Introduction

Water adsorption plays a crucial role in numerous industrial processes, particularly in dehydration of gases1,2. Efficient removal of water from gas streams is crucial to prevent corrosion, equipment malfunction, and the formation of gas hydrate3,4. The petrochemical industry, a cornerstone of global energy supply and chemical production, processes critical gases such as natural gas (NG), olefins, and paraffins5,6. During production, NG is typically saturated with water vapor from underground reservoirs7, while olefin/paraffin steams often contain water vapor introduced during methanol-based conversion processes8,9. In both vapor and liquid forms, water presents significant challenges—causing corrosion (especially in the presence of acid gases) and forming methane hydrates that can clog pipelines2,10,11. Therefore, effective dehydration is essential to ensure compliance with industrial standards and specifications.

Adsorptive dehydration using solid adsorbents offers a desirable, low-cost, and energy-efficient solution, with the potential to achieve dehydration and/or purification in a single operation10,12. An ideal water sorbent should: (i) exhibit high water sorption capacity, (ii) maintain long-term stability and recyclability across multiple water adsorption-desorption cycles, (iii) regenerate at low temperatures, and (iv) possess highly competitive water sorption over other gases. In industrial settings, dehydration is typically conducted via temperature swing adsorption, with water adsorption occurring at ambient temperatures (20–35 °C) and regeneration at elevated temperatures7,12. Conventional adsorbents such as molecular sieves (e.g., zeolites 4 A, 5 A, and 13X) are widely used in these processes13,14. However, they require energy-intensive regeneration (e.g., heating above 200 °C) to maintain sufficient working capacity between cycles. Metal-organic frameworks (MOFs), as newer type of crystalline sorbents, have shown promise for water-adsorption-related applications1,15,16,17,18,19. Nevertheless, due to the nature of their inherent coordination bonds, most MOFs exhibit only moderate or poor hydrothermal stability20,21,22. Even the most stable MOFs reported to date will inevitably suffer from slow decomposition or degradation over time, especially under repeated exposure to elevated temperatures (typically >100 °C) during reactivation cycles for a long duration (months to years)1,21,23. This deficiency severely impedes their practical use in real-world applications requiring thousands or even millions of adsorption-desorption cycles. Moreover, MOF synthesis often requires expensive ligands and toxic organic solvents, which severely limit their large-scale and cost-effective production at industrial scale24,25,26,27. In addition, some MOFs, despite of high water uptake, also adsorb other molecular species, hampering their selectivity.

In conventional porous adsorbents, dehydration of industry gas mixtures is typically achieved by physisorption and/or weak chemisorption on pore surfaces28,29. In contrast, water adsorption can also occur in nonporous or nearly nonporous compounds. A notable example is hygroscopic salts, such as CaCl2, which absorb large amounts of water into its bulk via hydration30. However, these materials tend to deliquesce over time, leading to desiccant leakage and subsequent corrosion of the dehydration unit31,32,33. They also require relatively high temperatures ( > 175 °C) for full regeneration.

Herein, we present a unique type of zero-dimensional (0D) molecular compounds, 0D-M(PyC)2(H2O)4 (abbreviated as M-PyC, where M = Mn, Co, Ni, Zn; PyC = 4-pyridinecarboxylic acid), that can efficiently address these limitations of existing desiccants. These molecular complexes undergo fully reversible and rapid water decoordination (desorption or activation) and recoordination (adsorption) processes, accompanied with a structural transformation between their crystalline and amorphous states (Fig. 1a). This transformation fundamentally differs from the “breathing” or “gating” transitions typically observed in flexible MOFs34, such as MIL-53 and DUT-8(Ni)35,36,37, during water-induced sorption which involve two crystalline phases with the same framework connectivity. In addition, the activated amorphous a-M(PyC)2 compounds (denoted as a-M(PyC)2, where "a" stands for amorphous) do not rely on pre-existing pores, reflecting their essentially nonporous nature. All M-PyC compounds feature a three-dimensional (3D) hydrogen-bonded network consisting of M(PyC)2(H2O)4 moieties interconnected by hydrogen bonds. Upon heating at 90–120 °C in ambient air, they lose terminal water molecules and transition into an amorphous state. Remarkably, upon exposure to water (liquid or vapor), they regain crystallinity and adsorb large quantities of water (27–30 wt%). Due to their nearly nonporous nature, the a-M(PyC)2 compounds do not adsorb NG or any hydrocarbon (HC) gases, effectively acting as ideal molecular sieves for these species (Fig. 1b). Furthermore, since water desorption involves a crystalline-to-amorphous phase transition, it can be achieved simply by rapid heating in air, without the need for any additional treatment or reactivation steps under inert atmosphere—as required for most other adsorbents—making the process economically highly viable. Most significantly, the amorphous nature of these molecular sorbents eliminates the stringent need to maintain crystallinity, which constrains the longevity of crystalline adsorbents such as MOFs and zeolites. As a result, such water adsorbents offer potentially unlimited reusability, paving the way for their practical deployment in real-world dehydration processes.

