Introduction

Glasses are ubiquitous in our everyday lives, but still pose fundamental questions about the nature of order in solids. Typically formed by the rapid cooling of a liquid, these amorphous solids have broad applications, with vitreous silica (SiO2) the most well-known example1. Most functional glasses are purely inorganic solids, restricting the range of functional properties feasible2. Recently, a number of metal-organic framework (MOF) glasses containing molecular components have been discovered by heating their crystalline counterparts to a melting point, followed by a rapid cooling, thereby expanding the potential applications of glass materials3,4,5.

MOF glasses combine the unique properties of crystalline MOFs6,7, such as high surface area and chemical versatility8, with the distinctive properties of a glass, such as formability9. This makes this class of molecular materials particularly interesting for applications that benefit from both porosity and an amorphous nature10,11,12, such as advanced gas separation13,14, hybrid electrochemical systems15,16, and photonic applications17.

However, there are currently only a handful of known MOF glasses3,18, primarily derived from a subclass of MOFs known as zeolitic imidazolate frameworks (ZIFs)19. ZIFs are distinguished by their tetrahedral network structure, which closely resembles that of silica and zeolites20. Like silica, these frameworks can melt, and by rapidly cooling the melt, the MOF glass is formed21,22. Conversely, unlike silica, the chemical versatility of these molecular-based solids allows the modification of both the metal centers (e.g., Zn, Co, Fe)23,24,25 and the organic ligands (e.g., imidazolates, benzimidazolates)26,27, causing variations in the melting temperatures. Still, the anionic imidazolate is thermally unstable in the temperature regime of melting. This instability means that melt quenching (MQ), even in inert conditions, can introduce impurities into the glass and is particularly acute for MOF glasses with paramagnetic transition metals, which typically contain impurities of metallic cobalt or Fe3O428,29. Additionally, thermal processing of MOF glasses requires careful control of the time-temperature process, as melt-quenching rates, re-melting of the glass, or direct collapse of the MOF structure, can have large effects on the final properties of the material, including bubble formation, micro-cracking, surface oxidation, and extent of decomposition30,31.

Several approaches have been investigated to bypass the melt-quenched process, such as the mechanochemical method, although it has only proven effective with Zn32. More recently, the desolvation of metal complexes has been studied to create MOF glasses33. In this work, we developed an alternative direct route to MOF glasses, through the reaction of metal precursor and imidazolate ligand34,35,36, that does not mirror synthetic routes for traditional inorganic glasses1, and overcomes the limitations associated with conventional methods for the preparation of molecular-based glasses. Our route allows the preparation of nearly impurity-free MOF glasses containing paramagnetic metals, which permits the exploration for the first time of their intrinsic optical, electronic, and magnetic function. MOF glasses simultaneously possess a long-range structure that is completely disordered, but with a strong short-range order37. This mirrors other topologically disordered materials, where the properties such as electronic or photonic band gaps emerge from the local uniformity, as occurs in amorphous silicon or disordered metamaterials38. Topologically amorphous antiferromagnets would therefore have collective magnetism. Although there are many amorphous magnets, most are ferromagnets with extensive site disorder, and the only structurally characterized example of an amorphous antiferromagnet is ionic FeF339. Thus, achieving a molecular amorphous antiferromagnet would allow for a chemical versatility inaccessible with metallic systems40, as demonstrated herein. Direct synthesis of MOF glasses thus not only expands the range of feasible materials, but also the range of functional properties that can be achieved in MOF glasses, as demonstrated with the realization of MOF-supported photodetectors and topologically amorphous antiferromagnets.

Results

Direct synthesis of ZIFs glasses

Heating at 300 °C for 6 h a combination of ferrocene (0.16 mmol) and a mixture of imidazole:benzimidazole (0.32 mmol), sealed under vacuum, results in the formation of an orange transparent glassy monolith of formula Fe(im)2–x(bim)x, denoted as dg-MUV-29 (dg = direct-glass), which can be prepared with different amounts of imidazole and benzimidazole (x = 0.2, 0.4, 0.5, and 0.6) (more details are in the SI). Note that this route produces polymeric cyclopentadiene as a byproduct.The amount of each ligand incorporated into the framework was determined by 1H NMR spectroscopy (Table S1). Larger amounts of benzimidazole (x ≥ 0.78) yield a new crystalline phase as a byproduct, denoted MUV-28, of formula Fe(im)1(bim)1, whereas the absence of benzimidazole (x = 0) produces the known crystalline MUV-24(coi)25 (Fig. 1 and Figures S7–S9). Interestingly, reaction at lower temperature (150 °C) yields mixtures of crystalline phases which have been identified as MUV-136, MUV-28 or IMIDFE34.

