Extreme mass-independent fractionation of mercury isotopes driven by atmospheric plasma-induced chemical process

Abstract

Mass-independent fractionation (MIF) of mercury (Hg) provides a powerful tracer for reconstructing Hg cycling in the Earth system. Although generally attributed to photochemistry, many natural isotope signatures remain inconsistent with the established mechanisms. Here, we present experimental evidence that atmospheric plasma-induced reduction produces odd Hg-MIF with magnitudes and patterns distinct from photochemical processes. Plasma reactions yielded far larger odd-MIF signatures than photoreduction, with Δ¹⁹⁹Hg spanning −78‰ to +28‰, the most negative Hg-MIF across all Hg-reduction experiments. The sign of Hg-MIF reverses with solution pH, yielding strongly negative Hg-MIF at pH ≥ 3.0 but positive Hg-MIF at pH ≤ 2.0. This reversal arises from competition between plasma-generated species, where solvated electrons drive negative magnetic isotope effects (MIE) at higher pH, while hydrogen radicals produce positive MIE at low pH. These findings reveal plasma chemistry as a fundamental but previously overlooked driver of Hg isotope fractionation. Given the ubiquity of natural plasmas such as lightning and aurora, plasma processes need to be integrated into models of Hg cycling in Earth’s atmosphere and surface environments.

Experiments provide evidence that plasma reactions drive a unique mercury mass-independent fractionation distinct from photochemistry, identifying an overlooked mechanism that may help explain unexplained isotope signatures in nature.

Introduction

Mercury (Hg) is the only toxic heavy metal that circulates globally in its elemental gaseous form, exerting widespread environmental impacts^1–3. Stable Hg isotopes, particularly mass-independent fractionation (Hg-MIF, e.g., Δ¹⁹⁹Hg)⁴, provide powerful tracers of Hg sources and transformation processes across Earth’s systems^5–7, from modern atmospheric Hg cycling², to ecological evolution^8–10, and paleo-volcanic events¹¹. To date, natural Hg-MIF signatures have been predominantly attributed to photochemical reactions in the atmosphere and surface environments^4,12–17. However, growing evidence shows that natural isotope anomalies can not be reconciled with light-driven processes alone. For example, pronounced negative shifts in Δ¹⁹⁹Hg observed before the evolution of terrestrial vegetation^8,18 and under polar night conditions^19,20, contradict photochemical expectations. Additionally, global Hg cycling models based on existing fractionation frameworks often fail to reproduce the isotopic compositions observed in natural samples^2,3,5,6,21. These inconsistencies highlight the limitations of a purely photochemical perspective and emphasize the need to explore additional non-photochemical mechanisms to explain the diverse Hg-MIF signatures in natural systems.

Plasma, ubiquitous on Earth and beyond (e.g., lightning, auroras, and solar wind), represents an important yet often overlooked driver of global elemental cycling^22–27. Previous studies have shown that plasma can generate bioavailable nitrogen^24,26 and phosphorus²⁵ at fluxes comparable to those of photochemistry. Moreover, plasma has been demonstrated to induce pronounced MIF in elements such as oxygen²⁸, magnesium²⁹, and titanium³⁰, etc. Unlike photochemical pathways driven by photon absorption, plasma induces isotope effects through non-thermal electron impacts and symmetry-driven reactions involving energetic radicals and excited complexes³¹. Despite these findings, plasma-induced Hg-MIF has not been reported. Our earlier work demonstrated that reactive species produced in atmospheric plasma-liquid systems efficiently reduce Hg²⁺ to Hg⁰ via pathways distinct from photochemistry^32–38. Here, we show that plasma-induced reduction, which is mediated by solvated electrons (${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$) and hydrogen radicals (H^•_(aq)) generated at the interface, produces extreme Hg-MIF signatures distinct from those produced by known photo-reduction processes. These findings identify plasma chemistry as a previously overlooked driver of Hg-MIF, offering a plausible explanation for enigmatic isotopic anomalies and providing new constraints for refining global Hg cycling models.

Results

Experimental design

To investigate Hg isotope fractionation during plasma-induced reduction, we constructed a customized atmospheric plasma reactor³⁹ designed to generate a solution anode glow discharge (SAGD). This configuration simulates the conditions of negative cloud-to-ground lightning, which accounts for ~95% of natural discharges²⁷. In this configuration, the Hg²⁺-containing solution itself acts as the anode, while a tungsten electrode serves as the cathode. A stable plasma is sustained using a high-voltage DC power supply operated at 30 mA (Fig. 1a; see Methods). This reactor permits precise control over the discharge current, solution pH, and discharge gas composition, enabling experiments under variable conditions that more closely replicate natural plasma environments. By conducting experiments across a range of pH values, we tracked the evolution of Hg isotopic compositions during the plasma-induced reduction of dissolved Hg²⁺ to elemental Hg⁰.

Fig. 1. Experimental design and isotopic signatures of plasma-induced Hg²⁺ reduction.

a Schematic of the solution anode glow discharge (SAGD) atmospheric plasma reactor used to simulate natural plasma analogous to negative cloud-to-ground lightning. b The plot of Δ¹⁹⁹Hg versus δ²⁰²Hg for samples from SAGD plasma-induced reduction experiments (reactant: Hg²⁺; product: Hg⁰; pH: 3.0 − 5.0), compared with those from representative photochemical reduction studies^4,12,40–42. Gray dashed lines represent reference values at Δ¹⁹⁹Hg = 0 and δ²⁰²Hg = 0, respectively. DOC dissolved organic carbon.

