Exploring the stability of BO_x at various inorganic supports

Abstract

Oxidative dehydrogenation of propane (ODHP) is a promising alternative route for producing light olefins, especially propene. Since the discovery of the exceptional activity of h-BN and other boron-based solids, their role in ODHP has attracted strong interest but remains insufficiently understood. Here, we provide the first direct experimental evidence of volatile boron oxide (BOₓ) species under ODHP conditions, revealed by TPD–MS and supported by XPS and solid-state NMR analyses. Advanced MAS ssNMR showed preferential coordination of BOₓ to Al in SiO₂–Al₂O₃ supports. BOₓ dispersion and stability were found to be strongly support-dependent: silica-supported BOₓ facilitates sublimation of boron oxides and propane activation at lower temperatures compared to γ-Al₂O₃ or SiO₂–Al₂O₃. Despite this, all systems follow identical selectivity–conversion trends. These results highlight a mechanistic pathway where volatile boron intermediates influence catalytic performance, advancing fundamental understanding and suggesting new strategies for designing selective, energy-efficient catalysts.

Boron-containing catalysts, particularly boron oxide (BOₓ) impregnated onto inorganic supports, have recently attracted attention due to their high selectivity and activity in the oxidative dehydrogenation of propane (ODHP), however, the underlying mechanisms are not fully understood. Here, the authors observe volatile BOₓ species under ODHP conditions, showing that BOₓ dispersion and stability depend on the nature of the inorganic support, with silica supports facilitating BOₓ sublimation and propane activation at lower temperatures compared to γ-Al₂O₃ or SiO₂–Al₂O₃.

Introduction

Boron-containing catalysts, particularly boron oxide impregnated onto various inorganic supports, have recently attracted significant attention due to their high selectivity and activity in the oxidative dehydrogenation of propane (ODHP)^1–6. Compared to conventional transition-metal or acid–base catalysts (e.g., vanadium oxides or alumina), boron-based catalysts typically require lower reaction temperatures and produce significantly lower amounts of CO₂. A plausible explanation suggested in the literature is that even small concentrations of boron oxide can suppress the dissociative adsorption of propane and oxygen, thereby reducing the overall propane conversion while markedly enhancing olefin selectivity^7,8.

Despite extensive research, the precise reaction mechanism of oxidative dehydrogenation of propane (ODHP) over boron-containing catalysts remains under debate, particularly concerning the identification of active sites and the interplay between gas-phase and surface-mediated processes. Early mechanistic proposals, such as those involving boron oxide supported on γ-alumina, suggested that surface-derived propyl radicals play a crucial role as key intermediates. These intermediates were considered responsible for enhanced propane conversion without compromising olefin selectivity⁷. Recent studies emphasize propane adsorption on the boron oxide (BO_x) phase, which exhibits liquid-like behavior under typical catalytic conditions. This adsorption process leads to propyl chemisorption and the subsequent formation of surface hydroxyl (B–OH) species. The regeneration of these active sites through dehydration is generally considered the rate-limiting step^9–11. Alternative analyses suggest that gas-phase reactions, initiated catalytically and occurring under specific thermodynamic conditions (e.g., temperature, reactor volume, oxygen partial pressure), might significantly contribute to the reaction pathway. These gas-phase processes align closely with mechanistic concepts initially proposed in the 1990s^12–14.

An intriguing possibility, not yet thoroughly explored in the literature, is the role of boron species released into the gas phase. Boron leaching at reaction temperatures has been clearly observed with inorganic supports such as silica and alumina¹⁵. It can be hypothesized that these leached boron species might directly participate in radical chain reactions, altering both the selectivity and efficiency of the ODHP process. Clarifying this unexplored aspect may significantly enhance our mechanistic understanding of the overall ODHP process and provide critical insights into optimizing catalyst performance and reaction conditions.

The extent of boron leaching from the BO_x phase supported on various oxides can serve as an indicator of chemical stability and the strength of the interaction with the support. Understanding the factors influencing boron leaching is essential, as they directly impact catalyst lifetime, regeneration potential, and the persistence of active boron species under ODHP conditions. For example, the stability of BO_x can be substantially enhanced by depositing an Al₂O₃ overlayer onto a SiO₂ support via atomic layer deposition¹⁵. In contrast, for BO_x directly supported on SiO₂, an initial monolayer boron loading of about 1.7 wt.% drops to 0.5 wt.% in the spent catalyst (after 24 h in reaction), clearly indicating poor stability and limited retention of the active phase under reaction conditions¹⁶. Moreover, the chemical identity of the leached boron species remains uncertain: these species tend to condense in cooler parts of the reactor, which hinders their direct identification and reliable quantification.