Fig. 1: Schematic illustration of crystallization-assisted water sorption for industrial gas dehydration.
Fig. 1: Schematic illustration of crystallization-assisted water sorption for industrial gas dehydration.The alternative text for this image may have been generated using AI.
Full size image

a Schematic representation of the reversible structural transformation between the amorphous and crystalline states, involving water-induced recrystallization, self-assembly, and hydrogen-bond dissociation during desorption. b Schematic of the M-PyC-based system for practical gas dehydration, associated with water adsorption/desorption, infinite reusability, and complete exclusion of hydrocarbon gases.

Results

Synthesis and crystal structure

Conventional synthesis of sorbent materials is usually conducted under demanding conditions, such as the use of organic templates, high temperatures, agitation and calcination for zeolites38, expensive, high boiling-point and often toxic organic solvents, and sometimes large quantities of modulators for MOFs39,40. These processes are often complex and environmentally hazardous, making scale-up production challenging and requiring multistep activation procedures41. In contrast, in this study, we designed a simple, one-step, easily scalable reaction in aqueous solution at room temperature and ambient pressure, using inexpensive metal salts and ligand to yield a series of pure and highly crystalline M-PyC molecular compounds.

The crystal structures of M-PyC were determined by single-crystal X-ray diffraction (SCXRD) analysis. M-PyC (M = Mn, Ni, Zn) were found to crystallize in the triclinic crystal system, space group P-1 (Supplementary Tables 1–3) with the molecular formula M(PyC)2(H2O)4. As shown in Fig. 2a, each metal center (e.g., Ni) is coordinated to two PyC ligands at the axial position via nitrogen atoms and to four water molecules at the equatorial positions. The hydrogen atoms in the coordinated water molecules further form hydrogen bonds with the oxygen atoms of the carboxylate group from adjacent PyC ligands, generating a 3D hydrogen-bonded network with an average hydrogen bond distance of 1.85 Å (Fig. 2b–c and Supplementary Figs. 1 and 2). The calculated hydrogen bond energies are 0.865 eV (highest) for Ni-PyC, 0.859 eV (slightly lower) for Mn-PyC, and 0.832 eV (lowest) for Zn-PyC (Supplementary Notes), which are consistent with their corresponding decomposition temperatures: 115 °C (highest) for Ni-PyC, 100 °C (slightly lower) for Mn-PyC, and 90 °C (lowest) for Zn-PyC (Supplementary Figs. 3–6). Furthermore, the cohesive energies per M-PyC molecule were determined to be 7.480 eV (highest) for Ni-PyC, 7.433 eV (slightly lower) for Mn-PyC, and 7.210 eV (lowest) for Zn-PyC (Supplementary Notes), reflecting contributions from the combined effects of strong hydrogen bonding and weak van der Waals interactions between the stacked ligand (PyC) rings.

Fig. 2: Crystal structure of M-PyC (M = Mn, Co, Ni, Zn).
Fig. 2: Crystal structure of M-PyC (M = Mn, Co, Ni, Zn).The alternative text for this image may have been generated using AI.
Full size image

a Single M(PyC)2(H2O)4 molecular unit. b Hydrogen bonds between two adjacent units within the network (red dashed lines). c Perspective view of the crystal structure of as-made M-PyC along the c axis. Metal (Mn, Co, Ni, Zn), cyan; carbon, gray; oxygen, red; nitrogen, blue; hydrogen, white.

Water-induced reversible structure transformation and competitive water adsorption

Upon activation, coordinated water molecules were removed from M-PyC, accompanied by the complete loss of hydrogen bonds. As a result, the 3D hydrogen-bonded network collapsed, and the structure decomposed into an amorphous form. Powder X-ray diffraction (PXRD) analysis confirmed that the activated samples, a-M(PyC)2, lost their crystallinity entirely (Fig. 3a and Supplementary Figs. 7–9), indicating a crystalline-to-amorphous transformation during activation. On the other hand, upon re-adsorption of water, all a-M(PyC)2 samples fully reverted to their crystalline phases, yielding PXRD patterns that match perfectly with the as-made M-PyC structures. This structural transformation, associated with water removal (dehydration) and re-coordination (rehydration), is not only fully reversible but can in principle be repeated indefinitely. Thermogravimetric analysis (TGA) confirmed that the weight loss around 90–120 °C corresponds to the release of four equivalents of water molecules (Supplementary Figs. 3–6). Additionally, porosity analysis revealed that the M-PyC compounds are nearly inaccessible to nitrogen gas. The Brunauer-Emmett-Teller (BET) surface areas, estimated from nitrogen adsorption data collected at 77 K on activated a-M(PyC)2 samples, are all below 5 m2/g (Supplementary Fig. 10), verifying their essentially nonporous nature.