Hg isotope fractionation during plasma reduction

Our most striking finding is that plasma-induced reduction produces exceptionally large odd Hg-MIF with an isotopic signature distinct from those of photochemical pathways. In these experiments, MIF occurred exclusively in odd isotopes (¹⁹⁹Hg and ²⁰¹Hg), with no detectable MIF in even isotopes (²⁰⁰Hg and ²⁰⁴Hg; Fig. S1). The most pronounced MIF appeared in residual Hg²⁺ at pH 4.0, where Δ¹⁹⁹Hg reaches -78‰ after ~90% reduction (Fig. 1b). To our knowledge, this represents the largest Hg-MIF reported for any reduction process. Between pH 3.0 and 5.0, residual Hg²⁺ consistently exhibits negative Δ¹⁹⁹Hg with positive δ²⁰²Hg, indicating depletion of odd-Hg isotopes concurrent with enrichment in heavy even isotopes. This inverse relationship places our data in distinct regions of δ²⁰²Hg-Δ¹⁹⁹Hg space (quadrants II and IV), contrasting with the first-third quadrant clustering characteristic of most photochemical systems^40,41 (Fig. 1b). For clarity, we define this MIF pattern as negative MIF ((-)MIF), in contrast to the positive MIF (( + )MIF) commonly observed in photochemistry^40,41. Photochemical reduction typically produces (+)MIF, whereas (-)MIF is rare and of much smaller magnitude. It has been observed only under specific conditions, such as for Hg complexed with thiol ligands (e.g., Hg(Cys)₂), where Δ¹⁹⁹Hg values are restricted to −1.3‰ to 0.89‰ (Fig. 1b)^40,41. Notably, the remarkably large MIF observed here (Δ¹⁹⁹Hg from −78‰ to +28‰) is accompanied by only modest mass-dependent fractionation (MDF) (δ²⁰²Hg from −0.81‰ to +1.6‰). The combination of exceptionally large negative odd-MIF and minimal MDF shifts demonstrates that plasma chemistry drives unprecedented and distinctive Hg-MIF.

pH dependence of Hg isotope fractionation

During plasma-induced reduction, the residue Hg²⁺ consistently exhibited increasing δ²⁰²Hg values under all pH conditions examined. In contrast, both the sign (±) and magnitude of odd-Hg MIF were highly pH-dependent. Under moderately acidic conditions (pH 2.3–5.0), the reaction produced pronounced (-)MIF, characterized by progressive depletion of odd isotopes in the residue Hg²⁺ (Fig. 2a, b). At lower pH (1.6 and 2.0), however, the residue Hg²⁺ became enriched in both ²⁰²Hg and ¹⁹⁹Hg, generating (+)MIF. In addition to this sign reversal, the magnitude of MIF is markedly suppressed under strongly acidic conditions, with the Δ¹⁹⁹Hg differences falling below 3.3‰. While such low-pH environments are rare in nature, they provide a valuable experimental window for probing the mechanism of plasma-induced Hg-MIF. All plasma reduction reactions followed pseudo-first-order kinetics, expressed as ln([Hg]/[Hg]₀) = -kt, with rate coefficient (k) decreasing as pH declined (from 1.19 h^-1 at pH 5.0 to 0.09 h^-1 at pH 1.6; Fig. S2). Overall, the isotopic variations conformed well to Rayleigh fractionation behavior (Fig. 2a, b), consistent with mass balance constraints.

Fig. 2. Plots of δ²⁰²Hg and Δ¹⁹⁹Hg values as a function of the remaining Hg²⁺ fraction under different conditions.

a Plasma reduction experiments conducted at pH 5.0 to 3.0; b Plasma reduction experiments conducted at pH 1.6–2.3; and c Electrochemical reduction experiment conducted at pH 3.0. Black dashed lines represent predictions from the Rayleigh fractionation models, while the gray dashed lines at δ²⁰²Hg = 0 and Δ¹⁹⁹Hg = 0 denote the reference for no isotopic fractionation.

Control experiments

The SAGD process, often characterized as glow-discharge electrolysis, involves not only plasma chemistry but also UV radiation and charge transfer⁴³, potentially introducing concurrent photochemical and electrochemical effects. To isolate these contributions, we conducted rigorous control experiments. Electrochemical reduction without plasma (Fig. S3a, Supplementary Text 1) rapidly reduced Hg²⁺ (Fig. S2), albeit with slower kinetics than the SAGD progress. However, this pathway produces negligible Hg-MIF (Δ¹⁹⁹Hg from -0.10 to +0.12‰) while yielding measurable MDF (δ²⁰²Hg shifts from -0.86 to 1.54‰, Fig. 2c). Furthermore, exposure to plasma-light irradiation alone, without direct contact between the plasma and the sample (Fig. S3b, Supplementary Text 2), resulted in negligible Hg reduction and consequently no isotope fractionation. Together, these control experiments demonstrate that the anomalously large odd-MIF signatures arise specifically from plasma-induced reduction, rather than conventional electrochemical or photochemical pathways.