Fig. 1: Structural analysis of compounds synthesized with varying imidazole:benzimidazole ratios.
figure 1

Schematic representation of the structures obtained under high temperature synthetic conditions (300 °C for 6 h) for MUV-24(coi), dg-MUV-29, and MUV-28. In the diagram, tetrahedral FeII centers are illustrated as orange tetrahedra. The imidazolate ligands are shown in light blue, and the benzimidazolate ligands in aquamarine blue. The corresponding powder X-ray diffraction pattern for each material is presented at the bottom, together with the calculated patterns of the crystalline materials. Note that the structure of dg-MUV-29 was arbitrarily created to illustrate the lack of order. Source data are provided as a Source Data file.

Full size image

No detectable Bragg diffraction peaks are observed in the different materials of the dg-MUV-29 family, consistent with a glassy phase. A similar trend has been recently reported using an alloy of two different meltable coordination polymers instead of two different linkers, and depending on the amount of seeding the recrystallization process can be prevented41. Furthermore, differential scanning calorimetry (DSC) measurements exhibit no endothermic signal related to framework melting, but instead display the distinctive glass transition corresponding to their glassy nature (Fig. 2a). The glass transition temperature (Tg), defined as the peak onset temperature of the endothermic calorimetric signal of the glasses, ranges from 205 °C to 245 °C (Fig. 2b) depending on the incorporated amount of benzimidazole (x), and is slightly higher than for MQ ag-MUV-2425 (Tg = 190 °C), of formula Fe(im)2. This effect is consistent with previous observations for the Zn analog, where Tg also increases with bim– concentration42. In all cases, an initial heating cycle up to 200 °C was performed to eliminate factors that could obscure the clear observation of the Tg, such as possible organic impurities or the sample-compartment effect, which might make the first scan particularly challenging. Upon further heating, the dg-MUV-29 decomposes at a temperature of 450 °C.

Fig. 2: Thermal properties of ZIF-derived glasses.
figure 2

a Thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) curves of dg-MUV-29-bim0.48 b DSC of the materials obtained via high-temperature synthesis with variable im⁻ to bim⁻ ratios. c Optical images of various members of dg-MUV-29. Optical scale bars are 100 μm. Source data are provided as a Source Data file.

Full size image

The lack of long-range order in the glasses is confirmed by the absence of sharp features in their X-ray total scattering data. Pair distribution functions (PDF or D(r)) for dg-MUV-29 were extracted from the total scattering data via appropriate corrections and subsequent Fourier transformation and are presented in Fig. 3a. Comparison of these data with crystalline Fe3(im)6(Him)2 and MQ ag-MUV-24 shows that the short-range correlations are similar, although a new peak (peak “C” at 2.44 Å in Fig. 3a) can be detected in the dg-MUV-29 family, corresponding to the distance between carbon atoms in the benzene ring of the benzimidazole ligand (Fig. 3b). The relative intensity of this peak increases as the amount of bim increases, confirming the successful incorporation of bim rings. The PDF peaks extend up to approximately 6 Å, which matches with the distance between neighboring FeII centers (5.8 – 6.1 Å in the crystalline phase MUV-24)25. This clearly indicates that the tetrahedral coordination of the FeII centers with imidazolate linkers is maintained in dg-MUV-29 materials, and that their chemical environment is similar to that of MQ ag-MUV-24. Importantly, the Bragg peaks attributed to crystalline impurities of Al and Fe2N observed in the MQ ag-MUV-2425 data at higher Q values are absent in dg-MUV-29 (Figure S13). These impurities prevented the study of the magnetic properties of ag-MUV-24. In contrast, the direct-glass synthesis avoids the melting process required in the melt-quenched protocol, thereby preventing typical degradation and partial oxidation observed in MQ ag-MUV-2425. The phase purity is unequivocally confirmed by Mössbauer spectroscopy, which is extremely sensitive to the geometry and oxidation state of iron centers. The characterization of the local environment of the metal ion in amorphous phases is very challenging due to the limited number of available techniques. Specific techniques such as 67Zn NMR has greatly advanced the understanding of ZnII centers within a random network, detecting short-range disorder in Zn-based amorphous glass ZIFs (ag-ZIFs), which has been a significant achievement in the field37. The incorporation of FeII centers permits the possibility of using Mössbauer spectroscopy to study for the first time the local environment of a glassy phase. The Mössbauer spectra of dg-MUV-29 measured at 80 K reveals two broad peaks with the same relative area but different widths (Fig. 3c). This is typical of amorphous materials where the FeII centers have the same coordination number but are located on sites with slightly different geometries and environments, leading to slightly different isomer shift (IS) and quadrupole splitting (QS)43. The spectrum can be fitted with a distribution of QS, and the average IS is consistent with tetrahedral coordination of all FeII cations as in MQ ag-MUV-2425. The fitting procedure with this model was tested by calculating χ2. For all spectra, calculated χ2 values are between 1.2 and 1.5, close to 1, indicating that the deviations between the estimated and observed number of counts are mainly due to statistical error. Additionally, Mössbauer spectroscopy make it possible to monitor the oxidation state of the iron centers over time, revealing a 20 % oxidation after 24 h of air exposure (Figure S21).