Fig. 3: Crystallization assisted water adsorption on Ni-PyC.
Fig. 3: Crystallization assisted water adsorption on Ni-PyC.The alternative text for this image may have been generated using AI.
Full size image

a PXRD patterns of the as-made, activated and rehydrated Ni-PyC samples, illustrating the reversible structural transformation between the water-coordinated crystalline phase and the activated/dehydrated amorphous phase. The intensity is reported in arbitrary units (arb. units). b Water adsorption-desorption isotherms of a-Ni(PyC)2 at 25 °C (P0: 32 mbar for water). c Water adsorption kinetics on a-Ni(PyC)2 at 25 °C and various relative humidity (RH). d Comparison of desorption kinetics for water-adsorbed Ni-PyC samples at different regeneration temperatures. e Adsorption isotherms of water vapor and various gases on a-Ni(PyC)2 at 25 °C, including CH4, C2H4, C2H6, C3H6, and C3H8. f Heat flow and estimated adsorption heats for selected molecules (H2O, CH4, C2H4, C3H6) on a-Ni(PyC)2 obtained from thermogravimetric differential scanning calorimeter (TG-DSC) measurements under ambient conditions.

We compared the water adsorption performance of a-M(PyC)2 with several commercial benchmark water adsorbents (alumina, zeolite 4 A, 5 A, and 13X) and two top-performing MOF-based water adsorbents, MOF-801 and CAU-10 (CAU, Christian-Albrechts-University)15,42,43. The water adsorption isotherm of a-Ni(PyC)2 exhibits a gradual increase at low pressures, followed by a pronounced inflection as the pressure reaches ~16 mbar (50% of RH), boosting the water uptake to its saturated capacity of nearly 300 mg g−1 (30 wt%) at 25 °C (Fig. 3b). The desorption branch of the isotherms clearly shows that water molecules remain undesorbed during the desorption process, suggesting their chemisorptive nature. Other isostructural a-M(PyC)2 compounds exhibit very similar water adsorption behavior (Supplementary Fig. 11). The water uptake capacity of a-M(PyC)2 is well comparable to benchmark water adsorbents at room temperature (25 °C) (Supplementary Table 4 and Supplementary Figs. 12–14): 318 ± 5 mg g−1 (alumina), 273 ± 7 mg g−1 (zeolite 4 A), 206 ± 12 mg g−1 (zeolite 5 A), 334 ± 2 mg g−1 (zeolite 13X), 359 ± 10 mg g−1 (MOF-801) and 368 ± 5 mg g−1 (CAU-10) but will well exceed all of them at higher temperatures as its uptake amount will remain more or less a constant while all other sorbents will inevitably undergo a significant decrease in water uptake due to their physisorption nature. For example, at 35 °C, a-Ni(PyC)2 takes up 290 mg g−1 of water (a decrease of ~3% compared to 25 °C), while the water uptake amounts for zeolite 4 A and zeolite 13X are 253 mg g−1 and 300 mg g−1, respectively, counting a decrease of ~10% and ~24%. The adsorption kinetics study on a-Ni(PyC)2 further confirmed its maximum H2O adsorption rate of 0.32 mmol g−1 min−1: the saturation is reached ( ~ 16 mmol g−1) within 50 min at 95% RH and 60 min at 75% RH (Fig. 3c). Although its adsorption kinetics is slower than zeolites (Supplementary Fig. 16) due to its distinctly different adsorption mechanism involving chemical sorption of water via crystallization-assisted reconstruction of its hydrogen-bonded network, rather than through rapid physisorption as in zeolites, a-Ni(PyC)2 exhibits nearly identical adsorption kinetics under ambient air and completely excludes other gases and maintains its full capacity under these conditions, confirming that the process can be carried out with the same efficiency and performance without the need of protective atmosphere (Supplementary Fig. 15). On the contrary, zeolites suffer from partial pre-adsorption of ambient moisture and/or atmospheric impurities, resulting in up to a 30% loss in water uptake when operated in open air (Supplementary Figs. 24–25). We further investigated the effect of a-Ni(PyC)2 particle size on adsorption rate. We synthesized samples of smaller particle sizes by varying the precursor concentration and performing differential centrifugation. Samples with smaller particles exhibit enhanced adsorption rate under different relative humidities (Supplementary Fig. 17).

To investigate desorption kinetics, we measured desorption curves at various temperatures in open air (Fig. 3d and Supplementary Fig. 18). Complete water desorption required ~60 min at 90 °C. Raising the temperature to 120 °C reduced desorption time significantly and achieved full desorption within 12 min, corresponding to a maximum desorption rate of 1.34 mmol g−1 min−1. These regeneration temperatures (90–120 °C) are significantly lower than those required for zeolites ( ~ 200–300 °C, Supplementary Figs. 19–21). Moreover, Ni-PyC can be fully regenerated within a very short time (30 min) in open air without a protective atmosphere (Supplementary Figs. 22 and 23), highlighting its unique capability for facile and energy-efficient regeneration. In contrast, conventional adsorbents such as zeolites 4 A and 13X require not only much higher temperatures and longer times for complete regeneration but also suffer from substantial losses in their uptake capacity (up to 30%) when exposed to open air. This reduction arises from their rapid adsorption of ambient moisture or atmospheric impurities, leading to partial pre-adsorption after regeneration in open air (Supplementary Figs. 24 and 25), thereby limiting their practicality under open-air conditions and necessitating a controlled or protective atmosphere.