Discussion

Large and distinct odd Hg-MIF induced by plasma reduction

One remarkable feature of our results is the exceptionally large MIF observed in odd Hg isotopes (¹⁹⁹Hg and ²⁰¹Hg) during plasma-induced reduction. The measured Δ¹⁹⁹Hg values span an unprecedented range from -78‰ to +28‰, far exceeding the MIF signatures typically reported for photochemical reduction (Fig. 1b). To better capture the intrinsic isotope effects beyond the apparent fractionation in raw isotope ratios, we quantified enrichment factors (E¹⁹⁹Hg) from the slope of Δ¹⁹⁹Hg versus ln(f)^5,44, where f is the fraction of remaining Hg²⁺. As shown in Fig. 3a, E¹⁹⁹Hg varies strongly in both magnitude and sign (±) and exhibits a clear dependence on pH. At pH 5.0 and 4.0, E¹⁹⁹Hg is strongly positive (+28.6‰ and +31.5‰, respectively). However, as the pH declined, E¹⁹⁹Hg decreased sharply, from +31.5‰ at pH 4.0 to +1.5‰ at pH 2.3, and became consistently negative (~ -2.1‰) at pH 2.0 and 1.6. This strong pH dependence, which closely correlates with the significant slowdown of Hg²⁺ reduction kinetics at low pH (Fig. 3b and Fig. S2). In contrast, the enrichment factors for MDF (e.g., E²⁰²Hg) show no clear trend with pH and remain relatively small, typically ranging from -0.5‰ to -1.1‰. Remarkably, to the best of our knowledge, the E¹⁹⁹Hg value of +31.5‰ observed at pH 4.0 represents the highest enrichment factor reported for any Hg reduction process to date (Fig. 3b).

Fig. 3. pH-dependent variation of odd-Hg mass-independent fractionation (MIF) enrichment factors (E¹⁹⁹Hg, ‰) during plasma-induced Hg²⁺ reduction.

a Δ¹⁹⁹Hg plotted against the fraction of Hg²⁺ remaining (f) for plasma-induced reduction experiments. E¹⁹⁹Hg, determined from the slope of the best linear fit between Δ¹⁹⁹Hg and ln(f), is used to isolate the MIF component by removing the contribution from mass-dependent fractionation (MDF), in contrast to the conventional total fractionation factor α that includes both MDF and MIF^5,44. Note: Data marked with open symbols were excluded from the fit to better illustrate the potential of the pH effect on E¹⁹⁹Hg, as prolonged plasma-induced reactions led to pH decreases (Fig. S2). The light purple shaded areas represent the reported ranges of photochemical Hg²⁺ reduction^12,40. b E¹⁹⁹Hg values and reduction rate coefficient (k: h⁻¹) plotted as a function of the pH. The blue and green shaded areas represent the reported ranges of E¹⁹⁹Hg for photochemical reduction of Hg²⁺ and MeHg, respectively^12,14,40. For better comparison, the E¹⁹⁹Hg value corresponding to the (-)MIF pattern in photochemical Hg²⁺-reduction with cysteine at pH 3.6 is also marked with a red dot⁴⁰.

Photochemical reduction experiments generally yield much smaller E¹⁹⁹Hg than those observed in plasma-induced reduction (Fig. 3b). Previous studies have reported that E¹⁹⁹Hg values for photochemical Hg²⁺ reduction typically range from -6.73‰ to +1.03‰^12,40, and from -16.23‰ to -0.08‰ for methylmercury (MeHg) photochemical reduction¹⁴. Notably, one of the most positive E¹⁹⁹Hg reported for photochemical Hg²⁺ reduction (+1.03‰) occurs in the presence of cysteine at pH 3.6⁴⁰; however, this value remains far below the +17.2‰ observed in our plasma-induced reduction experiment at pH 3.0. In photochemical systems, both the sign and magnitude of MIF are strongly influenced by ligand identity and pH⁴¹. While slightly positive E¹⁹⁹Hg can occur under specific conditions, most photochemical reductions of Hg²⁺ and MeHg exhibit clearly negative E¹⁹⁹Hg values. Systems dominated by oxygen-containing ligands consistently produce (+)MIF (negative E¹⁹⁹Hg value) across the tested pH range (3-7)⁴¹, where thiol-ligands can generate either (+)MIF or (-)MIF depending on both pH and redox conditions⁴⁰. In contrast, our plasma-induced reduction experiments, conducted in an aqueous matrix involving only oxygen-containing ligands (OH and H₂O) and no thiol or organic ligands, consistently yielded positive E¹⁹⁹Hg values at pH ≥2.3. Moreover, even small changes in pH induced significant shifts in both the magnitude and the sign of Hg-MIF. Collectively, these results demonstrate that plasma-induced reduction processes produce significant MIF, with enrichment factors that vary with pH and differ clearly from those in photochemical systems.