In this work, we focus on identifying the boron species leached from BO_x/support catalysts and evaluating the stability of the impregnated BO_x phase on fumed silica (SiO₂), γ-alumina (Al₂O₃), and mixed silica–alumina (SiO₂–Al₂O₃ with Si/Al = 5) supports. We also discuss the potential catalytic implications of BO_x stability and leaching behavior. The study demonstrated that the volatility of BO_x complexes depends on interactions with the support and correlates with the observed catalytic activity in the oxidative dehydrogenation of propane. Pulse experiments, in which boric acid solution was directly injected into the hot zone of a catalyst-free reactor, showed a significant increase in propane conversion and the production of propene and ethene immediately after injection of a small amount (5 μl) of solution. This observation indirectly demonstrates that BO_x particles in the gas phase promote the conversion of propane to olefins.

Results and discussion

The catalytic performance of BO_x on SiO₂, Al₂O₃, and mixed SiO₂–Al₂O₃ supports is shown in Fig. 1A, B. The support composition strongly affects the onset of propane conversion. Both catalysts containing Al₂O₃ require higher temperatures (by about 80 °C) to initiate conversion, whereas olefin selectivity remains essentially unchanged across supports (it depends more on conversion than on temperature). These trends agree reasonably well with previous reports for BO_x on SiO₂ and Al₂O₃^7,8,11,15,16. The thermal stability of the supported BO_x phase was investigated using temperature-programmed desorption (TPD) coupled with mass spectrometry (MS). Figure 1C illustrates the dependence of the QMS signal intensity on temperature, highlighting changes in desorption rates and the evolution of surface-bound species from the catalysts. As the temperature increases, distinct features appear in the QMS signal, corresponding to the progressive release of specific boron-containing fragments, primarily identified as B₂O₃ and B(OH)₃. These clearly resolved mass-spectrometric signals indicate notable differences in the thermal stability and surface interactions of boron species across Al₂O₃ and SiO₂ supports, respectively. Moreover, both the temperature and quantity of desorbed B₂O₃ and B(OH)₃_₃ can be linked with the observed catalytic behavior; BO_x/Al₂O₃ requires higher reaction temperatures than BO_x/SiO₂ to achieve comparable propane conversions (cf. Fig. 1A, C), consistent with differences in the thermal stability of active boron species.

Fig. 1. The catalytic performance of BO_x/support materials and the volatility of BO_x species.

Temperature dependence of propane conversion (A) and selectivity (B) to propene (black), ethene (red), and their sum (green). C TPD–MS analysis of boron species released from the BO_x/support materials; the typical reaction-temperature range (Tr) is highlighted by the box. D Dosing experiment in which 5 µL of water or aqueous H₃BO₃ solution is injected directly into the gas stream into an empty catalytic reactor at 500 °C, causing a ~25% spike in propane conversion and formation of propene and propane monitored by mass spectrometry.

The link between catalytic activity and the thermal stability of the BO_x phase is not self-evident. To further probe the role of boron species in the gas phase, 5 µL of an aqueous H₃BO₃ solution containing 3.2 µmol of boron was injected into an empty reactor (6 cm³ free volume) under reaction conditions identical to those used in catalytic tests. This boron amount is approximately 70-fold lower than that present in 100 mg of the catalyst. The reactor volume was comparable to that used in experiments with loaded catalysts (Fig. 1D). Propane conversion increased by approximately 20%, accompanied by a corresponding increase in olefin concentrations (propene and ethene) in the gas. Estimated olefin selectivity was very similar to that observed over the supported BO_x catalysts. The return of concentrations of monitored substances to the pre-injection level is relatively slow (at least several minutes). To verify that the enhancement of propane conversion is caused by the addition of boron to the gaseous reaction mixture and not by the presence of water, the metal surface of the needle, or anything related to the injection process itself, 5 µL of pure, redistilled water was dosed into the reactor as a blank. It is clear that the injection of water causes only very small changes in the concentration of both propane (an increase in conversion of about 5%), propene, and ethene, and the system very quickly (within 30 s) returns to the level of concentrations of the monitored components before the injection. Under these conditions, H₃BO₃ is expected to rapidly dehydrate to HBO₂ and subsequently to B₂O₃; although the equilibrium vapor pressure of B₂O₃ at ODHP temperatures is very low, boron-containing volatile species can still form in the presence of water^17,18. Pulse experiments demonstrated that the presence of BO_x species in the gas phase dramatically enhanced the conversion of propane to propene and ethene. Approximately 70 times less boron than is present on 100 mg of catalyst dosed into the reaction mixture by the pulse is able to cause propane conversion that exceeds the conversions obtained on supported catalysts in our catalytic tests. Although the contribution of surface BO_x species in reaction catalyzed by solid catalyst cannot be excluded, this dosing experiment, together with the TPD results, points to a significant role of gas-phase BO_x species in the radical chain mechanism.