For practical dehydration applications, an optimal adsorbent must simultaneously achieve high water uptake, rapid release/regeneration at low temperatures, and highly competitive water adsorption against HC molecules—specifically the main components of natural gas like methane (CH4) and important chemical feedstocks such as ethylene (C2H4) and propylene (C3H6). To assess competitive adsorption, we measured the adsorption isotherms of CH4, C2H4, C2H6, C3H6, and C3H8 on a-M(PyC)2. Water vapor, interacting strongly via M-OH2 bonds and, as described in detail above, can be readily accommodated, while all HC gases are effectively sieved out, showing negligible adsorption (Fig. 3e and Supplementary Fig. 26). Notably, activated a-M(PyC)2 compounds exhibit remarkably high uptake ratios of water vapor over HC gases, e.g., 138.7 for H2O/CH4, 51.1 for H2O/C2H4 and 40.4 for H2O/C3H6 at 25 °C, demonstrating distinct advantages to zeolites and alumina currently used in industry, as well as some of the best performing MOFs (Supplementary Figs. 27–35 and Supplementary Table 4). Differential scanning calorimeter (DSC) measurements confirmed the binding affinities, with adsorption heats of 49.40, 3.92, 6.30, and 7.03 kJ mol−1 for H2O, CH4, C2H4, and C3H6 on a-Ni(PyC)2, respectively (Fig. 3f), fully consistent with the observed selective adsorption behavior.

In situ probing of water adsorption-assisted recrystallization

To elucidate the relationship between the unique structure formation-deformation behavior associated with water adsorption-desorption in M-PyC compounds, we performed detailed structural analysis of a-Ni(PyC)2 by probing its recrystallization process as a function of time and humidity using humidity-dependent PXRD. As shown in Fig. 4a, at a high RH (90%), adsorption took place immediately as soon as exposing the activated sample to water vapor. The compound began to regain its crystallinity within a few minutes and completed water recoordination within 30 min. Decreasing the RH increased the time required to reach complete recrystallization (Fig. 4a and Supplementary Figs. 37–39). The recrystallization process was highly sensitive to the changes in humidity: an immediate response was observed when a sample initially placed under a low humidity (35% RH) was subsequently exposed to high humidity (90% RH) (Supplementary Fig. 40). Furthermore, in situ PXRD measurements using a humidity- and temperature-controlled cell were performed to monitor the reversible structure transformation between crystalline and amorphous states during water-induced sorption (Fig. 4b). Starting from the amorphous form (activated a-Ni(PyC)2), exposure to water vapor restored its crystalline hydrogen-bonded network, as indicated by the reappearance of characteristic peaks in the PXRD patterns collected at various time intervals. During the subsequent regeneration, the hydrogen-bonded network gradually decomposed upon heating, eventually losing all terminal water molecules and transforming into the amorphous phase, as evident by the disappearance of characteristic diffraction peaks in the PXRD patterns over a short time period. This recrystallization-decrystallization process was fully reversible and reproducible, as demonstrated by multiple consecutive cycles of water vapor adsorption-desorption (Supplementary Fig. 41).

Fig. 4: Visualization of structural transformation in Ni-PyC induced by water sorption.
Fig. 4: Visualization of structural transformation in Ni-PyC induced by water sorption.The alternative text for this image may have been generated using AI.
Full size image

a Humidity-dependent PXRD patterns of Ni-PyC, illustrating the reversible structural transformation between water-coordinated crystalline and dehydrated amorphous phases. Bottom panel: PXRD patterns of a-Ni(PyC)2 at 90% RH collected at different times; top panel: a-Ni(PyC)2 at 65% RH collected over times. b In situ PXRD patterns of Ni-PyC showing its reversible structure transformation between the crystalline and amorphous states. The activated sample was exposed to H2O under 50% RH at room temperature for 5 h and subsequently regenerated by heating to 150 °C at a ramping rate of 10 °C min−1 and was kept at 150 °C for 30 min. Real-time PXRD patterns of the sample were recorded during these processes. The intensity is reported in arbitrary units (arb. units).

To gain a deeper insight into the structural evolution of Ni-PyC during the amorphous-to-crystalline transformation, we conducted in situ IR spectroscopy on the as-made sample during activation and rehydration process to investigate structures lacking long-range order. Vibrational bands were assigned by comparison with the PyC ligand (Supplementary Fig. 42). The results show that key pyridine vibrations (CC/CN stretches and CH deformations) were retained, while new bands at 1750–1300 cm−1 appeared in as-made Ni-PyC (Fig. 5, blue curve), indicating the presence of carboxylate (COO-) groups. A strong band at 1382 cm−1 corresponded to the symmetric COO- stretching mode, while the counterpart vas(COO-) band typically appeared above 1500 cm−1. Broad OH stretching bands at ~3405, 3265, and 2905 cm−1 reflected hydrogen-bonded water; notably, the largely red-shifted 2905 cm−1 band with respect to gas phase value (vas, 3756 and vs, 3657 cm−1) suggested strong hydrogen bonding44,45. A wagging mode of H2O at 938 cm−1 further indicated its coordination of water molecules to Ni2+ via oxygen atoms.