Fig. 3: Structural characterization of modified ZIF-derived glasses using pair distribution function (PDF) analysis and Mössbauer spectroscopy.
figure 3

a X-ray PDF in the form of D(r) of dg-MUV-29-bim0.48, dg-MUV-29-bim0.18, ag-MUV-24 and IMIDFE. b Schematic representation of the benzimidazolate bridge coordinated to two FeII centers and the connectivity between tetrahedral FeII ions with benzimidazolate linkers. Orange spheres represent Fe atoms, blue spheres represent nitrogen atoms, and black spheres represent carbon atoms. Hydrogen atoms have been omitted for clarity. Mössbauer spectra of dg-MUV-29-bim0.48 (c) ag-MUV-24 (d) and IMIDFE (e) samples, displayed with lines overlaid on the experimental data. These lines represent the sum of doublets and sextets, or distributions of quadrupole splitting (QS) (see Table S2), and are slightly shifted for clarity. Source data are provided as a Source Data file.

Full size image

Chemical diversity

The versatility of the direct-glass synthesis also provides the flexibility to incorporate linkers with diverse functionalities into the glassy phase. Until now, incorporating bulky benzimidazolate ligands with iron derivatives has been impossible because access to the cag topology is required, and that phase has remained elusive. Using the direct-glass synthesis, various analogs have been successfully prepared, denoted dg-MUV-29-X, where X is a different substituent in the position 5 of the benzimidazole ligand (X = NH2, CH3, Br, and Cl).

For the synthesis of dg-MUV-29-X, ferrocene (28 mg, 0.15 mmol) was combined with imidazole (Him) (15.3 mg, 0.23 mmol) and a 5-X-benzimidazole (0.07 mmol) and sealed under vacuum in a 4 mm diameter layering tube. The mixtures were heated at 300 °C for 6 h, producing an orange glass in each case corresponding to dg-MUV-29-NH2, dg-MUV-29-CH3, dg-MUV-29-Br, and dg-MUV-29-Cl, similar in appearance to dg-MUV-29 (Fig. 4a). The same synthetic fraction was maintained in all cases, with x ≈ 0.5. The incorporated amount of each ligand into the framework was determined by 1H NMR spectroscopy (Table S4 and Section S3.3) and closely matches the synthetic amount. The amount of incorporated Xbim can be varied in a controlled manner by adjusting the ratio of the reactants. The incorporation of the different benzimidazolate ligands into the glassy material can be clearly followed for dg-MUV-29-Cl and dg-MUV-29-Br using EDX-SEM, showing that Cl and Br are uniformly distributed throughout the monolithic framework (Fig. S47-S48). PXRD measurements of the resultant monolithic phases revealed the absence of any Bragg diffraction for all analogs (Fig. 4b).

Fig. 4: Structural analysis of compounds synthesized with different benzimidazole derivatives.
figure 4

a Optical images of different dg-MUV-29-X monolithic glass. Scale bars are 100 μm. b X-ray powder diffractograms of dg-MUV-29-X. c Differential scanning calorimetry (DSC) first upscan of dg-MUV-29-X family. To calculate the Tg onset, the material was heated to 200 °C to avoid process that could obscure the Tg. Source data are provided as a Source Data file.

