Uncapping energy transfer pathways in metal-organic frameworks through heterogeneous structures

Abstract

Metal-organic frameworks (MOFs) are porous, crystalline materials known for their structural versatility and high thermal stability, making them promising candidates for light-emitting diode applications. Distinct classes of MOFs, such as multivariate (MTV)-MOFs and MOF-on-MOFs, introduce heterogeneity by incorporating multiple ligands within a single unit cell (MTV-MOFs) or by stacking different MOFs on top of each other (MOF-on-MOF). Although these strategies improve their properties, the mechanisms of energy transfer between their heterogeneous components and their effects on optical properties, such as quantum yields, remain poorly understood. In this study, we demonstrate that MOF heterostructures significantly improve quantum yield compared to single-ligand-based MOFs. Using Zr-UiO-67 as the base MOF, we generated an array of MOF-on-MOFs and analogous MTV-MOFs utilizing other Zr-MOFs constructed from 4,4’-azobenzenedicarboxylate or 4,4’-stilbenedicarboxylate linkers. Our results demonstrate that, although the emission color coordinates for the stilbene-based materials are identical, the MTV-MOFs increase the quantum yield from 8.2% to 10.2%, whereas the MOF-on-MOFs reach a quantum yield of up to 40.0%.

Distinct classes of MOFs, such as multivariate (MTV)-MOFs and MOF-on-MOFs, introduce heterogeneity by incorporating multiple ligands within a single unit cell (MTV-MOFs) or by stacking different MOFs on top of each other (MOF-on-MOF). Here authors show that MOF heterostructures significantly improve quantum yield compared to single-ligand-based MOFs.

Introduction

The imperative need for advanced, efficient fluorescent materials is driven by the substantial energy consumption of lighting systems. In the United States, lighting and display technologies consume approximately 213 billion kilowatt-hours annually¹. This significant energy usage contributes to around 13% of global CO₂ emissions, predominantly resulting from the burning of fossil fuels, which generate 80% of the country’s electricity². In light of this, developing more efficient lighting and display technologies is critical for reducing energy consumption and mitigating climate change. Metal-organic frameworks (MOFs) have garnered significant attention for their potential in lighting and display devices due to their high crystallinity, modularity, and stability^3–7. Of particular interest are luminescent MOFs, composed of lanthanides and/or aromatic and π-conjugated ligands^8–11, exhibiting enhanced light absorption and emission properties^9,12–15. While various strategies have been explored to develop luminescent MOFs—such as employing multiple linkers, incorporating metals, or encapsulating luminescent guest molecules, oxides or other species within their pores—the quantum yield of MOFs remains low due to several factors. These include symmetry, defects, π-π stacking, aggregation-induced quenching, and Energy Transfer^8,16–18. Despite ongoing efforts to enhance quantum yields through metal doping, modulation of guest molecules, and mixed linker systems, substantial challenges remain. These challenges include reproducibility, cost and environmental impacts, particularly when rare earth metals are incorporated^9,19–26.

While lanthanide-based MOFs have demonstrated high emission efficiencies, with quantum yields exceeding 90%^8,26–31, their widespread use is limited by the high cost and environmental impact associated with lanthanide extraction^31,32. In contrast, MOFs that rely on ligand-centered emission typically achieve significantly lower quantum yields, generally not exceeding 25%⁸. An alternative strategy for achieving high quantum yields with MOFs involves multivariate (MTV)-MOFs, which utilize a combination of organic ligands to generate a single-phase material with a range of functionalities^33,34. MTV-MOFs differ from defective MOFs in that, while defects in conventional MOFs typically arise from missing linkers or metal clusters, MTV-MOFs instead introduce substitutions at ligand or cluster sites without creating vacancies^35,36. Despite this innovative approach, the quantum yields of MTV-MOFs remain low. For example, JLNU-7 (JLNU: Jilin Normal University; [(Zn₄O)(L)(naphthalene-2,6-dicarboxylate)]·4H₂O·DMF), which incorporates the ligands naphthalene-2,6-dicarboxylate and L: methanetetra(tetrakis[4-(carboxyphenyl)oxymethyl]methane into an MTV-MOF structure, achieves a quantum yield of 11%³⁷. State-of-the-art materials like InGaN exhibit quantum yields around 70%³⁸. For MOFs to be competitive, they must achieve similar levels of efficiency. Heterogeneous MOFs, such as MTV-MOFs, have significant progress to make before they can be integrated into device structures. A key limitation of the MTV-MOF approach is the energy transfer between closely situated luminescent ligands that can quench the emission of one or more lumiphores^39,40. Consequently, new methods of generating heterogeneous MOFs that preserve the excellent color modulation of MTV-MOFs while enhancing efficiency are essential. Recently, there has been increasing interest in a novel approach known as MOF-on-MOF growth, which aims to synergistically combine the properties of multiple MOFs. In MOF-on-MOF growth, a “core MOF” is initially synthesized and then used as a seed for the growth of a secondary “shell MOF” on its surface. The primary advantage of this approach is that the “shell” and “core” MOFs interact only at their interface. Since energy transfer is distance-dependent, the MOF-on-MOF approach can reduce quantum yield loss by increasing the distance between fluorophores while still maintaining the benefits of having multiple fluorophores linked together^41–44. In MTV-MOFs, ligands are densely packed and randomly distributed within the unit cell. In contrast, MOF-on-MOF structures feature ligands that are spatially separated by a grain boundary formed during the epitaxial growth of the shell MOF. Previous studies examining the electronic and optical properties of MOF-on-MOFs have explored their application in photon upconversion, catalysis, and structural characterization^{39,40,42,45–47}. In 2021, Haldar et al. exploited energy transfer between layers to generate an energy cascade resulting in red emission, facilitated by doping organic acceptors between the layers³⁹. In these layered materials, the energy transfer efficiency reached 75%, resulting in an overall material having a quantum yield of 10.5%. However, enhancing the efficiency of the transfer steps alone may not suffice. Recently, Lee et al. conducted a study on the energy transfer mechanisms of an MTV-MOF, varying the integration levels of nanographene moieties⁴⁸. The MOF, known as UMOF-2 [UMOF-2: Unconventional MOF-2 composed of Cu and hexaphenyethyllbenzene (HPB) or hexabenzocoronene (HBC)], was incorporated with varying percentages of HBC. Increasing the HBC content not only changed the material’s color from blue to green but also resulted in an enhanced efficiency of energy transfer, which reduced the overall emission intensity as HPB transferred energy to HBC. In other words, despite improving energy transfer, the overall quantum yield was decreased. Energy transfer thus acts as a double-edged sword, offering attractive capabilities to modify the optical properties of MOFs for various applications—from color modulation to turn-off sensing—but also imposing limitations on material efficiency. Although numerous materials discussed thus far place emitters in close proximity to facilitate energy transfer, the complete implications of this arrangement have not been extensively studied.