Fig. 5: Probing water-induced recrystallization of amorphous Ni-PyC by in situ infrared spectroscopy.
Fig. 5: Probing water-induced recrystallization of amorphous Ni-PyC by in situ infrared spectroscopy.The alternative text for this image may have been generated using AI.
Full size image

The as-made Ni-PyC was activated at 120 °C and subsequently rehydrated with ~20 Torr water vapor (middle panel). Top black line: difference spectrum obtained by subtracting the spectrum of the activated sample (dark brown) from that of the hydrated sample (light red). Bottom panel: the spectrum (gray) of the crystalline Ni-PyC sample. Top panel: spectrum (purple) of the reactivated sample showing removal of adsorbed water upon annealing at 120 °C. Notation and acronym: ν = stretching, δ = deformation, β = bending (scissoring), as = asymmetric, s = symmetric, ip = in-plane, oop = out of plane, and py = pyridine.

Upon activation at 120 °C, these water-related bands disappeared, and the νs(COO-) band blue-shifted and broadened (Fig. 5, brown curve), overlapping with pyridine CC stretching at 1420 cm−1, reflecting a crystalline-to-amorphous transition. Rehydration by exposure to ~20 Torr water vapor gradually restored the broad water stretching bands above 2800 cm−1, along with the wagging band at 938 cm−1 (Fig. 5, dark- to light-red curves), similar to the as-made structure. This suggests water insertion and binding to Ni2+ as bridging units. The νs(COO-) band gradually shifted back to 1382 cm-1, indicating that COO- groups now interacts with H2O through forming hydrogen-bond. The formation of hydrogen-bond is also clearly evidenced by the evolution of water stretching bands, e.g., red-shifting of the 3380 cm−1 band to 3374 cm−1 and gradual appearance of 2905 cm−1 band within 2 h. Spectral sharpening confirmed the restoration of crystallinity upon H2O insertion. Furthermore, the vas(COO-) band at 1630 cm−1 vanishes upon water adsorption, replaced by a new band at 1590 cm−1 (difference spectrum), suggesting a red shift of ~40 cm−1 upon rehydration. A final dehydration at 120 °C removed the inserted water again, reverting the spectrum to the amorphous state (Fig. 5, purple curve).

Dehydration under conditions mimicking industrial settings

To validate the practical use of M-PyC compounds for industrial dehydration, dynamic column breakthrough experiments were carried out on Ni-PyC for different HC gases under different humidity levels. The results are in excellent agreement with the adsorption strength trends determined from single-component gas (vapor) adsorption isotherms. In all cases, water was preferentially adsorbed onto the column packed with a-Ni(PyC)2, while HC gases including CH4, C2H4, and C3H6 were immediately eluted from the humid mixtures (Fig. 6). The H2O retention times for different HC gases were similar (19 h g-1 at 25% RH, 14 h g-1 at 50% RH, and 12 h g-1 at 75% RH). Regardless of the humidity level, CH4, C2H4, and C3H6 were excluded (no retention in the column) immediately, revealing a negligible adsorption of these gases and extremely competitive adsorption for water vapor. Remarkably, at RH values of 75%, 50%, and 25%, the experiment yielded 2.97 L g−1, 3.97 L g−1, and 5.92 L g−1 of methane, respectively, with water content below 1 ppm (0.0001%) in all three cases (Supplementary Figs. 43–45), well below the industrial standard for deep dehydration ( < 147 ppm)28. Similar productivities were also achieved for ethylene and propylene during dehydration (Supplementary Figs. 46–51). Additionally, humidity-dependent methane breakthrough experiments confirmed the immediate elution of methane from the a-Ni(PyC)2 column (Supplementary Fig. 52), in contrast to the varying retention behaviors observed for other zeolites (e.g., 0.5 h g−1 of methane retention for zeolite 4 A and 1.5 h g-1 for zeolite 13X), even though zeolite 4 A exhibits a longer water elution time (Supplementary Fig. 53). This further highlights the highly competitive water adsorption performance of the M-PyC compounds. To evaluate its recyclability for real-world applications, we also conducted multiple consecutive breakthrough experiments on the same a-Ni(PyC)2 sample. As shown in Supplementary Fig. 54, the dynamic breakthrough curves remained nearly identical across cycles, demonstrating the excellent dehydration reusability of a-Ni(PyC)2.

Fig. 6: Experimental dehydration of important industrial gases by Ni-PyC.
Fig. 6: Experimental dehydration of important industrial gases by Ni-PyC.The alternative text for this image may have been generated using AI.
Full size image

a Breakthrough curves of methane under different humidity conditions (25%, 50%, and 75% RH). b Breakthrough curves of ethylene under different humidity conditions (25%, 50%, and 75% RH). c Breakthrough curves of propylene under different humidity conditions (25%, 50%, and 75% RH).