Several mechanisms have been proposed to explain Hg-MIF, primarily the magnetic isotope effect (MIE) and the nuclear volume effect (NVE)^45,46. The ratio of Δ¹⁹⁹Hg to Δ²⁰¹Hg provides a key diagnostic for distinguishing between these mechanisms and elucidating reaction pathways. The MIE, a kinetic isotope effect, typically produces Δ¹⁹⁹Hg/Δ²⁰¹Hg ratios ranging from 0.85 to 1.46, with both the magnitude and sign of the MIF influenced by reaction conditions¹⁴. In contrast, the NVE, which arises from differences in nuclear charge radii, yields a theoretical ratio of 1.59 ± 0.05 (2SE)^12,41 and generally produces much smaller anomalies (<0.4‰)⁴⁷. As shown in Fig. 4, the Δ¹⁹⁹Hg/Δ²⁰¹Hg slopes obtained from the plasma-induced reduction process vary with pH. From pH 5.0 to 3.0, the slope remains nearly constant at ~0.91 (Fig. 4a) despite a roughly linear decline in reduction rate (Fig. S2). At lower pH, however, the slope first decreases to 0.72 at pH 2.3, then increases to 1.20 and 1.27 at pH 2.0 and 1.6, respectively. This shift coincides with the observed reversal in MIF sign (from negative to positive MIF), indicating a change in the dominant reaction pathway.

Fig. 4. Relationships between Δ¹⁹⁹Hg and Δ²⁰¹Hg obtained from plasma-induced reduction experiments at different pH conditions.

Dashed lines in panels (a, b) denote linear regressions for each dataset. c Comparison of Δ¹⁹⁹Hg/Δ²⁰¹Hg slopes and E¹⁹⁹Hg from plasma-induced reduction experiments in this study with reported slopes from photochemical studies^{4,12,14,40,41}. The gray- and pink-shaded areas indicate the characteristic slope ranges for the nuclear volume effect (NVE) and magnetic isotope effect (MIE), respectively, as observed in photochemical Hg reduction processes¹⁴. PL FAR Pony Lake fulvic acid¹⁴.

The Δ¹⁹⁹Hg/Δ²⁰¹Hg ratios and the large MIF magnitudes observed here (E¹⁹⁹Hg up to +31.5‰), strongly support MIE as the primary mechanism governing MIF during the plasma reduction process. Negative MIF corresponds to Δ¹⁹⁹Hg/Δ²⁰¹Hg ratios <1.0 (0.72–0.91), whereas positive MIF is characterized by values > 1.2. At pH 2.3, the Δ¹⁹⁹Hg/Δ²⁰¹Hg ratio reaches a minimum, suggesting a transitional regime in which both (-)MIF and (+)MIF pathways are active, with the former still dominating. At lower pH, the (+)MIF pathway becomes increasingly prominent, consistent with the rising slope. This transition is further reflected in the magnitude of MIF: at pH 2.3, the absolute value of E¹⁹⁹Hg reaches a minimum (+1.79‰), while deviations from this pH in either direction significantly enhance the extent of MIF.

Control of Hg-MIF by competing reductants in plasma reduction

In our plasma-liquid reduction systems, the two primary reductants are ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ and H^•_(aq)⁴³. Because the liquid phase serves as the anode, electrons generated in the plasma are injected into the solution and become solvated, forming ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$, in which the electron is stabilized within a cage of oriented water molecules⁴³. With a standard reduction potential of -2.88 V, ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ is one of the most powerful reducing agents and can drive rapid Hg²⁺ reduction. By comparison, H^•_(aq) generated either through plasma-driven dissociation of water molecules or proton scavenging of ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ (${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ + H⁺ → H^•_(aq)), also exhibits strong reducing power (E⁰ = -2.31 V). Kinetic data further indicate that ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ is the dominant reductant under neutral to mildly acidic conditions: the rate constant for its reaction with Hg²⁺ (~2.3*10¹⁰L/(mol_•s)) is about an order of magnitude greater than that of H^•₍_aq₎ (~2.3*10⁹L/(mol_•s))⁴⁸. Although ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ can reach high local concentrations (>1 mM), it is rapidly scavenged by protons under acidic conditions (Table S1)⁴⁹. Electron spin resonance spectroscopy (ESR) measurements confirm the presence of ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ in our system and show a marked decrease in its abundance with decreasing pH (Fig. S4, Supplementary Text 3). Together, these observations indicate that${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ dominates Hg²⁺ reduction at pH≥2.3, whereas enhanced proton scavenging at lower pH (2.0-1.6) suppresses ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ and shifts the reduction toward H^•_(aq)^43,49. This pH-dependent switch in the dominant reductant closely parallels the observed MIF sign inversion.

We conducted targeted experiments to investigate the roles of key reactive species in governing Hg-MIF. To specifically assess the contribution of ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$, we used NaNO_2, a well established efficient ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ scavenger⁴⁹. At pH 4.0, addition of 10 or 50 mmol/L NO₂^– significantly suppressed Hg²⁺ reduction, decreasing the rate constant from 0.93 h^–1 (control) to 0.19 and 0.16 h^–1 (Fig. 5a), respectively. Importantly, the maximum Δ¹⁹⁹Hg shifted from -78.3‰ (control) to -1.29‰ and -0.88‰, with the corresponding E¹⁹⁹Hg dropping from +31.5‰ to +1.86‰ and +0.69‰ (Fig. 5b/g). These results strongly implicate ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ as a key driver of the large (-)MIF observed in the plasma-induced reduction. We further test this interpretation at pH 2.3 by adding acetone (1% v/v), which is an even more efficient ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ scavenger but reacts much more slowly with H^•_(aq) (Table S1). Similar to NO₂^–, Hg²⁺ reduction was significantly inhibited. However, in contrast to NO₂^– addition experiments, the isotopic signature switched from negative to positive MIF (Fig. 5b/e): the maximum Δ¹⁹⁹Hg shifted from -2.19‰ to +1.00‰, while E¹⁹⁹Hg changed from +2.32‰ to -1.22‰ (Fig. 5g). This sign reversal reinforces the conclusion that ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ drives (-)MIF, whereas suppression of ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ allows H^•_(aq) mediated pathways to produce (+)MIF.