In this work, we generated MTV-MOFs and MOF-on-MOFs to examine how the spatial arrangements of heterogeneous components within a MOF structure—whether closely integrated in MTV-MOFs or more distantly separated in MOF-on-MOFs—affect their optical properties (Fig. 1). In the case of MOF-on-MOFs, our approach involves growing a shell around the core by submerging the core MOF in a solution of shell MOF precursors and structure directing surfactants and subsequently conducting solvothermal treatment. This method facilitates the formation of a shell MOF that envelops the surface of the core, akin to assembling two MOFs together like Lego bricks. It also allows the generation of multiple shell layers in all directions. UiO-67—UiO: University of Oslo; Zr₆O₄(OH)₄(BPDC)₆, where BPDC^2-: 4,4’-Biphenyldicarboxylate linker—is an extensively studied luminescent MOF with a low quantum yield value of 3.8% ± 25% and high thermal stability attributed to its zirconium-carboxylate bonds served as the core building block for our MOF-on-MOF materials^49–51. For growth, we chose two MOFs with a cubic structure similar to UiO-67: Zr-AzoBDC made of Zr(IV) and 4,4’-Azobenzenedicarboxylate linkers (AzoBDC^2-), and Zr-StilBDC composed of Zr(IV) and 4,4’-Stilbenedicarboxylate linkers (StilBDC^2-). Zr-AzoBDC was selected for the presence of the azo bond (N=N bond), aiming to reduce the band gap of the MOF heterostructure and potentially induce red emission⁵². Similarly, the C=C double bond of stilbene was introduced to increase the number of π-bonds in the MOF orbitals, altering the energy compared to UiO-67. We generated two MOF-on-MOF heterostructures: MOF-on-MOF-BA, and MOF-on-MOF-BS, where the naming scheme denotes the initial letters of the ligands used in the core and shell, respectively. These MOF-on-MOF heterostructures were studied alongside two MTV-MOF analogs composed of the AzoBDC^2-, StilBDC^2-, and BPDC^2- linkers, with ratios matching closely those of the MOF-on-MOF counterparts. These were labeled as MTV-BA and MTV-BS to maintain consistency across the experiments. By investigating both MOF-on-MOF and MTV-MOF approaches, we aim to discern the differences and similarities between these two approaches and gain a better understanding of how chromophores interact with each other. We hypothesize that MTV-BA and MTV-BS will exhibit greater ligand interactions, as energy transfer primarily operates over short distances, whereas MOF-on-MOF-BA and MOF-on-MOF-BS would offer fewer opportunities for such interactions.

Fig. 1. Schematic representation of the materials synthesized in this project.

Three distinct carboxylate ligands were coordinated with Zr(IV) to generate UiO-67 (B; blue), Zr-AzoBDC (A; red), and Zr-StilBDC (S; magenta), each distinguished by its characteristic tag. The introduction of ligand A or S into the UiO-67 framework yields MTV-MOFs with arbitrary ligand distributions. Alternatively, synthesizing a MOF in the presence of a pre-formed MOF results in hierarchical MOF-on-MOF structures, where Zr-AzoBDC (red) or Zr-StilBDC (magenta) selectively assembles atop the UiO-67 (blue) core. It is important to note that these simplified structural representations serve as conceptual models to illustrate the formation of heterogeneous MTV-MOF and MOF-on-MOF architectures. They do not necessarily reflect experimentally observed factors such as precise MOF ratios, lattice mismatches, or ligand distribution patterns.

Results

Synthesis and characterization of UiO-67, Zr-AzoBDC, and Zr-StilBDC

The synthesis of the UiO-67 core was conducted based on previous work⁵³. UiO-67 is a Zr(IV)-based framework composed of BPDC^2- linkers, crystallizing in a Fm-3m space group to form a cubic unit cell with an edge length of 26.7734 Å^54–56. Known for its high thermal stability due to the strength of the Zr—O bonds, UiO-67 is an ideal candidate for seeding the growth of other MOFs onto its surface^{41,42,47,53,57}. Two MOFs, Zr-AzoBDC and Zr-StilBDC, were selected for growth on the UiO-67 core due to their similar coordination properties. Zr-AzoBDC crystallizes in a F23 (196) space group with a cubic unit cell and an edge length of 29.8623 Å, whereas Zr-StilBDC forms a cubic crystal with a Pn-3 (201) space group and an edge length of 30.0322 Å.