To further assess dehydration under realistic conditions, we performed breakthrough experiments on a simulated natural gas mixture (CH4/C2H6/C3H8/C4H10 = 80:10:5:5, v/v/v/v) at varying humidity levels (Supplementary Figs. 55–57). The a-Ni(PyC)2 compound effectively captured H2O vapor from the gas mixture, with H2O retention times for dehydration recorded as 17 h g-1 at 25% RH, 16 h g−1 at 50% RH, and 12 h g−1 at 75% RH. Moreover, because HC streams may contain other impurities such as CO2, we examined the CO2 tolerance of a-Ni(PyC)2. The compound exhibits negligible CO2 uptake at 25 °C (Supplementary Fig. 58). We further introduced ~2% CO2 into the simulated natural gas (under 50% RH) during breakthrough experiments (Supplementary Fig. 59) and observed no reduction in dehydration performance. To assess tolerance to sulfur-containing contaminants, the Ni-PyC molecular compound was immersed in 1 M HS- aqueous solution. The PXRD analysis confirmed its structural integrity and high resistance to these harsh conditions (Supplementary Fig. 60). Overall, these results underscore the excellent dehydration performance and impurity tolerance of the Ni-PyC molecular adsorbent across relevant HC gas mixtures.

Recyclability and chemical stability for scalable industrial processes

Many MOFs reported for water adsorption applications lack long-term recyclability and sustained chemical stability, particularly those featuring open metal sites46,47,48. Such materials often experience gradual loss of crystallinity or become amorphous after brief chemical exposure. In contrast, M-PyC compounds represent a rare group of molecular chemisorbents for which crystallinity is no longer required for their activated form, making them highly promising candidates for water-sorption-based applications. To evaluate the recyclability of Ni-PyC, we performed 100 consecutive adsorption-desorption cycles over a time period of 2 months. The water uptake capacity remained constant throughout the entire process, confirming its superior reusability (Fig. 7a). In addition, chemical stability tests under various harsh conditions demonstrated the robustness of the as-made Ni-PyC sample over extended periods. As confirmed by PXRD analysis (Fig. 7b), Ni-PyC retained highly crystallinity after 3 months in open air, 1 day in hot water (80 °C), 3 days in boiling water (100 °C), 2 months in hydrothermal autoclave reactor (100 °C), and 1 day in highly acidic and basic aqueous solutions, highlighting its exceptional durability.

Fig. 7: Recyclability and chemical stability for scalable industrial processes.
Fig. 7: Recyclability and chemical stability for scalable industrial processes.The alternative text for this image may have been generated using AI.
Full size image

a Adsorption-desorption recyclability tests of water on a-Ni(PyC)2 over 100 consecutive sorption cycles (adsorption at 25 °C and reactivation at 120 °C). b PXRD patterns of Ni-PyC after treatment under various conditions. The intensity is reported in arbitrary units (arb. units). c Scale-up synthesis of Ni-PyC product (1.62 kilograms) obtained from a one-pot rapid mixing of aqueous solutions, along with molded pellets of Ni-PyC samples. d Radar chart of six key factors used for the evaluation and guidance of desiccants during the dehydration process.

Another critical aspect is the scalability of candidate materials for industrial production49. The use of inexpensive, commercially available raw chemicals and eco-friendly reaction conditions renders M-PyC compounds well suitable for large-scale synthesis. As an illustrative example, we successfully prepared 1.62 kilograms of Ni-PyC in a simple one-pot reaction at room temperature, using water as the sole solvent. The reaction was completed in just 1 h with a quantitative yield ( > 99%, Supplementary Fig. 61). The product was readily formulated into uniform pellets (Fig. 7c) and exhibited high crystallinity (Supplementary Fig. 62). Overall, Ni-PyC offers a compelling combination of high water uptake capacity, facile synthesis, easy scalability and unlimited recyclability. Very importantly, the entire adsorption-desorption process can be conducted in open air without the need of protection, which is impossible for most other water sorbents. These attributes position the M-PyC compounds strong candidates for industrial dehydration applications (Fig. 7d).

Discussion

In the quest for highly efficient and economically viable water adsorbents for industrial deep dehydration of hydrocarbon feedstock gases, we have discovered a series of molecular (0D) coordination complexes M-PyC (M = Mn, Co, Ni, Zn), featuring 3D hydrogen-bonded networks. Unlike conventional water sorbents that are typically highly crystalline, these molecular “water-trapping” agents operate in an amorphous state. Their unique crystallization-assisted chemisorption of water, coupled with a fully reversible structural transformation between crystalline and amorphous states via rapid heating (desorption) and water trapping (adsorption) in open air, endows them with both indefinite recyclability and remarkable selectivity—capturing large amounts of water while completely excluding hydrocarbons, including key feedstock gases such as methane, ethylene and propylene. Other highly advantageous features include rapid regeneration at relatively low temperatures, cost-effective, facile, and environmentally friendly synthesis that is easily scalable to the kilogram level, and, most notably, the elimination of the stringent framework stability requirements typically imposed on crystalline adsorbents. Together, these attributes render the M-PyC compounds as truly competitive candidates for practical implementation in industrial dehydration processes.

Unlike conventional molecular adsorption processes that depend on maintaining crystalline structures, this work introduces a conceptually distinct approach: employing amorphous molecular complexes as chemisorbents for exceptionally selective and robust water adsorption. The non-crystalline nature of these molecular water sorbents offers potential solutions to longstanding challenges faced by crystalline sorbent materials (e.g., MOFs, COFs, HOFs), including framework degradation, reliance on expensive chemical reagents, operation under protective environment, and use of toxic organic solvents.