Fig. 5. Reactive-species control of Hg isotopes mass-independent fractionation (MIF): switching between ${\text{e}}_{\left(\mathrm{aq}\right)}^{-}$ and ${\mathrm{H}}_{\left(\mathrm{aq}\right)}^{\bullet }$-driven reduction pathways govern the sign and magnitude of Hg-MIF.

a–c First-order kinetic fits of Hg²⁺ reduction under different plasma reduction conditions (A, pH=4.0; B, pH=2.3; and C, pH=2.0). d–f Corresponding isotopic variations in δ²⁰²Hg and Δ¹⁹⁹Hg for the experiments shown in a-c (d, pH=4.0; e, pH=2.3; and f, pH=2.0). In panel d, data from the experiment with NO₂^- addition are plotted against the right Y-axis, as the Δ¹⁹⁹Hg values were significantly smaller than those in the pH 4.0 control experiment (plotted against the left Y-axis). (g) Characteristics of E¹⁹⁹Hg changes under different reaction conditions. h Relationship between reaction rate constants and E¹⁹⁹Hg under varying conditions. i Slopes of Δ¹⁹⁹Hg vs Δ²⁰¹Hg obtained from different plasma reduction experiments. The black dashed lines in panels a-c indicate regression fits, and the gray dashed lines in panels d-f denote reference values at Δ¹⁹⁹Hg = 0, δ²⁰²Hg = 0, and E¹⁹⁹Hg = 0.

While a shift in Hg speciation coincides with the transition of Hg-MIF between pH 2-3 (Fig. S5), this change itself does not act as the causative factor for the Hg-MIF sign reversal. Within this pH range, the dominant hydrated Hg²⁺ species shift from [Hg(H₂O)_n]²⁺ (n = 1-7) at pH <3.0 to hydrolyzed [Hg(OH)₂(H₂O)_n] at pH > 3.0 (Fig. S5)^50,51. This speciation change is unlikely to govern the MIF sign reversal, as demonstrated by solution cathode glow discharge (SCGD) experiments, which employ an opposite polarity to our SAGD system and consistently yield positive MIF (Fig. S6, Supplementary Text 4). Therefore, changes in Hg speciation are unlikely to govern the MIF sign reversal. To further test the role of H^•_(aq), we introduced H₂ (50 ~ 100 mL/min) into the Ar discharge, which promotes the generation of aqueous H^•_(aq)^52,53. Under both pH 2.3 (normally yielding (-)MIF) and pH 2.0 (normally yielding (+)MIF), H₂ addition accelerated Hg²⁺ reduction and altered isotopic signatures (Fig. 5c/f). At pH 2.0, increasing H₂ flow enhanced the magnitude of (+)MIF, with E¹⁹⁹Hg shifting from -1.92‰ to -4.59‰ (Fig. 5g). At pH 2.3, H₂ addition reversed the MIF sign from negative to positive, with E¹⁹⁹Hg changing from +2.32‰ to -2.25‰ (Fig. 5d/g). Although H₂ addition can alter plasma temperature, the observed MIF sign reversals are primarily driven by a shift in the dominant reducing species from ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ to H^•_(aq). This shift modulates the spin-dynamics of the radical pair intermediates. Notably, good linear relationships were observed between the first-order rate constant (k) and E¹⁹⁹Hg for both pH 4.0 + NO₂^– and pH 2.0 + H₂ experiments (Fig. 5h). Despite opposite slopes, these results clearly show that Hg-MIF is kinetically controlled and reflects the rate-determining step of Hg²⁺ reduction under each condition (Fig. 6, Eqs. 4 and 5). Corresponding changes in Δ¹⁹⁹Hg/Δ²⁰¹Hg further support a pathway transition: the slope drops from 0.91 to 0.80 at pH 4.0 but rises from 0.80 to 1.04 at pH 2.3, consistent with a switch from ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$ to H^•_(aq)-dominated reduction. In contrast, the minimal slope change at pH 2.0 indicates no alteration of the primary pathway (Fig. 5i).

Fig. 6. Proposed reaction pathways and mechanisms for the observed Hg isotopes mass-independent fractionation (MIF) during plasma-induced reduction under different pH conditions.