These three materials—UiO-67, Zr-StilBDC, Zr-AzoBDC—serve as the building blocks for the MTV-MOF and MOF-on-MOF approach to luminescent modulation. Powder X-ray diffraction (PXRD) patterns for all three MOFs were obtained and showed good agreement with their corresponding simulated patterns (Fig. 2a). Fourier Transform Infrared Resonance (FT-IR) spectra confirmed the formation of the MOFs, as evidenced by the disappearance of the stretching band associated with protonated carboxylic acid functional groups in the MOF spectra (Fig. S1). Thermogravimetric analysis (TGA) demonstrated high thermal stability of all three MOFs: UiO-67 is stable up to 500 °C, while both the Zr-StilBDC and Zr-AzoBDC are stable up to 400 °C (Figs. S2,  S3, Fig. S4). Additionally, TGA conducted in air for UiO-67 quantified the residual ZrO₂ content in the sample. After accounting for solvent loss, the final mass percent was determined to be 38%, which is slightly higher than the expected 34.93%^58,59, suggesting the presence of missing linker defects. Similarly, the TGA profiles for Zr-StilBDC and Zr-AzoBDC yielded final mass percentages of 43% and 34%, respectively. The higher residual weight in Zr-StilBDC could suggest the presence of a hexagonal close-packed (hcp) phase; however, PXRD analysis did not reveal an additional peak in the low 2θ region, indicating that this phase, is not present⁵⁸. Furthermore, the TGA profile of Zr-StilBDC reveals a pronounced solvent loss, consistent with literature reports⁶⁰. The observed weight loss occurs at a temperature corresponding to the boiling point of the solvent system, confirming that it originates from the solvent removal rather than ligand decomposition. Nitrogen (N₂) adsorption isotherms collected for the three MOF building blocks confirmed that Zr-StilBDC and Zr-AzoBDC are microporous, displaying characteristic Type I isotherms (Figs. 2b,  S5), while UiO-67 exhibited a step at 0.1 P/P₀, indicative of the presence of two distinct pore types, consistent with previous literature reports⁵³. UiO-67 exhibited a BET surface area of 2,224 ± 2 m²/g, whereas Zr-AzoBDC and Zr-StilBDC displayed surface areas of 1,166 m²/g and 1,134 m²/g, respectively. While the BET surface area of Zr-AzoBDC aligns well with literature reports⁶¹, Zr-StilBDC displays lower surface area⁶⁰. However, its purity was confirmed through PXRD, TGA, and EA analyses, with the latter showing excellent agreement with the theoretical stoichiometry (Table S2)^53,60,61. The relatively low BET surface area observed for Zr-StilBDC may be attributed to morphological factors. Scanning Electron Microscopy (SEM) images confirmed that all UiO-67, Zr-StilBDC, and Zr-AzoBDC exhibit uniform crystal sizes and high homogeneity, suggesting that structural features or textural effects could impact their surface area (Figs. S6– S8). Building upon the optimized synthetic protocols for UiO-67, Zr-AzoBDC, and Zr-StilBDC, MTV-MOFs and MOF-on-MOF structures were successfully synthesized.

Fig. 2. Solid state characterization of MOFs and MOF heterostructures.

a PXRD patterns for (bottom to top) UiO-67 simulated, Zr-StilBDC simulated, UiO-67, Zr-StilBDC, MTV-BS, and MOF-on-MOF-BS. b PXRD patterns (bottom to top) for UiO-67 simulated, Zr-AzoBDC simulated, UiO-67, Zr-AzoBDC, MTV-BA, and MOF-on-MOF-BA and c. Nitrogen isotherms collected at 77 K for MOF-on-MOF-BS (green), MTV-BS (magenda), Zr-StilBDC (red), and UiO-67 (black).

Generation of MTV-MOFs and MOF-on-MOF Heterostructures

For the generation of the MTV-MOFs, linkers were carefully added in a 2:1 ratio (BPDC^2-:AzoBDC^2- or BPDC^2-:StilBDC^2-) (Fig. S9 and  S10). MTV-BS and MTV-BA were obtained with 86.7% and 84.7% yields, respectively. MTV-BS exhibits significant peak broadening in the low-angle region of the PXRD pattern (2θ = 5.17°). The lowest-angle peak, corresponding to the (111) reflection, remains prominent as in both UiO-67 and Zr-StilBDC. However, the (111) reflection in MTV-BS appears to have broadened to the extent that it is unresolved from the (200) peak at approximately 5.97°. This observation suggests potential layering or congregation of ligands within the MTV structure (Fig. 2a). Based on the position of the (111) peak, we calculated a unit cell parameter of 29.62 Å, which is slightly smaller than the expected 30.03 Å for Zr-StilBDC due to the inclusion of shorter BPDC^2- linkers. Similarly, the PXRD pattern for MTV-BA was analyzed using the same methodology. The (111) reflection generates the lowest-angle peak, located at 4.97°, corresponding to a calculated unit cell parameter of 30.80 Å. Like MTV-BS, MTV-BA also exhibits broadening of the (111) peak, making resolution between it and the (200) peak challenging. FT-IR further confirmed the generation of the MTV-MOFs (Fig. S1), with MTV-BS showing significant broadening of the 1654 cm⁻¹ and 1597 cm⁻¹ stretches due to the overlap of BPDC and StilBDC carbonyl carbon C=O vibrations in the MTV-MOF heterostructure (Fig. S1). MTV-BA also exhibits broadening in the 1602 cm⁻¹ carbonyl carbon C=O stretch (Fig. S1) Nitrogen isotherms of the MTV-BA and MTV-BS revealed BET surface areas of 912 and 752 m²/g (Fig. S11, Fig. 2b), confirming that the MTV-MOFs maintain their porosity. The isotherms exhibited a Type I profile, indicative of microporosity. Additionally, SEM images confirmed the uniform phase distribution of both MTV-MOFs (Fig. S13 and  S14). The characterization of MTV-MOFs confirms that the BPDC^2- linker was successfully incorporated into MOFs containing AzoBDC^2- or StilBDC^2- linkers. Based on the PXRD, FT-IR, BET surface areas, and SEM, it is apparent that the mismatch between ligands has not interfered with the formation of the MTV-MOFs. Additionally, TGA was performed on both MTV MOFs. MTV-BS exhibits stability up to 400 °C, and based on the decomposition data, we observed a residual 25% ZrO₂, which is lower than the observed residual mass for UiO-67 and Zr-StilBDC. This discrepancy suggests an excess of the ligand in the structure (Fig. S15). MTV-BA was analyzed in a comparable manner, showing stability up to 400 °C (Fig. S16) and a ZrO₂ residue of 45%. These TGA results indicate solvent losses consistent with those of the pristine MOFs, confirming that the various samples are comparable. To determine the ligand ratios in the MTV-MOFs, they were digested in concentrated NaOD, and the resulting solutions were analyzed using ¹H-NMR (Fig. S9). Analysis of the ¹H-NMR of these solutions is challenging due to pH fluctuations in the basic digestion medium, which induce significant chemical shifts between the free ligands and their post-digestion counterparts⁶². First, we obtained NMR spectra of the individual ligands to establish their relative chemical shifts: BPDC^2- and AzoBDC^2- linkers exhibit two doublets, while StilBDC has two doublets and a singlet. In the NMR spectra of MTV-BA, for example, there are four peaks in the aromatic region, with four of them being doublets. Based on the NMR spectra after MTV-MOF digestion, we identified the order of peaks in the aromatic region as Proton D (AzoBDC^2-), Proton B (BPDC^2-), Proton C (AzoBDC^2-), and finally proton A (BPDC^2-). To confirm our peak assignments, we sequentially spiked the NMR samples with BPDC^2- and AzoBDC^2- (Fig. S10). MTV-BA follows this D-B-C-A pattern, with B and A having a ratio of 1:1 ratio and D and C also having a 1:1 ratio. The B:D ratio was found to be 2:1, therefore, the core:shell ratio was 2:1 (Table S1). For MTV-BS, the expected order of peaks is B-E-A-F-G, with E, F, and G corresponding to the protons on the StilBDC^2- linker (Fig. S9). The ratio of B:E gave us a core:shell ratio of 1.8:1, which is close to 2:1 as in MTV-BA (Table S1). To further validate the results obtained from ¹H-NMR, we performed EA on the MTV-MOFs, which yielded compositional ratios consistent with the NMR findings (Table S2). Specifically, MTV-BS exhibited a BPDC^2-:StilBDC^2- ratio of 1.6, while MTV-BA showed a ratio of 1.8.

