comparative, sensing measurements were performed for MOF-808. Similarly, the MOF-808 powder was activated by purging N2 gas for 20 min. According to Supplementary Figure