Framework CSMCRI-11 was synthesized as colourless rectangular crystals under solvothermal conditions by reaction of bifunctional ligand 4- (1H-imidazol-1-yl) benzoic acid (HL), 4,4′-bipyridine (bpy) and Cd(NO₃)₂·4H₂O in the molar ratio of 0.2:0.2:0.16. As-synthesized CSMCRI-11 crystal was examined by single-crystal X-ray diffraction (Supplementary Table S1) which displays monoclinic space group C2/c. The asymmetric unit contains two Cd(II) ion, one bpy linker, two deprotonated ligand (L ⁻¹ : hereafter L ) and one nitrate anion. Both the metal centres exhibit different coordination environments. For instance, Cd(1) is coordinated with four oxygen atoms from the carboxylate group of L and two nitrogen atoms (N atoms N3 and N5) from the pyridyl ring of bpy linker in distorted octahedral geometry (CdO₄N₂). The Cd-O bond length range from 2.292 to 2.419 Å while Cd-N bond length ranges from 2.285 to 2.297 Å. (Zhai et al., 2019).

Two alike Cd (1) ions are bridged through two carboxylate groups in syn-anti fashion to form a binuclear [Cd₂(CO₂)₄] subunit (Scheme 1) with Cd(1)-Cd(1) spacing of 3.986 Å. On the other hand, Cd (2) atom is in entirely different coordination milieu, and reveals distorted octahedral geometry (CdO₄N₂) via ligation with four imidazolyl nitrogen atoms of L in the equatorial plane, and two oxygen atoms of nitrate anions. Here, Cd-O bond length is 2.358 Å, and Cd-N bond length varies from 2.296 to 2.318 Å. Every deprotonated ligand is in coordination with [Cd₂(CO₂)₄] subunits and Cd (2) ions forming two-dimensional (2D) layer along the crystallographic b axis.

The weavy layer acts as a roof with rhombus window (Figure 1A) of size 16.4 × 18.8 Ų (considering atom to atom connection). Two bpy linkers connects every [Cd₂(CO₂)₄] units, leading to a double-pillar layered three-dimensional (3D) structure with the formation of large voids (dimension: 11.4 × 24.5 Ų) along the c axis. Such huge pores mutually instigate two-fold interpenetration (Figure 1D) to the overall structure of CSMCRI-11. Nevertheless, rectangular cavities of dimension 6.54 × 18.09 Ų still exists along ab plane. Also, pillaring bpy linkers are accompanied by string π–π interaction (centroid–centroid distances in the range: 3.794 Å), imparting high rigidity to the 3D porous network.

These porous channels accommodate DMF and H₂O solvent molecules, which are highly disordered. However, their presences were alternatively ascribed from a combination of PLATON calculation, thermogravimetric weight loss, elemental analysis, and IR spectral data that corresponds to the molecular formula of CSMCRI-11 [Cd_1.5(L)₂(bpy)(NO₃)]·DMF·2H₂O. Regardless of two-fold interpenetration, the solvent accessible voids of CSMCRI-11 is estimated to be 31.6% of the total crystal volume.

The purity of bulk phase and structural integrity of CSMCRI-11 was established from strong correlations of peaks between powder X-ray diffraction (PXRD) pattern of as synthesised framework and simulated one (Figure 2A), obtained from crystallographic data. The FT-IR spectral data exhibits IR band of the >C═O (carbonyl) for guest DMF solvent at 1663 cm⁻¹ (Supplementary Figure S4). Also, a broad absorption peak centred around 3300 cm⁻¹ corresponds to lattice water molecules. Thermal stability of the framework was assessed from thermogravimetric analysis (TGA) under inert atmosphere (Supplementary Figure S5), showing a two-step weight-loss. The first weight loss was observed in the temperature range 30–100 ^oC and corresponded to 3.89% of the initial mass, accounting for the removal of lattice water. The second weight loss of 8.89% features loss of DMF solvents. A sharp weight loss beyond 350 ^oC relates to framework decomposition. For, generation of the guest-free (activated) framework (hereafter 11a), the as-synthesised crystals of CSMCRI-11 were immersed in methanol for three days (by exchanging the solvent three times a day), and solvent exchanged framework was heated under vacuum at 120 ^°C for overnight. The FT-IR spectra analysis of 11a shows absence of all the peaks related to H₂O or DMF solvents. Furthermore, the TGA curve of 11a did not reveal any weight loss up to 320 ^°C, supporting its robust nature at high temperature (Supplementary Figure S5). To further corroborate stability of the structure, 50 mg of as-made CSMCRI-11 was taken in a vial and exposed to common organic solvents (methanol, acetonitrile, dichloromethane, tetrahydrofuran, acetone), and water. Importantly, PXRD patterns of these exposed samples remained unchanged (Figure 2B). We further investigated the hydrolytic stability of the framework by exposing it to saturated water vapours at room temperature. A time-dependent PXRD (Supplementary Figure S6) study up to 10 days was carried out at a regular interval, which revealed maintenance of structural integrity throughout, and substantiates to sufficient hydrolytic stability of the MOF. The stability aspect was further assessed by leaving 11a in the open air for 15 days. Remarkably, PXRD pattern (Figure 2B) remained unchanged and corroborates to its high robustness.