The MOF-on-MOF heterostructures were synthesized using the procedures outlined in the methods section. MOF-on-MOF-BS was isolated with a 94.5% yield, while MOF-on-MOF-BA was obtained with a 72.2% yield. The PXRD pattern for MOF-on-MOF-BS revealed a broad peak at 4.92° (Fig. 2a), corresponding to a single phase. A comparison of this pattern with the simulated patterns for UiO-67 and Zr-StilBDC suggests that the broad peak likely represents the (111) reflection, albeit shifted to a lower angle (5.90° for UiO-67). The observed d-spacing of 17.9 Å is larger than the 15.0 Å value for the same peak in UiO-67, corresponding to a unit cell parameter of 31.0 Å. This indicates that the MOF-on-MOF-BS has a larger unit cell compared to that of Zr-StilBDC. Moreover, the broadening of the peak in the MOF-on-MOF-BS sample, as evidenced by the increase in Full Width at Half Maximum (FWHM) from 0.49° for UiO-67 to 1.15° for MOF-on-MOF-BS, highlights a reduction in long-range order while maintaining short-range order. This behavior is consistent with the expected structure of a MOF-on-MOF nanosized core-shell material. The PXRD pattern for MOF-on-MOF-BA closely resembles that of MOF-on-MOF-BS but exhibits an even greater broadening of the (111) reflection. The peak position at 5.21° corresponds to a unit cell parameter of 30.8 Å, which is larger than the expected 29.86 Å unit cell of the Zr-AzoBDC shell MOF. The FT-IR spectrum of the MOF-on-MOF-BS exhibited stretches of similar broadening as seen in MTV-BS, although to a lesser extent (Fig. S1). The reduced broadening may be due to the higher concentration of Zr-StilBDC on the surface of the material. The stretch at 656 cm⁻¹ showed possible overlap with a UiO-67, and new stretches, not present in Zr-StilBDC, appeared at 433 cm⁻¹ corresponding to changes in the fingerprint region. Changes in the vibrations within the fingerprint region are challenging to assign to specific functional groups, however, they can be attributed to changes in the overall molecular composition⁶³. In the case of MOF-on-MOF-BA, the carbonyl carbon C=O vibrations at 1678 cm⁻¹ were of lower intensity and displayed several overlapping and broad stretches. New stretches also appeared in the 750 cm⁻¹ region, consistent with UiO-67, which are absent in MTV-BA. The ¹H-NMR spectrum of the digested MOF-on-MOF-BS exhibits distinct peaks corresponding to both the ligands present in UiO-67 and Zr-StilBDC, confirming the successful incorporation of both ligands. Like the ¹H-NMR results observed for the MTV-MOFs, a notable shift in chemical shift (ppm) is present due to the altered pH conditions resulting from NaOD used in the digestion process. By analyzing peak assignments, we determined a 2:1 integration ratio between the BPDC^2- and StilBPDC^2- linkers. Likewise, for MOF-on-MOF-BA, a 2:1 ratio was observed between the BPDC^2- and AzoBDC^2- linkers (Fig. S9), further supporting the structural integrity and composition of these heterostructures. Elemental analysis confirmed the findings from NMR studies, yielding ratios consistent with expected values (Table S2). Specifically, MOF-on-MOF-BS exhibited a UiO-67 to Zr-StilBDC ratio of 1.9, while MOF-on-MOF-BA displayed a UiO-67 to Zr-AzoBDC ratio of 2.1. Thermogravimetric analysis was conducted to evaluate the thermal stability and compare the residual mass of MOF-on-MOFs with theoretical expectations. Both MOF-on-MOF-BS and MOF-on-MOF-BA demonstrated thermal stability up to 400 °C (Fig. S17 and S18). MOF-on-MOF-BS and MOF-on-MOF-BA exhibited residual masses of 45% and 47%, respectively—both exceeding their expected values of 34.15% and 34.03% respectively—suggesting the presence of structural defects, likely due to missing linkers. The BET surface area analysis confirmed that the MOF-on-MOF-BS exhibits a surface area of 710 m²/g (Fig. 2b). The N₂ isotherm displayed a type IV profile exhibiting a hysteresis loop, likely due to the nano-sized particles of the sample. It is well established that hierarchical structures often exhibit reduced surface areas due to constricted mesoporous openings⁶⁴. The N₂ isotherm of MOF-on-MOF-BA revealed a BET surface area of 1106 m²/g, with a similar isotherm shape and hysteresis loop as MOF-on-MOF-BS (Fig. S12). The SEM images of MOF-on-MOF-BA (Fig. S19) depict uniform particles consistent with the morphology of UiO-67, whereas MOF-on-MOF-BS (Fig. S20) exhibits smaller yet uniformly sized particles. Due to their nanometer particle size, high-resolution SEM was insufficient for structural differentiation; therefore, Transmission Electron Microscopy (TEM) was conducted for both MOF-on-MOF-BS (Fig. 3a) and MOF-on-MOF-BA (Fig. 3b). Despite the moderate resolution of TEM, the observed contrast differences in MOF-on-MOF-BA clearly distinguish core and shell regions, where the nitrogen-containing shell (due to the N=N functionality) provides sufficient contrast against UiO-67, making the core distinctly visible. Furthermore, electron diffraction patterns confirm crystallinity and short-range order, supporting our structural model. The total particle size of MOF-on-MOF-BA was measured at 97.40 nm, with an 80.73 nm core diameter, indicating that the shell extends 16.67 nm beyond the core. In contrast, distinguishing the core-shell structure in MOF-on-MOF-BS proved more challenging due to the compositional and structural similarities between Zr-StilBDC and UiO-67 (Fig. 3a). However, subtle density variations across the particle suggest the presence of a core-shell structure, further supported by electron diffraction patterns indicating a coherent crystalline framework (Fig. S21). MOF-on-MOF-BS exhibited an overall particle diameter of 81.77 nm, with a core size of 76.56 nm and a shell thickness of 5.21 nm. After characterizing the properties of these heterogeneous MOF structures, we proceeded to investigate their optical properties, focusing on how variations in ligand ratios and spatial arrangements influence their quantum yield. The observed nanometer-scale particles further support the broad peaks seen in the PXRD patterns—nano crystallite sizes are well known to induce peak broadening⁶⁵. Additionally, the formation of a shell phase in MOF-on-MOF heterostructures might disrupt long-range order, further contributing to the PXRD peak broadening. While defects in the lattice could also lead to peak broadening, EA and NMR data rule out significant structural defects, given the close agreement in ligand composition across all characterization techniques.