Full size image

DSC measurements of the different analogs of the dg-MUV-29-X family reveals the absence of endothermic peaks corresponding to framework melting but instead show the characteristic Tg of the glassy nature of the material (Fig. 4c). The glass transition is nearly identical across the samples, and reversible, allowing for multiple cycles without any alteration, as also observed in dg-MUV-29 (Fig. S44-S45). The possibility of incorporating bulkier ligands into the iron framework allows further tunability of the system and has an effect on the CO2 gas sorption capability of the glasses (Fig. S49).

Topologically disordered amorphous antiferromagnets

Based on the combined results of all the characterization techniques, we can conclude that the direct-glass synthesis method enables the production of a high-purity FeII phase, which cannot be achieved through conventional melt-quenching methods. This level of purity allows for precise measurement of magnetic correlations of the first phase-pure MOF glass containing paramagnetic ions, providing insight into the nature of the magnetic phase of dg-MUV-29. Bulk magnetic measurements confirmed antiferromagnetic interactions typical of FeII imidazolate MOFs44 (Fig. 5a), as indicated by Curie-Weiss fitting (TCW = –24.7(4) K, implying J = 16.5(3) K with a moment slightly larger than the spin-only value for high-spin tetrahedral FeII (C = 4.19(1) emu mol–1 K, equivalent to g = 2.36)45. Unlike crystalline iron imidazolates, no clear ordering transition at higher temperatures was observed in static susceptibility measurements, confirming the phase purity. AC susceptibility measurements (Fig. 5b) revealed a clear cusp at 5.2(2) K, consistent with the presence of cooperative magnetic interactions. The magnetization as a function of the field was linear, indicating moderately strong antiferromagnetic interactions with no ferromagnetic components (Fig. S22).

Fig. 5: Magnetic properties of ZIF-derived glasses.
figure 5

a Thermal dependence of magnetic susceptibility in the temperature range 2–300 K of dg-MUV-29. The experimental data (expt., black line) have been fitted (red dotted line) following the Curie-Weiss model. b Temperature dependence of the a.c. susceptibility of dg-MUV-29 measured at different frequencies. c Mössbauer measurements at different temperatures of dg-MUV-29. The lines over the experimental points are the estimated distributions of quadrupole splittings (QS). d I(Q) data, offset by 5000; dots represent the experimental data of dg-MUV-29, and the line represents the Monte Carlo (MC) simulation. The MC simulation is matched to temperature, with 1.8 K data plotted against 13 K. An empirical scale, offset, and linear background have been applied to all MC data, fixed to the lowest temperature, and distances are scaled by the experimentally determined metal-metal distance r0 = 6.20 Å. e Real space spin correlation function derived from MC simulations showing antiferromagnetic alternation that decays with increasing distance, r Source data are provided as a Source Data file.

Full size image

The Mössbauer spectrum of dg-MUV-29 at 4 K shows magnetic hyperfine splitting (Fig. 5c), revealing strong magnetic correlations consistent with the transition observed in the magnetization data. The magnetic hyperfine interactions were found to be complex, requiring detailed modeling of the hyperfine interaction parameters, including the quadrupole interaction (Δ = –2.42 mm s–1), asymmetry parameter (η = 0.70), and magnetic hyperfine field (Bhf = 15 T). These parameters suggest significant orbital magnetic momentum and confirm the high-spin FeII state (Fig. 5c). These findings support the hypothesis that dg-MUV-29 is a continuous random network (CRN) antiferromagnet, which would be a new type of magnetic phase. These materials exhibit a completely disordered long-range structure while maintaining a strong degree of local uniformity. This implies that exchange disorder, i.e., variations in the strength and sign of magnetic interactions, are small, and hence the inherent topological amorphousness of the network is driving the observed magnetic properties. Although there are many amorphous ferromagnets, amorphous antiferromagnets are extremely rare. To further investigate this hypothesis, we conducted semi-classical Monte Carlo simulations using the well-studied Wooten-Winer-Weaire CRN model of amorphous silicon46, simulating a 512-spin periodic approximant of the tetrahedral continuous random network (Section S2.9). We note that different materials with tetrahedral CRN structures, though sharing family relationships, are likely to have distinct topological features (e.g., ring statistics), that could play a role in their magnetic function, and so improved structural models will no doubt shed further light on their behavior47. The simulated susceptibility was in quantitative agreement with experimental data, particularly at high temperatures. Below Tc = 13 K, the simulated susceptibility showed a cusp and deviation consistent with a long-range ordering transition, at a temperature which mirrors that of analogous crystalline systems, but higher than observed for dg-MUV-29.