Standard XPS measurements were also performed on fresh and annealed samples (Fig. 2A). The measured B 1s binding energies of fresh samples closely match values reported previously in the literature^16,19–23. For BOₓ supported on Al₂O₃, the B 1s peaks ranged from approximately 192.5 to 192.9 eV and exhibited minimal variation with BO_x loading up to monolayer coverage. This downward shift in B 1s binding energy relative to B₂O₃ (~193.3 eV) is consistent with theoretical predictions, which indicate that Al–O–B linkages reduce the B 1s binding energy compared to environments where boron is coordinated exclusively to B–O–B structural motifs, as in B₂O₃ or B(OH)₃ (see Tables S2 and S3). Moreover, the absence of crystalline B₂O₃ or B(OH)₃ in XRD analysis (see Fig. S1) suggests well-dispersed BO_x species on the Al₂O₃ support. The XPS signals exhibit no significant changes upon thermal annealing during the TPD experiment, indicating that surface changes do not induce substantial alterations in boron coordination in the BO_x/Al₂O₃ material.

Fig. 2. XPS and MAS ssNMR characterization of BO_x species on investigated catalysts.

A B 1s XPS signal of BO_x/support materials as received (fresh) and after TPD experiment (annealed). ¹¹B MAS spectra of B BO_x/Al₂O₃ and C BOₓ/SiO₂ (18 kHz, 14.1 T). The black and red lines correspond to experimental spectra and to the sum of the best-fit quadrupolar line shapes. Individual components are shown as filled color patterns. D ¹¹B-REDOR dephasing curve (for details see ref. ²³) for BO_x/Al₂O₃ (red) and BOₓ/SiO₂-Al₂O₃ (blue) (31.25 kHz, 11.7 T) compared with SIMPSON simulation of the dephasing curve for a two-spin system using dipolar coupling of 0.7 and 1.2 kHz and quadrupolar coupling of ²⁷Al 3.1 and 6.4 MHz corresponding to octahedral (Al_O) and tetrahedral (Al_T) coordination environment of Al.

In the case of BOₓ/SiO₂, XRD analysis reveals the presence of crystalline B₂O₃ and/or B(OH)₃ phases even at monolayer BO_x loadings, indicating partial phase separation rather than full dispersion of the BO_x species. The corresponding B 1s XPS signal appears in the range of 193.1–193.7 eV, consistent with boron fully coordinated by oxygen. Theoretical calculations (Table S3) predict that when boron is integrated into the SiO₂ network through Si–O–B linkages, the B 1s binding energy shifts slightly higher than in B₂O₃. This reflects the stronger binding of the B 1s electron in such environments, supporting the interpretation that the observed XPS signal is consistent with either molecular B₂O₃/B-OH species or BO_x phase supported on SiO₂. Upon annealing at ~727 °C, the BO_x/SiO₂ samples exhibited significant shifts of the B 1s peak toward higher binding energies, not observed for Al₂O₃-supported counterparts. The thermally induced shift in B 1s binding energy is consistent with the partial desorption of loosely bound BO_x clusters, which increases the relative proportion of Si–O–B linkages. This change in coordination environment leads to stronger B 1s electron binding and, consequently, higher binding energies. DFT calculations indicate (see Supplementary Information for more details) that previously proposed fourfold-coordinated boron species, stabilized by adjacent protons (i.e., B(OSi)₃OH), are not thermodynamically favorable on the SiO₂ surface^11,16.