Fig. 3. Microscopy images.

TEM images provide evidence of the core-shell morphology in the synthesized MOF-on-MOF structures. a TEM images of MOF-on-MOF-BS reveals a well-defined core-shell architecture, highlighting the successful assembly of the fluorescent shell ligand onto the UiO-67 core. b Similarly, TEM images of MOF-on-MOF-BA confirm the formation of a distinct core-shell structure, demonstrating the controlled growth of Zr-AzoBDC on UiO-67.

Optical properties

To understand the interactions between ligands within the MTV-MOF and the MOF-on-MOF heterostructures and their impact on optical properties, we collected their UV-Vis reflectance and fluorescence spectra. We began by investigating the azo-based family of materials. As the core material, UiO-67 exhibits a reflectance peak at 318 nm (Fig. 4a), originating from the π-π* transitions of the H₂BPDC ligand. In contrast, Zr-AzoBDC shows absorption at 320 nm, 277 nm, and 224 nm, also attributed to the ligand, along with a unique absorption at 470 nm. MTV-BA exhibits the same absorption as Zr-AzoBDC, with the absorption at 224 nm blue shifting to 213 nm. In contrast, MOF-on-MOF-BA shows a red shift of the 224 nm peak to 303 nm. This suggests that the MOF-on-MOF growth method can slightly relax the high-energy transition. Using the UV-Vis spectra, Tauc plots were generated following the methodology outlined by Makula et al., where K-M absorption spectra were utilized (Fig. S22)⁶⁶. A linear extrapolation was performed to determine the band gap value at the x-axis. UiO-67 exhibited a band gap of 3.56 eV, whereas Zr-AzoBDC, MTV-BA, and MOF-on-MOF-BA showed band gaps of 1.99 eV, 1.97 eV, and 2.04 eV, respectively. The minimal variation in band gap among the three materials is likely attributed to consistent ratio between AzoBDC^2- and BPDC^2- linkers in both the MOF-on-MOF and MTV-MOF samples. Both MOF heterostructures display lower band gaps compared to UiO-67, which is particularly noteworthy in the case of MOF-on-MOF-BA, where none of the color properties typical of UiO-67 are evident. Another consequence of the similar electronic structures of the azo-based materials is the absence of detectable fluorescent emission in any of the samples. Notably, we observed no emission, indicating that measurements of quantum yield and lifetimes were not feasible even MOF-on-MOF-BA, despite its inclusion of the UiO-67 core.

Fig. 4. Absorption of MOFs and MOF heterostructures.

a UV-Vis spectra of UiO-67 (black), Zr-AzoBDC (blue), MTV-BA (orange), and MOF-on-MOF-BA (purple), and b UV-Vis spectra of UiO-67 (black), Zr-StilBDC (red), MTV-BS (magenta), and MOF-on-MOF-BS (green).