To investigate the local spin correlations, we conducted powder neutron diffraction measurements using diffractometer D20 at the ILL on a deuterated sample across a broad temperature range (1.8 to 100 K) (Fig. 5d). Subtraction of the highest temperature data (100 K) revealed short-range magnetic correlations with a distinct broad peak at 0.79 Å–1. This peak is lower than the first sharp diffraction peak from structural diffraction, and occurs at a similar Q to the crystalline antiferromagnet analogs36, suggestive of antiferromagnetic spin-spin correlations with a correlation length similar to the structure itself. Comparison of the experimental neutron diffraction data with our Monte Carlo simulations, using the metal-metal distance determined from PDF measurements, r0 = 6.20 Å, showed qualitative agreement, particularly from simulations near the ordering transition of the Monte Carlo simulations at Tc = 13 K. However, we note that deviations between experiment and simulation at low temperatures highlight the challenges of modeling amorphous materials, in particular the finite size of our periodic approximant and complexities beyond the minimal model of nearest-neighbor antiferromagnetism (Section S2.11). Our Monte Carlo simulation shows spin and quadrupole correlations indicating a speromagnetic state (Figs. 5e and S32), characterized by a disordered arrangement of magnetic moments (spins) on a local scale, while still maintaining no net magnetization across the entire system48.

The combined results from neutron scattering and magnetometry strongly indicate that this material behaves as an amorphous antiferromagnet with a continuous random network structure and predominantly nearest-neighbor Heisenberg interactions. The reported magnetic dg-MUV-29 family is an unusual kind of amorphous magnet, where local exchange and single-ion anisotropy disorder is expected to be small, meaning the intrinsic topological disorder dominates their properties. Unlike previous amorphous antiferromagnets43, this material exhibits minimal exchange disorder because each Fe ion has a similar local environment with four nearest neighbors, and Fe is the sole framework-forming cation49,50. This contrasts with amorphous metals, which typically exhibit significant exchange and site disorder40.

Scalability and processability

The ability to bypass the melting process and eliminate the need for a crystalline intermediate enables easier scalability compared to the conventional melt-quenching approach. In melt-quenching, a monolith forms by fusing individual crystals upon reaching the liquid state. In contrast, this method allows the amorphous material to directly form during synthesis, increasing its uniformity. This enables the production of large quantities of glass in film form simply by increasing the size of the layering tube used as a reactor. To demonstrate the scalability and processability of dg-MUV-29, we have used it as a substrate for a centimeter-scale thin-film photodetector based on methylammonium lead iodide (MAPbI3), replacing the conventional glass typically used. MAPbI3 is an archetypal material within the broader metal halide perovskite family, a class of semiconductor with promising applications in optoelectronic devices such as photodetectors, among others51. MOFs have been used in combination with perovskites as additives (blended with the active material), host materials, or functional layers in order to improve the device performances17,52,53,54, but their use as substrates in optoelectronic devices remains elusive, as producing large-area homogenous and monolithic MOFs is challenging. As most perovskite optoelectronic devices are fabricated on soda-lime glass, dg-MUV-29 is used here as the substrate for a simple proof-of-concept lateral perovskite photodetector. After isolating a centimeter-size monolithic dg-MUV-29 (see Fig. S35), we used vacuum deposition55,56 to process the perovskite on top of the glass by co-sublimation of PbI2 and MAI, followed by evaporation of gold contacts through a shadow mask (Fig. 6a, b). As shown in Fig. S38, the diffraction pattern and absorption profile confirm the formation of MAPbI3 on dg-MUV-29. The current of the resulting dg-MUV-29/perovskite photodetector was measured as a function of the applied field under illumination (white light, 100 mW·cm−2) and in the dark, as shown in Fig. 6c. The device showed the expected photoconductive behavior, with the current increasing more than three orders of magnitude upon illumination, with an on/off ratio > 1000 at 0.4 V/μm. Note that the variation between several devices is minor, indicating a good reproducibility of the detector fabricated on dg-MUV-29. The on-current was found to be also stable for several consecutive measurements (Fig. 6d), which indicates an overall good quality of the MAPbI3 film grown on the dg-MUV-29 substrate, as well as the mechanical stability of the MOF itself. Only a minor difference is observed between devices deposited on commercial float glass and the dg-MUV-29 (Fig. S37), likely arising from the different optical quality of the two materials.