Methods

Materials

All chemicals and gases were purchased commercially and used without further purification. 4-pyridinecarboxylic acid (PyC, 99%), manganese acetate tetrahydrate (Mn(CH3COO)2·4H2O, 99%), cobalt acetate tetrahydrate (Co(CH3COO)2·4H2O, 99%), nickel acetate tetrahydrate (Ni(CH3COO)2·4H2O, 99%), and zinc acetate tetrahydrate (Co(CH3COO)2·4H2O, 99%) were purchased from Alfa Aesar. Nitrogen, carbon dioxide, methane, ethylene, ethane, propylene, and propane gases were obtained from Praxair with a purity of 99.9 + %.

Synthesis of M-PyC (M = Mn, Co, Ni, Zn) crystals

4-pyridinecarboxylic acid (PyC, 1 mmol) was first dissolved in deionized water (15 mL). An aqueous solution (5 mL) of nickel acetate tetrahydrate (Ni(CH3COO)2·4H2O, 0.5 mmol) was then added to the PyC solution in a 25 mL glass vial and stirred for 1 h at room temperature. The resulting blue crystals of Ni-PyC were washed with water, filtered, and dried in air. Syntheses of Mn-PyC (colorless crystals), Co-PyC (orange crystals), and Zn-PyC (colorless crystals) followed the same procedure, with the metal salt replaced by manganese acetate tetrahydrate (Mn(CH3COO)2·4H2O), cobalt acetate tetrahydrate (Co(CH3COO)2·4H2O), or zinc acetate tetrahydrate (Zn(CH3COO)2·4H2O), respectively.

Scalable synthesis of Ni-PyC

In a typical procedure, 4-pyridinecarboxylic acid (8 mol, 0.985 kilograms) was dissolved in 18 L deionized water in a 22 L container. Ni(CH3COO)2·4H2O (6 mol, 1.490 kilograms) was then added to the aqueous solution and stirred at room temperature for 1 h. The crystalline product was collected by filtration and dried in air. A dry product weighing 1.62 kilograms was obtained, corresponding to a qualitative yield ( > 99%, ligand-based). For pellet molding, a portion of the Ni-PyC powder was mixed with 1 wt% of methyl cellulose as a binder material, and an appropriate amount of water was added to form a uniform paste (e.g., 5 g of Ni-PyC, 2 mL of deionized water, and 50 mg of methyl cellulose). The paste was manually formulated into pellets. Each pellet was approximately 5 mm in diameter and 50 mg in weight.

Physical characterizations

Powder X-ray diffraction (PXRD) measurements were performed on a Rigaku Ultima IV automated diffraction system using Cu-Kα radiation (λ = 1.5406 Å) over a 2 Theta range of 3–40°, with a scan rate of 3° min-1. The operating powder was 40 kV/44 mA. Thermogravimetric analysis (TGA) of the M-PyC samples was conducted using a TA Instrument Q5000IR thermogravimetric analyzer. For each run, about 10 mg of powder sample was loaded onto a platinum pan and heated under a nitrogen flux at a heating rate of 5 °C min-1.

Gas (vapor) sorption measurements

Before gas sorption experiments, fresh samples were activated at 120 °C for 2 h under dynamic vacuum conditions. BET surface area values were estimated based on nitrogen adsorption data collected at 77 K. Single-component water vapor adsorption isotherms, as well as adsorption isotherms for CH4, C2H4, C2H6, C3H6, and C3H8 gases, were recorded using a Micromeritics ASAP 3Flex adsorption analyzer (Micrometrics Instruments), with a water bath used to maintain the sorption tube at a constant temperature of 25 °C.

Water adsorption kinetics

Dynamic water vapor adsorption kinetics measurements were performed using a homemade gravimetric adsorption unit modified from a TGA Q50 thermogravimetric analyzer (TA Instruments). Pure nitrogen or open air was used as the carrier gas. The absorbed amount was monitored by recording the weight change of the sample over time using the TGA computer system. About 20 mg of Ni-PyC sample was first activated at 120 °C under a gas flow for 1 h to remove residual water molecules. The temperature was then cooled to 25 °C, and a secondary gas stream, passed through a bubbler filled with pure liquid water, was mixed with the dry gas stream. The resulting mixed gas stream (total flow rate of 40 mL min−1) was introduced into the adsorption chamber, and the adsorption process was monitored until equilibrium (saturated capacity) was reached. The recorded weight change were converted into transient normalized uptake curves, reflecting the adsorption kinetics over time.

Multicomponent column breakthrough tests

Breakthrough experiments were carried out using an automated mixed-gas breakthrough apparatus (3 P MIXSORB). About 100 mg of Ni-PyC sample was loaded into the column. The packed adsorbent was activated by heating to 120 °C for 2 h under helium purging (10 mL min−1). After stopping the helium flow, humidified feed gases (pure CH4, pure C2H4, pure C3H6, or mixed CH4/C2H6/C3H8/C4H10 at 80:10:5:5, v/v/v/v)) were pass through the adsorption bed at a flow rate of 3.5 mL min−1. The outlet gas was analyzed using a mass spectrometer (MKS). The breakthrough amount (qi) was calculated by integrating the outlet flow rate f(t) over time according to the following equation:

$${{{\rm{q}}}}_{{{\rm{i}}}}=\,\frac{{\int }_{0}^{{{{\rm{t}}}}_{0}}{{\rm{f}}}({{\rm{t}}}){{\rm{dt}}}}{{{\rm{m}}}}$$
(1)

where m is the mass of the adsorbent used for the test.