For the stilbene-family of materials, we observed a nuanced difference. Zr-StilBDC, MTV-BS, and MOF-on-MOF-BS all exhibited strong absorption at 374 nm, and 370 nm, respectively (Fig. 4b). Additionally, their spectra revealed a secondary absorption centered at 314 nm. The longer wavelength absorption peak is likely attributed from n-π* (low energy) transitions, whereas the peak at 314 nm is indicative of π-π* transitions. From the Tauc Plots (Fig. S22), we observe a slight variation in their band gap values. Zr-StilBDC exhibits a band gap of 2.95 eV, whereas MTV-BS and MOF-on-MOF-BS show band gaps of 2.73 eV and 3.01 eV, respectively. This indicates that the band gap of the MOF-on-MOF-BS aligns more closely with Zr-StilBDC than with MTV-BS. Interestingly, despite these differences in band gaps, there was no apparent change in the fluorescence spectra between MTV-BS and MOF-on-MOF-BS. The emission spectra for UiO-67, Zr-StilBDC, MTV-BS and MOF-on-MOF-BS are shown in Fig. 5a. Although solvent molecules confined within the pores could potentially influence emission, TGA confirms that all samples exhibit similar solvent losses. Moreover, since all materials were synthesized in DMF and subjected to identical synthesis and washing procedures, any residual solvent molecules are unlikely to impact the comparative analysis of these materials. Further, we immersed MOF-on-MOF-BS and MTV-BS in water and in ethanol and monitored their fluorescence spectra. Notably, the water suspension exhibited a more pronounced red shift compared to the ethanol suspension (Fig. S23), indicating a solvent-dependent modulation of the excited-state environment. In contrast, the solid-state materials displayed no observable shift in emission, supporting the conclusion that the encapsulated guest molecules are chemically identical across both heterostructures. All materials were excited at a wavelength of 390 nm to ensure uniform comparisons between the materials. UiO-67 emits at 460 nm (2.69 eV), which is slightly lower in energy than the emission peak of Zr-StilBDC at 454 nm (2.73 eV). For MTV-BS, the λ_max is nearly identical at 455 nm (2.72 eV), while MOF-on-MOF-BS exhibits an emission that nearly overlaps with MTV-BS despite their different band gap values. Interestingly, the intensity of emission nearly doubles in both MTV-BS and MOF-on-MOF-BS compared to Zr-StilBDC. Due to a combination of emission broadness, intensity, and wavelength shift, the CIE coordinates indicate a noticeable color change from UiO-67 and Zr-StilBDC to MTV-BS and MOF-on-MOF-BS. The color coordinates for UiO-67 are (0.174, 0.192), which shift to (0.156, 0.166) for MTV-BS and (0.156, 0.167) for MOF-on-MOF-BS (Fig. 5b). This significant shift aligns with the color values observed for Zr-StilBDC (0.157, 0.170). Overall, this demonstrates that the electronic structures of the MTV-BS and MOF-on-MOF-BS are like each other and resemble that of Zr-StilBDC. Given that both MTV-BS and MOF-on-MOF-BS contain a highly fluorescent ligand in their structures, it is expected that the emission profiles of these MOF heterostructures reflect to the characteristics of the stilbene ligand. However, as previously observed, there is a noticeable difference in emission intensity. Zr-StilBDC exhibits significantly lower emission intensity compared to either MOF-on-MOF-BS or MTV-BS. To further investigate this, we compared the emission profiles of Zr-StilBDC, MTV-BS, and MOF-on-MOF-BS with those of the pure ligands. Both K-M and fluorescence spectra confirmed that the MOFs exhibited ligand centered emission with no fluorescent contribution from the metal (Fig. S24,  S25). To gain more insights into this observation, we examined both the lifetime and quantum yields of these MOFs. MTV-BS shows a shorter lifetime of 4.05 ns compared to MOF-on-MOF-BS at 5.74 ns (Fig. S26). Given that UiO-67 exhibits a fluorescence decay lifetime of 18.95 ns, we estimated the efficiency of the energy transfer using the equation: $E=1-\frac{{\tau }_{DA}}{{\tau }_D}$ where E represents the transfer efficiency, τ_DA represents the excited-state lifetime of the donor-acceptor pair within the heterogenous MOFs, and τ_D represents the lifetime of UiO-67 (donor). Applying this relationship, we calculated E = 0.79 for MTV-BS and E = 0.70 for MOF-on-MOF-BS. These findings indicate a higher energy transfer efficiency MTV-BS, consistent with greater quenching and reduced overall emission. Furthermore, this trend aligns with the spectral overlap data (Fig. S27), which shows a more pronounced overlap between the absorption spectrum of MTV-BS and the emission spectrum of UiO-67, further supporting more efficient energy transfer in the MTV structure. The shorter lifetime of MTV-BS could be attributed to the closer proximity of the ligands within the structure, potentially facilitating more efficient energy transfer and consequently shorter emission lifetimes. Moreover, in line with our observation on the emission intensity (Fig. 5a), both MTV-BS and MOF-on-MOF-BS demonstrate higher quantum yields compared to UiO-67 or Zr-StilBDC. UiO-67 exhibits the lowest quantum yield at 3.0% ± 25% (Table 1). Previous literature studies report comparable quantum yields for UiO-67, encouraging the validity of the data^49,67. Zr-StilBDC exhibited a higher quantum yield (8.2% ± 25%) compared to UiO-67 (Table 1). Interestingly, MTV-BS showed a higher quantum yield than both UiO-67 and Zr-StilBDC with 10.2% ± 25%. Notably, MOF-on-MOF-BS outperformed both materials, achieving a quantum yield of 40.0% ± 25% (Table S2, Table 1). The MOF-on-MOF strategy is particularly compelling, as it employs the same metal nodes and linkers as MTV-BS yet achieves a significantly higher quantum yield. This enhancement suggests that the heterostructure architecture plays a critical role in optimizing photophysical properties beyond what is attainable through a homogeneous mixed-ligand approach. As a control, we prepared a physical mixture of UiO-67 and Zr-StilBDC in a 2:1 ratio, ensuring thorough blending. This mixture exhibited a significantly lower quantum yield of 5.5%, likely due to inefficient orbital overlap between UiO-67 and Zr-StilBDC, leading to the absorption of emitted light by adjacent UiO-67 particles (Table 1). Notably, the quantum yield of this physical mixture remains lower than that of MTV-BS, even when considering the ±25% uncertainty, and is substantially below that of MOF-on-MOF-BS. These results highlight the advantages of integrated heterostructures, demonstrating that simple physical mixing fails to achieve the enhanced optical properties observed in MOF-on-MOF architectures. We hypothesize that the MTV-BS and MOF-on-MOF-BS samples utilize the structural separation between the UiO-67 (or BPDC^2-) and ZrStilBDC (or StilBDC^2-) to align the chromophores to prevent random reabsorption. Overall, the low quantum yield of the physical mixture underscores the significant advantage of incorporating these ligands into either MTV-MOF or MOF-on-MOF structures. In systems involving energy transfer, each step of energy transfer typically diminishes the overall yield, with literature reporting transfer efficiencies ranging from 50% to 70%^39,68. While our transfer efficiencies are close to these reports, our heterogeneous MOFs demonstrate significantly higher quantum yields than their base materials, with the MOF-on-MOF heterostructure featuring larger separations between chromophores exhibiting the highest quantum yield. Using both quantum yield and lifetime data, we observed that the rate of radiative decay increases from 2.518 ns⁻¹ to 6.98 ns⁻¹ from MTV-BS to MOF-on-MOF-BS. In the MTV-BS approach, the proximity of the linkers enhances the potential of energy transfer. In contrast, the MOF-on-MOF approach separates the linkers, restricting energy transfer primarily to the interface of the two crystals due to the distance limitation in energy transfer. Despite similar emission and absorption spectra, indicating comparable color modulations resulting from interactions between the two ligands in both MOF-on-MOF and MTV-MOFs, significant differences are observed in their lifetimes and quantum yields.

Fig. 5. Fluorescent properties of MOFs and MOF heterostructures.

a Fluorescence emission of UiO-67 (black), Zr-StilBDC (red), MTV-BS (magenta), MOF-on-MOF-BS (green), and the physical mixture (dark cyan) and b CIE color coordinates for each MOF structure and heterostructure.

Table 1.