The stability of the BO_x phase supported on various materials was also evaluated using straightforward methanol-washing experiments. The boron content, as measured by ICP-MS and XPS, for fresh (as-synthesized) and methanol-washed materials, is summarized in Table S1. The data show a distinct support-dependent trend in boron removal following methanol washing. Approximately 90% of boron is extracted from BO_x/SiO₂, while only around 50% is removed from BO_x/SiO₂–Al₂O₃, and just 30–40% from BO_x/Al₂O₃. These variations indicate significantly stronger boron–surface interactions on alumina-rich supports compared to silica. The boron loss quantified by bulk-sensitive ICP-MS is corroborated by parallel reductions in the B 1s signal observed via surface-sensitive XPS, demonstrating qualitatively consistent trends across both techniques.

The different character of the BO_x phase on the inorganic supports containing ²⁷Al was also proved by solid-state NMR (Fig. 2B, C). The ¹¹B spectra of BO_x/Al₂O₃ are very similar to reference spectra of sodium aluminoborate glass (Figure S6), which can be deconvoluted to major trigonal BO₃ species with high C_q (approx. 2.7 MHz) and minor tetrahedral BO₄^– with symmetric environment and therefore low C_q (0.9 MHz)²⁴. In contrast, the ¹¹B spectrum of BO_x/SiO₂ resembles the spectrum of amorphous B₂O₃ (see Table S5 for comparison) and consists of two signals of trigonal BO₃ species, both with C_q about 2.7 MHz, which were assigned to boroxol ring and BO₃ chains²⁵. The hypothesis of the strong binding between BO_x and Al₂O₃ is further supported by ¹¹B²⁶-REDOR solid-state NMR experiment²⁷, which reveals strong effective dipolar coupling between ¹¹B and ²⁷Al for both BO_x/Al₂O₃ and BO_x/SiO₂–Al₂O₃ sample (Fig. 2D)²⁶. Here, the effective dipolar coupling phenomenologically reflects the distance between ¹¹B and ²⁷Al as well as the number of ²⁷Al in the proximity of ¹¹B. The value of 0.7 kHz for BO_x/SiO₂–Al₂O₃ corresponds well to sodium aluminoborate glass (with statistical distribution of ¹¹B and ²⁷Al)^24,28. In the case of BO_x/Al₂O₃, the effective dipolar coupling of d_eff = 1.2 kHz indicates even stronger interaction between ¹¹B and ²⁷Al. We highlight here that the REDOR build-up curve is slightly different for tetrahedral and octahedral Al sites (both of which are present in the material; see ESI), but simulations using SIMSON²⁹ agree reasonably well with experimental data in both cases (see also Figure S8).

Interestingly, even a low aluminum content (Si/Al = 5) in mixed BO_x/SiO₂–Al₂O₃ catalysts resulted in boron leaching behavior similar to pure γ-Al₂O₃. This suggests that boron preferentially binds to the oxygens connected to Al in the mixed SiO₂–Al₂O₃ oxide.

To conclude, our results reveal pronounced support-dependent differences in the thermal stability of the BO_x phase. BO_x species are considerably more stable and interact more strongly with Al₂O₃ than with SiO₂. This is evidenced by reduced boron leaching during methanol washing and by a higher degree of boron incorporation into the bulk of γ-Al₂O₃. Even small amounts of Al₂O₃ in mixed SiO₂–Al₂O₃ supports substantially enhance BO_x stability.

Temperature-programmed desorption further confirms these differences. At ODHP temperatures below 500 °C, significantly fewer boron-containing fragments, assigned to B₂O₃ and B(OH)₃, desorb from BO_x/Al₂O₃ than from BO_x/SiO₂, consistent with stronger BO_x–support interactions on alumina. In line with this, catalytic testing shows that BO_x/SiO₂ reaches measurable propane conversion at temperatures approximately 80 °C lower than BO_x/Al₂O₃ and BO_x/SiO₂–Al₂O₃ catalysts. This trend matches the relative ease of boron release into the gas phase. Notably, the selectivity–conversion relationship in propane ODH is essentially identical for all three systems.