The purity (c) of the breakthrough gas was calculated using the following equation:

$${{\rm{c}}}=\frac{{{{\rm{q}}}}_{{{\rm{C}}}{{{\rm{H}}}}_{4}}}{{{{\rm{q}}}}_{{{{\rm{H}}}}_{2}{{\rm{O}}}}+\,{{{\rm{q}}}}_{{{\rm{C}}}{{{\rm{H}}}}_{4}}}{{\rm{or}}}\frac{{{{\rm{q}}}}_{{{{\rm{C}}}}_{2}{{{\rm{H}}}}_{4}}}{{{{\rm{q}}}}_{{{{\rm{H}}}}_{2}{{\rm{O}}}}+\,{{{\rm{q}}}}_{{{{\rm{C}}}}_{2}{{{\rm{H}}}}_{4}}}{{\rm{or}}}\frac{{{{\rm{q}}}}_{{{{\rm{C}}}}_{3}{{{\rm{H}}}}_{6}}}{{{{\rm{q}}}}_{{{{\rm{H}}}}_{2}{{\rm{O}}}}+\,{{{\rm{q}}}}_{{{{\rm{C}}}}_{3}{{{\rm{H}}}}_{6}}}$$
(2)

Heat flow measurements

The heat flows during the adsorption process were measured for water vapor, methane, ethylene, and propylene on Ni-PyC using a differential scanning calorimetry (DSC) system (TA Instruments). Measurements were conducted by feeding a flowing stream of gas (water vapor, methane, ethylene, or propylene) over about 10 mg of activated sample at 25 °C. Prior to water vapor measurements, the baseline was recorded under dry nitrogen flow at 25 °C. Subsequently, nitrogen gas was passed through a saturator containing liquid water at 25 °C, and the resulting water vapor was introduced into the system while the DSC signal was recorded. The calorimetry curves were collected by real-time detection of the heat flow changes inside the chamber during the adsorption process. The enthalpies (ΔH) were determined by integrating the heat flow over time, using the following equation:

$$\Delta {{\rm{H}}}=\frac{{{\rm{S}}}\times {{\rm{m}}}}{6\times {10}^{-2}\times {{\rm{n}}}}$$
(3)

where ΔH is the enthalpies (kJ mol−1), S is the integrating area (W min g-1), m is the sample mass (g), and n is the amount of adsorbed gas (mol).

Humidity-dependent powder X-ray diffraction

Humidity-dependent PXRD measurements coupled with water vapor adsorption were performed on a Bruker D8 Advance system (Cu Kα radiation, λ = 1.5406 Å) equipped with a temperature-humidity chamber for adjusting the humidity level. About 20 mg of as-made Ni-PyC was first activated at 120 °C for 1 h, and then placed into the humidity chamber to adsorb water vapor at different humidity levels for various time periods. PXRD patterns were collected after the sample was exposed to different humidity conditions.

In situ PXRD measurements were carried out on a Bruker D8 Advance diffractometer equipped with an XRK 900 in situ cell. The Ni-PyC sample was loaded into the cell, activated under nitrogen flow, and subsequently cooled to 25 °C for water adsorption. Humid nitrogen (50% RH) was then introduced to the cell to promote water uptake accompanied by recrystallization-assisted structural changes. The sample was subsequently heated to dehydrate the water-saturated compound, with PXRD patterns recorded continuously throughout the entire process.

In situ infrared spectroscopy

In situ IR measurements coupled with water vapor adsorption were performed on a Nicolet iS50 infrared spectrometer equipped with a liquid nitrogen-cooled mercury cadmium telluride (MCT-A) detector and a temperature controller. The spectrometer was fitted with a vacuum cell placed in the main compartment, positioning the sample at the focal point of the infrared beam. To avoid direct pressing that might damage the crystalline structure, the Ni-PyC sample (5 mg) was prepared as a slurry by mixing with a small amount of water and pasted onto a KBr pellet. After drying the sample by blowing a nitrogen stream for a few minutes, it was directly attached onto the KBr pellet and transferred into the cell, which was connected to a vacuum line for evacuation and a water vapor line for controlled moisture exposure to enable adsorption. The sample was activated by heating to 120 °C under vacuum, then cooled back to room temperature for water vapor adsorption measurements. All spectra were recorded in transmission mode over a frequency range of 600–4000 cm−1 with a spectral resolution of 4 cm−1.

Recyclability tests

Recyclability tests of Ni-PyC were performed using a gravimetric adsorption analyzer over 100 consecutive water adsorption-desorption cycles. About 50 mg of Ni-PyC sample was first activated at 120 °C for 1 h prior to the experiments. After cooling to room temperature, water vapor generated from a liquid water bubbler was introduced into the adsorption chamber. Once water adsorption reached equilibrium, the water-adsorbed sample was desorbed at 120 °C for 1 h to reactivate it for the next adsorption-desorption cycle.