Quantum yields, CIE coordinates, and lifetimes for the fluorescent MOF structures and heterostructures collected in the solid state

Sample	Quantum Yield (%)	CIE Coordinates	Lifetimes (ns)
UiO-67	3.0	(0.174, 0.192)	18.95
Zr-StilBDC	8.2	(1.57, 0.170)	N/A
MTV-BS	10.2	(0.156, 0.166)	4.05
MOF-on-MOF-BS	40.0	(0.156, 0.167)	5.74
Physical Mixture	5.5	(0.158, 0.165)	N/A
Sodium salicylate	Experimental quantum yield: 62.0 Theoretical quantum yield: 62.0 – ref. 67

Discussion

This study investigates the transfer of energy between chromophores within MTV-MOF and MOF-on-MOF heterostructures. By examining the optical properties, lifetimes, and quantum yields of analogous MTV-MOFs and MOF-on-MOFs, we observed that MOF-on-MOF heterostructures exhibit greater emission efficiency. We hypothesize that this difference arises from the spatial arrangement of chromophores, minimizing energy transfer mechanisms—however further computational studies are needed to confirm this mechanism. The similar structures of H₂BPDC, H₂StilBDC, and H₂AzoBDC suggest that their carboxylic acid groups vary only slightly in acidity, affecting their interactions within the framework. Further, Zr-AzoBDC and Zr-StilBDC (29.86 Å and 30.03 Å) are longer than UiO-67 due to the increased length of the H₂AzoBDC and H₂StilBDC ligands, meaning that AzoBDC^2- and StilBDC^2- will not fit perfectly into the heterogeneous structure. Predicting preferential bonding in MTV-MOFs is challenging due to kinetic factors, which necessitate additional thermodynamic computations. However, the PXRD patterns for MTV-BA and MTV-BS indicate that both ligands are integrated within a single crystal, suggesting homogeneous distribution. In contrast, MOF-on-MOF synthesis involves the epitaxial growth of a secondary MOF over a core MOF, leading to greater ligand separation while preserving interfacial interactions at the core-shell boundary. This structural arrangement results in a single broad peak in the PXRD pattern. The addition of CTAB as a structure-directing agent facilitated the formation of MOF-on-MOF samples with uniform particle morphology, as confirmed by SEM, while TEM analysis further supported the presence of the core-shell architecture. One of our experimental objectives was to achieve crystallization of MTV-MOFs and MOF-on-MOFs with the same ligand ratio, ensuring that any observed differences in emission, quantum yield, and lifetime could be attributed to ligand orientation rather than composition. NMR analysis was employed to determine the incorporation ratio of the linkers, revealing a 2:1 core-to-shell ratio. In all cases, EA experiments confirmed the ligand ratios seen in the NMR studies. The dominant presence of the core likely accounts for the relatively small shell dimensions observed in TEM, measuring 5.21 nm for MOF-on-MOF-BS and 16.67 nm for MOF-on-MOF-BA. Both MTV-BS and MOF-on-MOF-BS exhibit identical emission profiles with CIE color coordinates of (0.156, 0.167). However, their lifetimes and quantum yields differ significantly: MTV-BS has a quantum yield of 10.2% and a lifetime of 4.05 ns, whereas the MOF-on-MOF-BS achieves a higher quantum yield of 40.0% and a lifetime of 5.74 ns. For comparison, a physical mixture of the Zr-StilBDC and UiO-67 showed a quantum yield of only 3.0% highlighting the superior performance of heterogeneous MOFs. MOF-on-MOF-BS achieves the same emission profile as MTV-BS but with greater emission efficiency, likely due to enhanced surface interactions and increased ligand spacing, resulting in reduced energy transfer between ligands. The observed increase in both quantum yield and excited-state lifetime from MTV-MOF to MOF-on-MOF supports a reduction in energy transfer as the primary factor driving the enhanced photophysical performance. If increased quantum yield were solely due to higher radiative rates, we would expect a shorter lifetime, which is not observed. While structural rigidity could reduce non-radiative decay, the MTV-MOF exhibits similar trends without such constraints, reinforcing the role of suppressed energy transfer. Future computational studies could provide mechanistic insights, though the complexity of the MOF-on-MOF unit cell presents a significant computational challenge. In addition to energy transfer, aggregation-induced quenching is one of the common hurdles for solid-state emitters and results from adjacent fluorescent species absorbing the light emitted by each other. Our findings are consistent with existing literature, where MOF-on-MOF heterojunctions have demonstrated enhanced performance, particularly in the context of photocatalysis. Previous studies have reported similar improvements in efficiency when using MOF heterostructures, highlighting the beneficial role of the heterointerface in facilitating charge separation and enhancing catalytic activity^47,69,70.

In solution-based fluorescent systems, dilution can effectively mitigate quenching effects; however, achieving the same in solid-state materials presents a greater challenge. The MOF-on-MOF strategy addresses this issue by distributing the fluorescent shell ligand across the UiO-67 core, minimizing aggregation and enhancing emission efficiency. This approach provides a framework for improving the quantum yield of heterogeneous MOF structures compared to their single-ligand counterparts. The findings of this study offer insights into the design and optimization of MOFs for advanced lighting applications, paving the way for the development of high-performance luminescent materials.

Methods

Materials

All chemicals were purchased from Sigma Aldrich, Avantor, VWR, Millipore Sigma, and Tokyo Chemical Industry and used without further purification.

Synthesis of UiO-67, Zr-StilBDC, and Zr-AzoBDC

Synthesis of UiO-67 followed literature procedures⁵³. Briefly, 22.3 mg (0.0957 mmol) ZrCl₄ and 165 µL HCl were suspended in 5 mL DMF prior to the addition of 30.76 mg (0.127 mmol) of 4,4’-Biphenyldicarboxylic acid (H₂BPDC) in the reaction vial. Following that the reaction mixture was heated to 80 °C for 12 h to obtain the white colored solid, which was washed with DMF, followed by acetone to obtain the MOF. Synthesis of Zr-StilBDC followed literature procedures with few modifications. ZrCl₄, (30 mg), DMF (2 mL), formic acid (0.3 mL), and H₂O (10 µL) were mixed in a vial and heated to 120 °C till a clear solution was obtained. In a separate vial, 4,4’-Stilbenedicarboxylic acid (H₂StilBDC) (35 mg) and DMF (3 mL) were mixed and heated at 120 °C to solubilize the ligand completely. Following that, the two clear solutions were combined and heated to 150 °C for 72 h in an autoclave. After the reaction had cooled to room temperature, the solid was separated using a centrifuge and then stirred in 20 mL DMF at 80 °C for 12 h then DMF was removed and replaced three times. Finally, the solid was collected using a centrifuge and stirred in 20 mL acetone for 12 h at room temperature. Afterward, it was collected and dried at 80 °C to obtain the MOF. Synthesis of Zr-AzoBDC followed the same procedure as Zr-StilBDC but in absence of 10 µL H₂O and 4,4’-Azobenedicarboxylic acid (H₂AzoBDC) was used as a ligand.