We therefore propose that the temperature required to initiate ODHP is governed by the amount of mobile boron species entering the gas phase. This proposal is strongly supported by the dosing experiment, in which injecting only 5 µL of aqueous boric acid into the reaction stream produced a rapid spike in propane conversion of a magnitude comparable to that of supported BO_x catalysts, while maintaining similar olefin selectivity. Supports that retain BO_x more strongly require higher temperatures to activate ODHP, whereas supports that release BO_x more readily enable activity at lower temperatures. This interpretation provides a coherent explanation for the observed temperature–conversion profiles and is consistent with the idea that volatile boron species participate in, or promote, radical ODHP pathways.

By demonstrating how inorganic supports govern both BO_x retention and propane conversion, this study points to an underexplored factor in ODHP catalysis. Future work should focus on in situ spectroscopic detection of gas-phase boron intermediates to validate or refute the proposed mechanism.

Methods

Boron-based catalysts studied in this paper were prepared by the impregnation method and labeled as follows: BO_x/support-weight% of boron. The following inorganic oxides were used as carriers: fumed silica (Sigma S5505), neutral gamma-alumina catalyst support (Alfa Aesar, Brockmann grade I, product number 11502), and silica–alumina catalyst support (grade 135, Sigma product code: 343358). The catalysts were prepared by impregnation. Typically, the aqueous solution of a certain amount of boric acid at 40 °C is stirred for 3 h and then dried at 60°C in a crucible open to air for 5 h. For low-concentration BO_x/Al₂O₃ samples, the degassed support was impregnated with isopropyl borate dissolved in isopropanol. All studied catalysts were subjected to basic characterization using X-ray powder diffraction to monitor the presence of crystalline phases, N₂ physical adsorption to evaluate porosity and specific surface area, and FT-IR spectroscopy to monitor changes in the amount and nature of surface OH groups. The chemical composition of the prepared catalysts was determined by ICP-MS after microwave high-pressure digestion. Detailed information about the preparation of each BO_x/support material, along with basic characterization, is provided in the Supplementary Information (SI) in Chapter 1. Materials, and 2. Supports and Catalysts Characterization.

The catalytic behavior of investigated boron-based catalysts in propane oxidative dehydrogenation was investigated in a plug-flow fixed-bed tubular reactor with an inner diameter of 15.5 mm, using 100 mg of catalyst with a grain size of 0.35–0.50 mm at the investigated temperature under atmospheric pressure. A reaction mixture consists of C₃H₈/O₂/He = 30/15/55 vol.% and a total flow rate of 20 mL min⁻¹. Before each catalytic test, a pretreatment was performed by He (11 mL min⁻¹) at the same temperature for 2 h. The blank test was conducted in an empty reactor under the same conditions as the test with catalysts. The reaction products were analyzed with an online gas chromatograph (Agilent 7890A) equipped with both a flame ionization detector (FID) and a thermal conductivity detector (TCD). The GS-Carbonplot GC column (50 m long with a diameter of 530 µm) was used for the separation of compounds. The injector and oven temperatures were set to 150 °C, and the detector temperature was 270 °C. Helium was used as the carrier gas with a flow of 3 mL min⁻¹.

To investigate the effect of the presence of BO_x particles in the gas phase on propane conversion, the catalytic reactor was modified to allow a small dose (5 µL) of aqueous boric acid solution to be injected into an empty reactor (without catalyst) using a Hamilton microsyringe. The experimental setup is shown in Fig. 4SA in the Supplementary Information. The reaction mixture was identical to that used for the steady-state catalytic test of the supported BO_x catalysts. A reaction mixture consists of C₃H₈/O₂/He = 30/15/55 vol.% and a total flow rate of 20 mL min⁻¹. A tubular reactor with a free volume of 6 cm³ was equipped with a side branch with a septum for liquid dosing with a Hamilton microdoser. In the first experiment, 5 µL of pure, redistilled water was dosed into the reactor as a blank. In the pulse experiment with boron, a solution was prepared by dissolving 0.4 g of boric acid in 10 mL of water. A 5 μL of the solution containing 3.23 μmol of boron (amount about 70 times smaller than the boron content in 100 mg of catalyst in the catalytic test) was dosed into the reactor through a septum using a Hamilton microsyringe. The reactor temperature was controlled by a temperature controller with a thermocouple placed in a well located in the middle of the reactor. The experiment was carried out at 500 °C. The composition of the gaseous mixture exiting the reactor was monitored using a quadrupole mass spectrometer (OmniStar 300GDS, Pfeiffer Vacuum) at 70 eV ionization employing a channeltron detector in Multiple Ion Detection in Time mode. The masses m/e = 4, 14, 15, 16, 18, 20, 26, 27, 28, 29, 32, 39, 40, 41, 42, 43, and 44 were monitored over time. The intensity was monitored every 1.5 s. The resulting data were evaluated based on the fragmentation pattern of individual substances (propane, propene, ethene, water, and O₂).