Strategic incorporation of electronically passive transition metal in association with conjugated organic struts is one of the potential strategies for the synthesis of luminescent MOFs. (Zhu et al., 2015; Zhang et al., 2018; Zhou et al., 2018) Owing to the excellent thermal and hydrolytic stability, together with its highest luminescence intensity in water (Supplementary Figure S7c), CSMCRI-11 was employed as a fluorescent probe in activated form for aqueous phase minute detection of water contaminants. Upon excitation at 265 nm, aqueous dispersion of 11a showed strong emission band at 426 nm that is attributed to mixed contribution of the intra-ligand charge transfer of π-electron-rich links, (Chen et al., 2007) ligand-to-ligand charge transfer (LLCT) (π−π* and n−π* transitions) (Luo et al., 2012; Wan et al., 2015) and adequate π−π stacking interaction between the ligands in the structure. (Zhao et al., 2016; Liu et al., 2019) Additional experiment showed that after filtration of MOF particles, no fluorescence (Supplementary Figure S7d) could be observed in the remaining solution supporting the fact that above emission originates as a result of dispersed MOF particles only. UV-visible spectrum (Supplementary Figure S7a) of 11a shows a redshift of 9 and 3 nm in comparison to bpy and HL, respectively that are associated with binding of linkers to the Cd(II) metal ions. Remarkably, the intensity of the photoluminescent (PL) spectrum of 11a was found to be > 90% higher than constituting ligands. Coordination interactions through Cd(II) centres enhance the rigidity of the structure and boosts the PL intensity to such extent.

Owing to high photoluminescence intensity, 11a has been employed as a probe for selective and sensitive detection of metal ions in the aqueous phase. 1 mg of powdered 11a was dispersed in 2 ml water and sonicated for 2 h to generate a uniform suspension (1 mg/2 ml dispersion of 11a for each titration) Photoluminescence spectra of the aqueous phase dispersion was recorded by incremental addition of a series of MCl_y metal salts (M: Ba²⁺, Fe³⁺, Fe²⁺, K⁺, Mn²⁺, Ca²⁺, Cd²⁺, Pd²⁺, Co²⁺, Al³⁺, Ni²⁺, Cu²⁺, Cr³⁺, Mg²⁺, and La³⁺) in water (2.5 mM). For all these cases, the luminescence spectra were recorded at an emission wavelength of 426 nm under constant stirring at room temperature upon incremental addition of analyte (20–120 µl), and every titration was performed in triplicates to ensure the reproducibility of the results. Surprisingly, a rapid and significant decrease in the luminescence intensity of 11a was detected (Figure 3A) only in case of Fe³⁺ ion solution, while remaining metal ions did not show any significant changes. The quenching efficiency of Fe³⁺ ion is found to be 91.6% (Supplementary Figure S8), whilst the rest of the metal ions showed almost nominal values. The quenching competence follows the trend (Supplementary Figure S8): Fe³⁺>Cr³⁺>Cu²⁺>Ni²⁺>Fe²⁺>La³⁺>Cd²⁺>Mg²⁺>Al³⁺>Mn²⁺>Ba²⁺>Ca²⁺>Co²⁺>Pd²⁺>K⁺.