Fig. 6: Perovskite photodetector fabricated on a ZIF-derived glass substrate.
figure 6

a Schematics and b photograph of the perovskite photodetector prepared on the dg-MUV-29 substrate. c Current vs. applied field for perovskite detectors on dg-MUV-29, recorded under illumination (orange) and in the dark (light blue). The thick red and blue curves are the mathematical average of the corresponding curves recorded under illumination and in the dark, respectively. d Current response of the device at an applied field of 0.4 V/μm when exposed to consecutive light pulses. Source data are provided as a Source Data file.

Full size image

Discussion

The solvent-free methodology presented here facilitates the direct synthesis of a previously unreported family of MOF-derived glasses, designated dg-MUV-29, eliminating the necessity for a pre-existing crystalline phase. This innovative approach yields high-purity FeII glasses while effectively mitigating the oxidation and degradation issues commonly associated with traditional melt-quenching techniques. In this work, we have aimed to demonstrate a comprehensive research approach involving three key aspects, where direct synthesis has played a pivotal role in advancing various areas. Firstly, the versatility of this synthetic technique allows for the incorporation of a diverse array of ligands featuring varying functional groups, leading to the development of the expanded dg-MUV-29-X family. Such adaptability broadens the scope of potential applications and avenues for future research involving these materials. Secondly, we have shown that the resulting free-standing films of glassy MOFs can be used as substrates for device integration, and have demonstrated this applicability with the preparation of a photodevice using dg-MUV-29 as a substrate. Finally, this methodology establishes a robust platform for probing the magnetic properties of molecular-based random networks, thereby enhancing our understanding of the magnetism inherent in disordered systems. In summary, this direct-glass synthesis method proves to be a highly effective strategy for advancing the development of glassy ZIFs, thus paving the way for new research initiatives and applications. The implications of glassy ZIFs in the study of disordered systems underscore their potential to significantly impact the field of materials science.

Methods

Synthesis

Ferrocene (0.15 mmol) was combined with a mixture of imidazole (Him) (range: 0.18–0.30 mmol) and benzimidazole (Hbim) (range: 0–0.12 mmol) or 5-X-benzimidazole (0.07 mmol), maintaining a total ligand amount of 0.30 mmol. The mixture was subjected to at least three Ar–vacuum cycles to remove moisture and oxygen, then sealed under vacuum in a 4 mm diameter layering tube. It was heated at 300 °C for 6 h, producing an orange glass. After cooling to room temperature, the tube was opened and unreacted precursors were removed using acetonitrile. Although the resulting materials remain stable in air for several hours and oxidize only slowly, sensitive measurements—such as neutron diffraction and Mössbauer spectroscopy—are carried out under strict inert atmosphere conditions to prevent any risk of oxidation.

X-ray powder diffraction

The samples were lightly ground in an agate mortar and pestle and used to fill 0.7 mm borosilicate capillaries that were mounted and aligned on a Bruker D8 Discover powder diffractometer, using Cu Kα radiation (λ = 1.54056 Å). The detector was an EIGER2 R 500 K, multi-mode 2D.

Differential scanning calorimetry

Differential scanning calorimetry (DSC) measurements were conducted using a TRIOS DSC 250 instrument. Activated samples (10–15 mg) were placed in an aluminum crucible (30 μL) with a pierced lid. All the measurements were performed under a continuous N2 flow of 50 ml min−1. The sample was initially heated to 40 °C, followed by a 15-minute isotherm to stabilize the environment. Subsequently, the sample was heated at a rate of 10 °C min−1. Upon reaching the final temperature, a 10-minute isotherm was performed to ensure a complete phase change. The sample was then cooled back at 10 °C min−1. It is important to note that DSC results are highly dependent on the experimental conditions. Therefore, when changing the equipment, parameters such as N2 flow, sample amount, or size of the hole in the crucible lid (prepared using a standard needle of approximately 0.6 mm diameter) may need to be adjusted to accurately observe the phase changes.