Synthesis of MTV-BS and MTV-BA

Synthesis of MTV-BS followed literature procedures with few modifications. ZrCl₄, (30 mg), DMF (2 mL), formic acid (0.3 mL), and H₂O (10 µL) were mixed in a vial and heated to 120 °C till the clear solution was obtained. In a separate vial, 4,4’-Stillbenedicarboxylic acid (H₂StilBDC) (16 mg) and DMF (2 mL) were mixed and heated at 120 °C to solubilize the ligand completely. Similarly, in another vial 4,4’-Biphenyldicarboxylic acid (H₂BPDC) (10.25 mg) and DMF (1 mL) were mixed and heated at 120 °C to solubilize the ligand completely. Following that, all three clear solutions were combined in an autoclave (ZrCl₄ solution, then StilBDC^2-, then BPDC^2-) and heated to 150 °C for 72 h in an autoclave. The washing steps were like the Zr-StilBDC. Synthesis of MTV-BA followed the same procedure as MTV-BS but in the absence of 10 µL H₂O, and 4,4’-Azobenedicarboxylic acid (H₂AzoBDC) was used as a ligand.

Synthesis of MOF-on-MOF-BS and MOF-on-MOF-BA

Synthesis of MOF-on-MOF-BS proceeded by suspending 15 mg UiO-67 and 300 µL of 0.01 M CTAB solution in 3 mL DMF with 20 min sonication. Separately, ZrCl₄, (18 mg), DMF (1 mL), formic acid (0.5 mL), and H₂O (10 µL) were mixed in a vial and heated to 120 °C till a clear solution was obtained. Also, in a separate vial, 4,4’-Stillbenedicarboxylic acid (H₂StilBDC) (22 mg), 300 µL of 0.01 M CTAB, and DMF (2 mL) were mixed and heated at 120 °C to solubilize the ligand completely. Following that, all three components were combined (UiO-67 solution first, then ZrCl₄, then StilBDC^2-) in autoclave and heated to 150 °C for 72 h in an autoclave. The washing steps were like the Zr-StilBDC. Synthesis of MOF-on-MOF-BA followed the same procedure as MOF-on-MOF-BS, but in the absence of 10 µL H₂O and 4,4’-Azobenedicarboxylic acid (H₂AzoBDC) was used as a ligand.

Characterization methods

The powder X-ray diffraction patterns (PXRD) were recorded on a Rigaku MiniFlex XRD diffractometer using monochromated Cu Kα radiation (λ = 1.5418 Å) at ambient temperature. Our conditions of measurement were from 3 to 30 degrees, with a step of 0.02 degrees and a speed of 2.0°/min. Voltage was 40 kV and the current was 15 mA. Simulated powder X-ray diffraction patterns were generated from the single crystal data using Mercury 3.0. The thermogravimetric analysis (TGA) was carried out with a Shimadzu Thermogravimetric Analyzer (TGA) under argon atmosphere, at a heating rate of 10 °C/min up to 600 °C for all measurements. The UV-Vis absorbance spectra were obtained using a PerkinElmer UV-Vis Spectrometer. The diffuse reflectance was collected by depositing the powders within quartz slides and the Kubelka-Munk function was applied to the raw data to eliminate scattering. Fourier-transform infrared spectroscopy (FT-IR) was conducted on a PerkinElmer Spectrum Two FTIR Spectrometer from 400 to 4000 cm^–1. Nitrogen adsorption-desorption isotherms were collected at 77 K and 1 bar using the micromeritics 3FLEX surface analyzer. Before collecting data, the samples were degassed at 150 °C for 18 h. The BET surface areas were estimated from the amount of N₂ adsorbed via the BET (Brunnauer-Emmett-Teller) equilibrium equation.

Method of quantum yield calculation

Spectra collection

All emission and excitation spectra were acquired using a Horiba QuantaMaster 8075-21 fluorometer equipped with an integrating sphere. The spectra are collected with slit widths of 0.9 nm for emission and excitation. Spectra are all excited at the same wavelength of 390 nm, except for the sodium salicylate standard which was collected at 370 nm excitation. Sodium salicylate was used as a standard and has a known quantum yield of 62.0%⁷¹. Diffuse reflectance of the excitation scans through the range 380-400 nm, while emission scans through 410-600 nm. Reflectance measurements are measured at 450 nm by exciting the sample at 450 nm and measuring the diffuse reflectance from 445 to 465 nm. Samples were measured in the solid state. The reference material was Teflon. The quantum yield was corrected for the use of Neutral Density Filters in the collection of the diffuse reflectance of the excitation, and for the reflectance of the powder in the region of luminescence.

Calculation of un-corrected Quantum Yield

$$\mathrm{\varphi }=\frac{\mathrm{Photons}\mathrm{Emitted}}{\mathrm{Photons}\mathrm{Absorbed}}=\frac{\mathrm{Intensity}\mathrm{of}\mathrm{Emission}}{\left(\mathrm{DR}\mathrm{of}\mathrm{Reference}-\mathrm{DR}\mathrm{of}\mathrm{Sample}\right)}$$

Final Quantum Yield Calculation using Sodium Salicylate Standard

$${\mathrm{\varphi }}_{\mathrm{final}}={\mathrm{\varphi }}_{Standard}\times \frac{I_{Unknown}{A}_{Standard}}{I_{Standard}{A}_{Unknown}}\times 100\%$$

Collection of Lifetimes Using Time Correlated Single Photon Counting (TCSPC)

All lifetimes were collected in a Horiba Quanta Master 8075-21 system with a Delta Diode DD-300 (312 nm) laser and DSS LN-Series Receiver Module. The laser was pulsed at 5 MHz. Calculation of the mean lifetime followed the procedure outlined in the literature⁷².

Author contributions

K.C.S. conceived the project and designed the experiments together with K.T.S. K.T.S. led the experimental work (synthesis and characterization), performed UV-vis and emission experiments and interpreted the data. A.K.Y. synthesized the materials and investigated their surface areas. All authors contributed to the writing and editing of the manuscript.

Peer review

Peer review information

Nature Communications thanks Sergio Carrasco and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The data that supports the findings of the study are included in the main text and supplementary information files. Source Data are provided with this paper. Raw data can be obtained from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-62809-4.