Quantitative analysis of the fluorescence quenching of 11a by Fe³⁺ ions was determined by Stern-Volmer (S-V) equation (Parmar et al., 2017a; Parmar et al., 2017b): I₀/I = 1 + K _sv [C] (I and I₀ are photoluminescence intensities of the 11a after and before titration with analytes, respectively; [C] is the molar concentration of analytes ions in mM; K _sv is known to be S-V constant in M⁻¹). Change in the PL intensity was recorded upon incremental addition of an aqueous solution of Fe³⁺ ions (2.5 mM) to the aforesaid MOF dispersion in water (Figure 3B). The bent S-V plot at higher Fe³⁺ concentration might be attributed to the dynamic and static quenching process. (Chen et al., 2018) For the calculation of K _SV, a linear correlation between the PL intensity and Fe³⁺ concentration was considered at lower concentration range (0–0.08 mM), which resulted remarkable value of 1.91 × 10⁴  M⁻¹ (Table 1) that stands well in literature reports for detection for Fe³⁺ ions (Supplementary Table S6). Further, limit of detection (LOD) was obtained from changes in the emission intensity on incremental addition of 10 µM solution of Fe³⁺ ions. By employing the standard equation 3σ/K (Nagarkar et al., 2014; Karmakar et al., 2017) (σ = standard deviation of initial intensity of the MOF without analyte for five consecutive blank measurements at 2 min intervals, K = slope of the linear curve in Supplementary Figure S9) the LOD value for Fe³⁺ turned out (Supplementary Table S3) to be 0.219 μM (corresponding to 166 ppb). Given quenching efficiency, K _SV value, and LOD are important and desired criteria for sensing application, Fe³⁺ ion sensing by 11a truly corroborates ultra-sensitive detection in comparison to literature reports (Supplementary Table S6).

To further corroborate real-time applicability, visual detection of Fe³⁺ ions was targeted in the solution phase. As divulged in Figure 7B, the high fluorescence of aqueous dispersion of 11a under UV light (365 nm) immediately quenches upon one drop addition of Fe³⁺ solution, and authenticates to solution-phase visual detection of Fe³⁺ions.

Given presence of other metal ions cannot be neglected in real systems, competitive analysis test (CAT) was considered. To a 2 ml suspension of MOF, 120 µl of interfering analyte (2.5 mM) was added first, followed by similar amount of Fe³⁺ ion. While PL intensity did not change during the addition of interfering analytes, the solution exhibited more than 90% decrease of emission upon second addition. Further, to ensure the reusability of the material, MOF particles were separated via centrifugation after sensing experiments, washed thoroughly with water followed by methanol, and air-dried. The intensity of the recycled material was restored up to five consecutive sensing-recovery cycles (Figure 4A). Structural integrity and purity of reused 11a were verified from unchanged PXRD patterns to that of pristine ones. In addition, the response time of 11a towards sensing of Fe³⁺ was studied by recording the time-dependent change in the luminescence intensity. (Goswami et al., 2019b). For this, 40 μl of Fe³⁺ solution (2.5 mM) was added to the MOF dispersion, wherein luminescence intensity (Figure 4B) dropped to 61.9%. The change in PL intensity was recorded for 120 s after a regular interval of 20 s. Remarkably, no significant change in the photoluminescence intensity was observed beyond 20 s, and affirms fast responsive Fe³⁺ ion detection in aqueous phase.