The schematic illustrates how reductant pathways diverge with pH: at higher pH, ${\text{e}}_{\left(\mathrm{aq}\right)}^{-}$ rapidly reduce Hg²⁺ species ([Hg(OH)₂(H₂O)_n]) to Hg⁰ via Path 1, generating negative magnetic isotope effect ((-)MIE). At lower pH, proton scavenging suppresses ${\text{e}}_{\left(\mathrm{aq}\right)}^{-}$ and shifts reduction toward H•_(aq), which reduce hydrated Hg²⁺ species ([Hg(H₂O)_n]²⁺) via Path 2, producing positive magnetic isotope effect ((+)MIE). Arrows indicate the dominant pathways, and “Hg²⁺_(aq)”denotes hydrated Hg₂⁺ species, [Hg(OH)₂(H₂O)_n] and [Hg(H₂O)_n]²⁺, which dominants at pH> 3.0 and pH<3.0, respectively. This distribution is consistent with both literature reports^50,51 and Visual MINTEQ 4.0 simulations (Fig. S5).

Based on these findings, we propose that ${\mathrm{e}}_{\left(\mathrm{aq}\right)}^{-}$-driven reduction primarily produces the (-)MIE pattern (path 1), while H^•_(aq)-driven reduction (path 2) likely generates (+)MIE in our plasma-induced reduction system (Fig. 6). These observations can be interpreted in the framework of radical-pair spin dynamics. At pH ≥3.0, ${\text{e}}_{\left(\mathrm{aq}\right)}^{-}$, which are present at high interfacial concentrations and with a long lifetime (~300 μs)^55,56, likely react with [Hg(OH)₂(H₂O)_n] to form singlet radical pairs (path 1). Magnetic isotopes (¹⁹⁹Hg and ²⁰¹Hg) in such singlet radical pairs undergo hyperfine coupling (HFC) driven singlet-to-triplet (S-T) interconversion more readily than nonmagnetic isotopes (²⁰⁰Hg, ²⁰²Hg, and ²⁰⁴Hg)⁴⁵. This hyperfine interaction arises from the coupling between nuclear spins and the local magnetic field generated by electron motion and spin within the radical pair, forming the physical basis of the MIE. Faster S-T conversion reduces their recombination to ground state precursor, thereby enriching odd-Hg in the product Hg⁰ and yielding a (-)MIE signature with Δ¹⁹⁹Hg/Δ²⁰¹Hg ≈ 0.9. Conversely, at lower pH, rapid scavenging of ${\text{e}}_{\left(\mathrm{aq}\right)}^{-}$ suppresses this pathway, and H^•_(aq)-driven reduction becomes dominant. The reaction of H^•_(aq) with [Hg(H₂O)_n]²⁺ (path 2) generates triplet radical pairs that also undergo T-S interconversion; in this case, however, magnetic isotopes are preferentially depleted in the product, yielding (+)MIE (Δ¹⁹⁹Hg/Δ²⁰¹Hg = 1.20 ~ 1.27). While the data clearly establishes a non-photochemical MIF pathway, the underlying spin-state dynamics and molecular interactions warrant further time-resolved spectroscopic investigation.

Implications for the global Hg cycle

Photochemical reactions have long been regarded as the dominant source of odd Hg-MIF in nature. Our experiments reveal a previously unrecognized, plasma-induced pathway. Unlike photoreduction, which typically produces modest positive or only limited negative anomalies (E¹⁹⁹Hg from -16.23‰ to +1.03‰)^12,40, plasma-induced reduction produces exceptionally large negative Δ¹⁹⁹Hg anomalies (E¹⁹⁹Hg up to ~ +31‰) with a consistent Δ¹⁹⁹Hg/Δ²⁰¹Hg slope of 0.91 at pH ≥ 3.0. Importantly, this mechanism remains operative under more realistic atmospheric conditions. In experiments conducted under an Ar/air (1:1) plasma, the residual Hg²⁺ still exhibited a pronounced positive E¹⁹⁹Hg signature (+11.5‰), albeit with slower kinetics compared to pure Ar-SAGD (reduction rate decreasing from 0.93 h^–1 to 0.23 h^–1; Fig. S7). These findings challenge the prevailing view that photochemistry is the sole driver of large Hg-MIF and introduce plasma chemistry as a new pathway for explaining isotopic signatures that have been difficult to reconcile with existing photochemical lab work. Such plasma-driven processes likely compete with, or even dominate, conventional photochemistry in specialized atmospheric environments.

This possibility is particularly relevant in polar regions. Significant negative Δ¹⁹⁹Hg values (as low as -3.45‰ to -5.08‰) have been reported in snow samples collected during, or immediately following, prolonged periods of darkness^19,20. Such observations are difficult to explain solely through photochemical Hg cycling because photochemical reactions are effectively suppressed during the polar night and early spring owing to the absence of sunlight. This unsolved discrepancy is further underscored by recent three-dimensional isotope models, which explicitly exclude polar regions from global simulations, acknowledging that the mechanisms governing high-latitude Hg-MIF remain poorly constrained⁵⁴. Importantly, auroral processes are already known to perturb the polar atmospheric redox budget through the generation of reactive NO_x and HO_x, and have been shown to drive up to 8% depletion of stratospheric ozone⁵⁵. In this context, plasma-induced mercury reduction may represent a plausible non-photochemical isotopic fractionation pathway in polar atmospheres. If plasma-induced reduction yields Hg⁰ with highly positive Δ¹⁹⁹Hg values, the residual Hg²⁺ pool must inevitably inherit the pronounced (-)MIF signatures. This mechanism provides a reasonable, albeit still qualitative, framework linking plasma chemistry to the strongly negative MIF signatures observed in polar snow and ice archives. This interpretation is consistent with the recent proposal by Song et al. (2026)⁵⁴, who identified plasma chemistry, specifically the missing lightning component in atmospheric isotope models, as the likely key to resolving isotopic anomalies that remain recalcitrant to photochemical explanations.

