Timothée Stassin

Comprehensive evaluation of geospatial foundation models (Geo-FMs) requires benchmarking across diverse tasks, sensors, and geographic regions. However, most existing benchmark datasets are limited to segmentation or classification tasks, and focus on specific geographic areas. To address this gap, we introduce a globally distributed dataset for forest aboveground biomass (AGB) estimation, a pixel-wise regression task. This benchmark dataset combines co-located hyperspectral imagery (HSI) from the Environmental Mapping and Analysis Program (EnMAP) satellite and predictions of AGB density estimates derived from the Global Ecosystem Dynamics Investigation lidars, covering seven continental regions. Our experimental results on this dataset demonstrate that the evaluated Geo-FMs can match or, in some cases, surpass the performance of a baseline U-Net, especially when fine-tuning the encoder. We also find that the performance difference between the U-Net and Geo-FMs depends on the dataset size for each region and highlight the importance of the token patch size in the Vision Transformer backbone for accurate predictions in pixel-wise regression tasks. By releasing this globally distributed hyperspectral benchmark dataset, we aim to facilitate the development and evaluation of Geo-FMs for HSI applications. Leveraging this dataset additionally enables research into geographic bias and generalization capacity of Geo-FMs.

Comprehensive evaluation of geospatial foundation models (Geo-FMs) requires benchmarking across diverse tasks, sensors, and geographic regions. However, most existing benchmark datasets are limited to segmentation or classification tasks, and focus on specific geographic areas. To address this gap, we introduce a globally distributed dataset for forest aboveground biomass (AGB) estimation, a pixel-wise regression task. This benchmark dataset combines co-located hyperspectral imagery (HSI) from the Environmental Mapping and Analysis Program (EnMAP) satellite and predictions of AGB density estimates derived from the Global Ecosystem Dynamics Investigation lidars, covering seven continental regions. Our experimental results on this dataset demonstrate that the evaluated Geo-FMs can match or, in some cases, surpass the performance of a baseline U-Net, especially when fine-tuning the encoder. We also find that the performance difference between the U-Net and Geo-FMs depends on the dataset size for each region and highlight the importance of the token patch size in the Vision Transformer backbone for accurate predictions in pixel-wise regression tasks. By releasing this globally distributed hyperspectral benchmark dataset, we aim to facilitate the development and evaluation of Geo-FMs for HSI applications. Leveraging this dataset additionally enables research into geographic bias and generalization capacity of Geo-FMs. The dataset and source code will be made publicly available.

Liquid- and vapour-phase conversion of aligned Cu(OH)₂ nanobelts into crystalline coatings of Cu-BDC and Cu-CDC MOFs, leads to differences in the crystal phase, morphology and crystallographic orientation.

Thin films of crystalline and porous metal–organic frameworks (MOFs) have great potential in membranes, sensors, and microelectronic chips. While the morphology and crystallinity of MOF films can be evaluated using widely available techniques, characterizing their pore size, pore volume, and specific surface area is challenging due to the low amount of material and substrate effects. Positron annihilation lifetime spectroscopy (PALS) is introduced as a powerful method to obtain pore size information and depth profiling in MOF films. The complementarity of this approach to established physisorption‐based methods such as quartz crystal microbalance (QCM) gravimetry, ellipsometric porosimetry (EP), and Kr physisorption (KrP) is illustrated. This comprehensive discussion on MOF thin film porosimetry is supported by experimental data for thin films of ZIF‐8.

Energy‐efficient indoors temperature and humidity control can be realised by using the reversible adsorption and desorption of water in porous materials. Stable microporous aluminium‐based metal–organic frameworks (MOFs) present promising water sorption properties for this goal. The development of synthesis routes that make use of available and affordable building blocks and avoid the use of organic solvents is crucial to advance this field. In this work, two scalable synthesis routes under mild reaction conditions were developed for aluminium‐based MOFs: (1) in aqueous solutions using a continuous‐flow reactor and (2) through the vapour‐assisted conversion of solid precursors. Fumaric acid, its methylated analogue mesaconic acid, as well as mixtures of the two were used as linkers to obtain polymorph materials with tuneable water sorption properties. The synthesis conditions determine the crystal structure and either the MIL‐53 or MIL‐68 type structure with square‐grid or kagome‐grid topology, respectively, is formed. Fine‐tuning resulted in new MOF materials thus far inaccessible through conventional synthesis routes. Furthermore, by varying the linker ratio, the water sorption properties can be continuously adjusted while retaining the sigmoidal isotherm shape advantageous for heat transformation and room climatisation applications.

MAF‐252, a non‐porous crystalline coordination polymer, is obtained from the solvent‐free reaction of ZnO with 3‐(2‐pyridyl)‐5‐(4‐pyridyl)‐1,2,4‐triazole. MAF‐252 can be synthesized in powder form and deposited as thin films, starting from ZnO powder and layers, respectively. Chemical vapor deposition (CVD) of MAF‐252 enables conformal and patterned thin films, even on high aspect ratio features.

The performance of modern chips is strongly related to the multi-layer interconnect structure that interfaces the semiconductor layer with the outside world. The resulting demand to continuously reduce the k-value of the dielectric in these interconnects creates multiple integration challenges and encourages the search for novel materials. Here we report a strategy for the integration of metal-organic frameworks (MOFs) as gap-filling low-k dielectrics in advanced on-chip interconnects. The method relies on the selective conversion of purpose-grown or native metal-oxide films on the metal interconnect lines into MOFs by exposure to organic linker vapor. The proposed strategy is validated for thin films of the zeolitic imidazolate frameworks ZIF-8 and ZIF-67, formed in 2-methylimidazole vapor from ALD ZnO and native CoO_x, respectively. Both materials show a Young’s modulus and dielectric constant comparable to state-of-the-art porous organosilica dielectrics. Moreover, the fast nucleation and volume expansion accompanying the oxide-to-MOF conversion enable uniform growth and gap-filling of narrow trenches, as demonstrated for 45 nm half-pitch fork-fork capacitors.