Elemental and chemical-state analysis of the samples was investigated using X-ray photoelectron spectroscopy (XPS) in ultrahigh-vacuum chambers (base pressure <10⁻⁷Pa) equipped with an Omicron EA−125 analyzer, with a normal detection angle configuration. Non-monochromatized X-ray sources were operated with an Al anode (hν = 1486.4 eV). All XPS measurements were performed at room temperature. Data fitting was performed in KolXPD software³⁰. The Shirley method has been applied for background subtraction in all cases.

Analysis of desorbed species was performed by temperature-programmed desorption (TPD) using a quadrupole mass spectrometer (QMS, Leybold-Inficon Transpector) to detect gas-phase species. The QMS was placed behind a differentially pumped nozzle in order to separate molecules desorbing directly from the sample surface from the majority of background contributions. The samples were heated from the rear by an enclosed electron beam. The temperature was ramped linearly during the TPD from 27 to 723 °C at 15 °C min⁻¹. All TPD signals related to boron-containing species were normalized to nominal concentrations of B in each respective sample.

Solid-state NMR experiments were conducted on a JEOL 600 MHz spectrometer equipped with a 3.2 mm HX MAS probe using (18 kHz MAS) and a Bruker Advance Neo 500 MHz with a 2.5 mm HXY probe configured for ¹H, ¹¹B, ²⁷Al (31.25 kHz MAS). All samples, except the Na₂B₅Al₂O₁₂ glass, were packed in a glovebox to avoid contact with ambient moisture. Spectra were acquired at room temperature without further regulation in all cases. Chemical shifts were referenced globally by setting the CH₂ signal of the external adamantane sample to 37.8 ppm. RF pulses were calibrated using a Na₂B₅Al₂O₁₂ glass sample and are listed in Table S4 in Supplementary Information.

For characterization, 1D ¹¹B spectra were used, soft-centra-transition selective pulses accommodating the sites with low C_q as well. For ²⁷Al spectra and REDOR experiments, the shortest possible pulses were employed. To suppress probe background, ¹¹B and ²⁷Al characterization spectra were acquired with a spin-echo detection scheme with the echo delay set to a single rotor period. Fitting of quadrupolar line shape to experimental ¹¹B spectra was performed using ssNake v1.4b³¹ software, and all parameters are listed in Table S5.

To understand the mesoscale composition of the studied material, we have tracked the dipolar coupling between ¹¹B and ²⁷Al using REDOR (Rotation Echo Double Resonance)²⁷ experiment in agreement with the typical workflow followed for glassy materials^26,32. Because of the limited sensitivity of ²⁷Al, we have performed ¹¹B-detected REDOR with ²⁷Al recoupling (¹¹B-REDOR). The ¹¹B REDOR was acquired using a standard Bruker “redori” pulse scheme with a single echo in the ¹¹B channel and train for 180˚ pulses in the ²⁷Al channel with xy8 phase cycle. The signal $S$ (with ²⁷Al recoupling pulses) and $S_0$ (without ²⁷Al recoupling pulses) were detected in interleaved mode and integrated over the whole ¹¹B signal width (individual ¹¹B sites were not separated).

Author contributions

Conceptualization: R.B., M.R., and O.B. Methodology: R.B., M.R., O.B., J.B., and V.J. XPS-TPD experiments: V.J., and M.W. MAS NMR experiments: J.B. Catalytic experiments: R.B., and K.K. Computational simulation: M.R., and O.B. Writing – original draft: M.R., and R.B. Writing – review and editing: R.B., M.R., O.B., V.J., and J.B.

Peer review

Peer review information

Communications Chemistry thanks Kazuhiko Amakawa and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All relevant data are provided in the Supporting Information along with the ZENODO archive with 10.5281/zenodo.16761393.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42004-026-01926-7.