Beyond modern environments, plasma chemistry may also provide a speculative perspective on paleoenvironmental Hg records. Current models typically assume mantle-derived Hg lacks MIF (Δ¹⁹⁹Hg = 0), attributing positive Δ¹⁹⁹Hg in pelagic sediments to photoreduction and negative ones in marginal marine sediment settings to vegetation uptake and subsequent weathering^8,11,18. However, the occurrence of negative Δ¹⁹⁹Hg excursions in pre-Ordovician sediments, prior to the emergence of extensive terrestrial vegetation, remains difficult to reconcile within this conventional photochemical framework. Existing photochemical models require unrealistically restrictive conditions (e.g., near-total light attenuation¹⁸), and often produce Δ¹⁹⁹Hg/Δ²⁰¹Hg ratios (1.46 ~ 1.89 for thiol-ligand photoreduction or photochemical oxidation processes^15,40) that differ substantially from the observed values of ~0.9^56,57. Our findings suggest that plasma processes, such as volcanic lightning⁵⁸ during large eruptions, may represent a potential non-photochemical source of (-)MIF in pre-vegetation intervals. Large explosive eruptions are capable of generating intense lightning activity, as illustrated by the 2022 Hunga Tonga eruption⁵⁹. Nevertheless, whether volcanic lightning can generate negative Hg-MIF signatures and preserve them in the geological record remains to be systematically evaluated.

To assess the potential global significance of this mechanism, we extrapolated our air-Ar experimental yields to modern lightning conditions using a flash rate of 44 ± 5 s^-1 and an average energy of 6.7 GJ per flash^60,61. This calculation suggests that lightning could theoretically reduce ~570 tons of Hg²⁺ annually, corresponding to approximately 9.5% of troposphere Hg²⁺ reduction fluxes (Supplementary Texts 5)^54,62. Clearly, these estimates represent an upper bound derived from optimized and continuous plasma conditions. In natural environments, lightning is highly transient and spatially heterogeneous, and atmospheric Hg²⁺ primarily exists in particulate-bound forms rather than in the well-mixed aqueous system employed here, all of which will strongly affect the actual fluxes⁶³. Nonetheless, plasma-induced reduction generates exceptionally high E¹⁹⁹Hg signatures (up to +31.5‰), far exceeding and opposite in sign to those of typical photochemical reduction (down to −2.75‰)¹². Consequently, even a minor contribution of plasma-induced Hg²⁺ reduction, whether driven by tropical lightning or auroral forcing, could measurably influence atmospheric Hg isotope signatures. Such effects may therefore necessitate modification of current global models, many of which rely on artificially adjusted source signatures (Fig. S8).

It should be acknowledged that no direct field evidence of plasma-induced Hg-MIF has been observed. This may reflect the intrinsic complexity of natural lightning processes, which likely involve both the reduction of Hg²⁺ and the oxidation of Hg⁰, leading to the partial re-oxidation of newly produced Hg⁰. Moreover, the highly transient nature of lightning channels, their spatial heterogeneity, and subsequent atmospheric mixing further hinder the preservation of distinct isotopic signatures. These factors likely explain why plasma-induced anomalies remain elusive despite their mechanistic plausibility demonstrated here. Nevertheless, related plasma-induced isotope effects have been documented in other systems. Laboratory simulations of dust-driven electrostatic discharges on Mars demonstrate that low-temperature plasma can induce large kinetic isotope effects across multiple element systems^64,65. Similarly, sulfur MIF observed in volcanic sulfate from large eruptions⁶⁶ has traditionally been attributed to stratospheric photochemistry, although plasma reactions associated with volcanic lightning may also induce similar isotope anomalies. These observations collectively support the broader possibility that plasma processes may represent an underappreciated source of MIF in natural environments.

Overall, our plasma-liquid reaction system offers a unique and versatile platform for investigating the mechanism of Hg-MIF. Unlike previous gas-phase plasma MIF studies (e.g., O, Mg, and Ti)^29,30, our setup operates at a plasma-liquid interface, allowing direct separation of Hg⁰ products from Hg²⁺ reactants, enabling reproducible isotopic measurements and controlled mechanistic tests. Importantly, it also more closely mimics natural lightning-driven processes at atmospheric interfaces. It can be extended to MIF studies of other elements (e.g., S, Cd, Pb, and Zn), thereby broadening plasma-driven isotope geochemistry.