Selective luminescent detection of Fe³⁺ in highly sensitive and fast responsive manner inspired us to explore further application of this framework towards sensing of anions. We performed anion detection considering potassium salts of a series of anions (NO₃ ⁻, SO₄ ²⁻, MoO₄ ²⁻, NO₂ ⁻, Cl⁻, F⁻, PO₄ ³⁻, SCN⁻, I⁻, MnO₄ ⁻, Br⁻, CrO₄ ²⁻ and Cr₂O₇ ²⁻) in water (2.5 mM). These solutions were separately added to the aqueous dispersions of 11a. Among others, a significant quenching of PL intensity was observed by hexavalent oxo-chromium anions (Cr₂O₇ ²⁻, CrO₄ ²⁻), whereas rest of the anions did not show any major changes in luminescence intensity. The order of quenching (Supplementary Figure S10) was found to be Cr₂O₇ ²⁻>CrO₄ ²⁻>MnO₄ ⁻>SCN⁻>MoO₄ ²⁻>I⁻>NO₃ ⁻>NO₂ ⁻>SO₄ ^2–>Cl⁻>Br⁻>PO₄ ³⁻. The quenching efficiency for Cr₂O₇ ²⁻ and CrO₄ ²⁻ are 96.13 and 87.15%, respectively. For quantitative determination of the extent of quenching, changes in luminescence intensity of 11a were recorded by incremental addition of anionic solution. Stern–Volmer constant was obtained (vide supra) via plotting [(I₀/I)-1] vs. concentration of the analyte that revealed linear curve at the lower concentration (Figure 5C,D) and deviation from linearity occurs as the concentration raises. K _sv values were calculated from linear portion of the curve and found to be 1.46 × 10⁴ and 2.18 × 10⁴ M⁻¹ for CrO₄ ²⁻ and Cr₂O₇ ²⁻, respectively. In addition, LOD (vide supra) was evaluated by the gradual addition of 10 μM aqueous anionic solutions to 2 ml MOF dispersion. From linear fitting of the graph (Supplementary Figure S11, Supplementary Figure S12) involving intensity against analyte concentration, LOD was calculated to be 179 and 114  ppb for CrO₄ ²⁻ and Cr₂O₇ ²⁻, individually. A comparison of MOF-based fluorescence sensors for the detection of lethal oxo-anionic Cr(VI) pollutants in water is tabulated in Supplementary Table S7, which indicates that quenching constants from this study rank one of the best values among existing reports. Essentially, such low detection limits render 11a a suitable candidate for monitoring of these lethal oxo anions in water.

Keeping in mind that real-system possesses many other interfering anions, we next probed selectivity of 11a towards detection of Cr₂O₇ ²⁻ and CrO₄ ²⁻ ions by performing CAT, maintaining standard protocol (vide supra). In a distinctive experiment, 120 µl of interfering analyte solution (2.5 mM) was added, which resulted almost insignificant changes to the emission behaviour of 11a. Quite in contrast, successive addition of same amount of individual oxo-anionic Cr(VI) solution to the above mixture (MOF + interfering anions) leads to a drastic turn-off response to the PL intensity (Figures 5F,G). It is imperative to state that we repeated this experiment three times to ensure the chronology of results. Such remarkably selective sensing of Cr(VI) oxo-anion is of true elegance and validates 11a as a potential and reliable probe for detection of Cr(VI) oxo-anions in aqueous phase.

Given reclamation and reusability of a heterogeneous sensory material is of crucial importance, 11a was centrifuged after sensing experiments, washed thoroughly with water, followed by methanol and finally air-dried. It was observed that PL intensity could be restored up to five sensing-recovery cycles (Figure 6A) and indicative of consistency of its performance towards hexavalent chromium ion detection. Structural robustness of 11a during repetitive sensing was additionally assured from intact positions of all peaks in its PXRD pattern (Supplementary Figure S13) after sensing experiment. Further, fast-responsive analyte test (FRAT) was conducted to confirm the response time. Decrease in the intensity was recorded upon addition of 40 µl of Cr(VI) oxo anion solution to the MOF dispersion as a function of time. Emission response decreased to 69.9 and 58.2% for CrO₄ ²⁻ and Cr₂O₇ ²⁻ ions (Figures 6B,C), respectively. Change in PL intensity was recorded up to 120 s after a regular interval of 20 s, which revealed no significant change in luminescence intensity beyond 20 s. These results further validated fast responsive sensing of both the Cr(VI) anionic species in water. Motivated by excellent solution-phase of detection, an attempt was made to visually detect Cr(VI) ions in the aqueous phase. MOF dispersion was exposed to UV light (265 nm), which resulted bright blue emission. Addition of just one drop of either CrO₄ ²⁻ or Cr₂O₇ ²⁻ ions instantaneously diminished (Figure 7B) this emission and supports real-time visual monitoring as well.