Thermogravimetric analysis

Thermogravimetric analysis (TGA) measurements were conducted using a TGA 550 instrument (TA Instruments). All the samples were loaded in a platinum pan, under air or N2 atmosphere. Initially, the sample was heated to 40 °C and held isothermally for 15 min to stabilize. Subsequently, the sample was heated at a rate of 10 °C min−1.

Mössbauer Spectroscopy

Mössbauer spectra were collected in transmission mode using a conventional constant-acceleration spectrometer from Wissel (Starnberg, Germany) and a 50 mCi 57Co source in a Rh matrix. The velocity scale was calibrated using α-Fe foil. Isomer shifts, IS, are given relative to this standard at room temperature. The absorbers were obtained by gently packing the samples into Perspex holders. Absorber thicknesses were calculated on the basis of the corresponding electronic mass-absorption coefficients for the 14.4 keV radiation57. The low temperature measurements at 4 K and above were performed with the sample in liquid He and in He exchange gas, respectively, in a Janis (Westerville, OH, USA) bath cryostat, model SVT-400. The spectra were fitted to Lorentzian lines using a non-linear least-squares method58. Line widths and relative areas of both peaks in each doublet were kept equal during the refinement. Distributions of quadrupole splittings were fitted according to the histogram method59.

X-ray total scattering

Neutron total scattering data were collected at room temperature using a PANalytical Empyrean laboratory diffractometer, equipped with an Ag-Kα source and focusing mirrors. The samples were placed in 1 mm diameter quartz glass capillaries. Each sample underwent multiple scans, with a cumulative data collection time exceeding 24 h per sample. Additionally, measurements were conducted on an empty capillary and the diffractometer background. The resulting X-ray total scattering patterns were processed using the GudrunX60 program to produce a normalized Pair Distribution Function (PDF), optimized to ensure, for example, that the low-r portion of g(r) oscillates around –1. The total scattering structure factor, S(Q), was Fourier transformed for 0.6 < Q < 18.5 Å−1 to obtain the PDF.

Neutron total scattering

X-ray total scattering data were collected at temperatures of 5 K (base), 10 K, 15 K (just above the transition), 30 K, 60 K, and 120 K. Each temperature required a collection time of at least 3 h per sample. The sample was measured using the ILL D20 instrument at a wavelength of 2.41 Å. The sample was predominantly deuterated, achieved by using deuterated imidazole (Dim, C3N2D4) and deuterated ferrocene [Fe(C5D5)2], with a minor hydrogen presence from benzimidazole (Hbim).

Magnetic susceptibility (SQUID)

Magnetic susceptibility measurements were conducted on powdered samples (5–20 mg) using a SQUID magnetometer (Quantum Design MPMS-XL-5 and MPMS-XL-7). Variable-temperature (2−300 K) direct current (d.c.) magnetic susceptibility measurements were performed in applied fields of 1.0 kOe, along with variable field magnetization measurements up to ±5 T at 2.0 K. Additionally, variable-temperature (2−10 K) alternating current (a.c.) magnetic susceptibility measurements were conducted in a ± 4.0 G oscillating field at frequencies ranging from 1 to 996 Hz, under a zero d.c. field.

Photodetector fabrication

CH3NH3I (MAI) and PbI2 were purchased from Luminescence Technology Corp. All materials were used as received. For MAPbI3 deposition, MAI and PbI2 precursors were simultaneously evaporated using the following process. The evaporation chamber employed has only two quartz crystal microbalances (QCMs), one close to the PbI2 source designated to monitor exclusively the PbI2 precursor (PbI2-QCM), with no cross reading, and a second one fixed at the height of the substrates to monitor simultaneously the total amount of PbI2 and MAI mass reaching the substrates (MAPbI3-QCM). Initially, only PbI2 is heated, and the temperature is fine tuned to lead to a stable sublimation rate of precisely 0.50 Å s−1 on both PbI2-QCM and MAPbI3-QCM. Then, MAI is heated, and the temperature adjusted so that the sublimation detected on the MAPbI3-QCM increases from the previous 0.50 Å s−1 to 0.66 Å s−1, while the rate on the PbI2-QCM is kept stable at 0.50 Å s−1. During the evaporation the pressure of the chamber was maintained at 5·10−6 mbar. Gold contact were evaporated in another vacuum chamber through a shadow mask, defining lateral electrodes with channel length of 500 µm and width of 500 mm. The devices were encapsulated with Al2O3 (30 nm), deposited by atomic layer deposition in a Arradiance reactor at 40 °C52.