To look into the sensing mechanism of Fe³⁺ ions and Cr(VI) oxo-anions, a series of experimental studies were performed. In all the cases, quenching of the PL intensity due to collapse of the framework was discarded on the basis of unaltered PXRD patterns (Supplementary Figure S13) after sensing. The possibility of exchange of Fe³⁺ ions with the constituting Cd(II) ions is nullified from fast-responsive detection, in a sense that trans metalation reaction in such a short time span is not possible (Figure 4A). Another probability includes adsorption of Fe³⁺ and/or Cr₂O₇ ²⁻/CrO₄ ²⁻ ions within the framework. To this end, 11a was separately immersed to aqueous solutions of Fe³⁺, Cr₂O₇ ²⁻and CrO₄ ²⁻ for about a day. Resultant products Fe³⁺@11a, CrO₄ ²⁻@11a and Cr₂O₇ ²⁻@11a were analysed by inductively coupled plasma optical emission spectrometry (ICP-OES), which validated no encapsulation of these ions even after prolonged exposure from the ratio of Cd (II) and other ions (Supplementary Table S2). Moreover, ICP-OES of the supernatant solution showed absence of Cd²⁺ ions and further corroborated no leaching of Cd(II) ions. Similarly, FT-IR analysis (Supplementary Figure S14) of the samples did not show any additional peak and eliminates the chances of encapsulation of these ions. As a matter of fact, the easy recyclability of 11a after every detection through simple washing also indicates weak surface interaction between the studied analytes and the MOF.

In this milieu, the competition for the excitation energy absorption between 11a and other ions may lead to the change of the luminescent signals. (Chen et al., 2017b; Dong et al., 2017). Figure 7A shows the UV absorption spectra of Fe³⁺, CrO₄ ²⁻ and Cr₂O₇ ²⁻ plotted together with absorption and emission spectra of 11a. For Fe³⁺, relatively broad and moderate adsorption was perceived covering both excitation and emission peak of 11a. Of note, other metal ions have no such conspicuous absorption overlapping with the framework. This leads us to conclude that Fe³⁺ ions can absorb the excitation and emission light of 11a, thereby decreasing its luminescence intensity. (Guo et al., 2019) On the other hand, competitive energy absorption was envisioned as a valid, alternative reason for optical response quenching. CrO₄ ²⁻/Cr₂O₇ ²⁻ have two absorption bands in the range 220– 450 nm that cover a wide range of excitation and emission bands of 11a. In contrast, other anions do not absorb in this particular wavelength region. So excitation energy will be strongly absorbed only by these hexavalent Cr(VI) anions, reducing UV-vis absorption of the 11a, and resulting in substantial quenching. (Yi et al., 2015) Conversely, absorption spectra overlap of other metal ion and anions with absorption spectra of 11a show no significant overlap (Supplementary Figure S15a, Supplementary Figure S15b) and validates selective detection of studied analytes. Indeed, the excellent extent of overlap between the absorption spectra of Fe³⁺, Cr(VI) oxo-ions and emission spectrum of 11a helps in producing fluorescence quenching owing to the resonance energy transfer (RET).

The ligand was synthesised and characterised by previously reported method (Refer to SI).

A mixture of Cd(NO₃)₂∙4H₂O (48.4 mg, 0.156 mmol), 4,4ʹ-bipyridyl (31.25 mg, 0.2 mmol) and 4-(1H-imidazol-1-yl)benzoic acid (HL) (37.23 mg, 0.19 mmol) was dissolved in N,N-dimethylformamide (DMF; 7 ml) sealed in 15 ml glass vial, and heated at 120 °C for 2 days. The colourless, rectangular crystals were isolated (Supplementary Scheme S2) in 65% yield.

Anal.Calcd. for C₃₃H₃₃N₈O₁₀Cd_1.5 = [Cd_1.5 (L) ₂ (bpy)(NO₃)]·DMF·2H₂O: C, 45.54; H, 3.82; N, 12.88%; found: C, 44.23; H, 3.92; N, 12.43%. FT-IR (KBr) analysis: 1303, 1399, 1563, 1612, 1663, & 3414 cm⁻¹ (Supplementary Figure S4).

As synthesised crystals were washed with fresh DMF and finally dried in air. The guest solvents in CSMCRI-11 were exchanged with methanol by soaking the crystals in methanol for 3 days followed by exchanging the solvent 3 times a day. The crystals were dried overnight under vacuum at 120 °C to generate a solvent-free framework 11a.

Anal.Calcd. for C₃₀H₂₂N₇O₇Cd_1.5 = [Cd_1.5( L )₂(bpy)(NO₃)] C, 47.34; H, 2.91; N, 12.88%. found: C, 46.17; H, 3.01; N, 12.45%.