Methods

Experimental setup for plasma-induced Hg reduction

It is known that the accumulation of positive and negative charges within numerous cloud particles induces the formation of a strong electric field culminating in lightning strikes^26,27. Inspired by this natural phenomenon, our work utilizes a DC glow-discharge plasma reactor to investigate the unique isotopic signatures of plasma-induced reduction processes, thereby facilitating a comparative analysis of our distinct fingerprint with that of actual lightning. The plasma-induced Hg²⁺ reduction is achieved at the plasma-liquid interface using the apparatus shown in Fig. 1a, which consists of a homemade SAGD plasma reactor and a trapping unit. The Hg²⁺ working solutions (100 μg/L) were prepared by diluting the GBW 08617 Hg standard (containing ~3% HNO₃) with ultrapure water (18.2 MΩ·cm). The pH was adjusted using double-purified HNO₃ and ultrapure water as required for each experimental run. The reactor, with a working volume of approximately 100 mL, employs a Pt wire immersed in the solution as the anode, while a microplasma serves as the cathode. The microplasma was initiated by temporarily reducing the distance between the tip of a tapered tungsten electrode (outer diameter, 2 mm; length, 10 cm) and the solution surface (100 μg/L Hg). The gap, or height of the discharge, was set to about 3 mm in the present study. A 5-kΩ resistor inserted between the Pt anode and the positive output of the high-voltage DC power supply (HB-Z202-200AC, Tianjin Hengbo High Voltage Power Supply Plant, China) was used to stabilize the discharge current. As shown in Fig. 1a, argon gas (100 mL/min, Ar₁) was continuously supplied as both the discharge gas and the carrier gas, transporting the Hg⁰ produced by the plasma reaction into the trapping solution (inverse aqua regia, 40% (v/v) HNO₃:HCl=2:1) for collection⁶⁷. After a certain time interval, the plasma was turned off and the system was purged with Ar (Ar₂, Fig. 1a) for 10 mins. Subsequently, 4 mL of reaction solution was withdrawn from the reactor and immediately mixed with 10 mL of trapping solution, and then kept in the freezer until isotopic analysis. These operations ensured that both the trapped Hg and the residual Hg²⁺ in the reactor had identical Hg speciation and matrix composition, thereby enabling reliable isotope measurements. Simultaneously, the trapping solution was also renewed at the time of subsampling and collected to determine the evolution of Hg isotope ratios in both reactants and products of plasma-induced reduction. In the simulated air discharge experiments, a stable plasma discharge cannot be sustained for extended periods in pure ambient air. Therefore, these experiments were performed at pH 4.0 and a current of 40 mA using a 1:1 Ar–air mixed gas atmosphere. For ease of comparison with previously reported photochemical studies, we converted this instantaneous mode into a cumulative mode during data processing. Specifically, the isotope composition of the products obtained in our experiments was recalculated to represent the corresponding values under the cumulative mode. Each experiment lasted 3 to 5 hours, depending on the reaction rate under different conditions.

All acids used in this study were purified by double sub-boiling distillation using Savillex’s DST-1000 sub-distillation system. Ultrapure water (18.2 MΩ cm, Millipore Company, America) was used for the preparation of solutions. Commercially available NaNO₂ and acetone were purchased from Sinopharm Chemical Reagent Co., Ltd. High-purity argon (99.999%) and compressed air were obtained from Wuhan Iron and Steel Group Oxygen Company, Wuhan, China. Hydrogen gas was supplied by a hydrogen generator (model TP-3030C, Beijing Beifen Tianpu Instrument Technology Co., Ltd., Beijing, China).

Mercury concentration and isotope composition measurements

Total Hg concentrations were measured using atomic fluorescence spectrometry (AFS 9700, Beijing Haiguang Co., China). Mercury isotope analysis was performed on a Nu Plasma II MC-ICPMS instrument (Nu Instruments, UK) at the State Key Laboratory of Geomicrobiology and Environmental Change. Six Hg isotopes (¹⁹⁸Hg, ¹⁹⁹Hg, ²⁰⁰Hg, ²⁰¹Hg, ²⁰²Hg, and ²⁰⁴Hg) and two Tl isotopes (²⁰³Tl, ²⁰⁵Tl) were measured simultaneously. A cold vapor generation (CVG) system were used for Hg sample introduction, with a Hg concentration of 5 μg/L. For mass bias correction, a 60 μg/L Tl standard solution was introduced via pneumatic nebulization. The sample introduction flow rate was set to 1.5 mL/min, yielding a signal intensity of ~1.5 V for ²⁰²Hg. Mass bias was corrected using both internal standardization based on the measured ²⁰⁵Tl/²⁰³Tl ratio and the sample-standard bracketing (SSB) method with the standard. In this study, all final Hg isotope compositions are reported in δ notation (permil, ‰) relative to the initial reactant (the GBW 08617 Hg standard), which is measured before and after each sample, as defined by:

$${\mathrm{\delta }}^{\mathrm{xxx}}\mathrm{Hg}=\left\{\frac{{\left(\genfrac{}{}{0ex}{}{\text{xxx}}{}\text{Hg}/\genfrac{}{}{0ex}{}{198}{}\text{Hg}\right)}_{\mathrm{sample}}}{{\left(\genfrac{}{}{0ex}{}{\text{xxx}}{}\text{Hg}/\genfrac{}{}{0ex}{}{198}{}\text{Hg}\right)}_{\mathrm{standard}}}-1\right\}\times 1000$$

where xxx is the mass of each Hg isotope between 199 and 204. Mercury isotope MIF is reported in “capital delta” notation (Δ^xxxHg, ‰) and is defined as: Δ¹⁹⁹Hg = δ¹⁹⁹Hg-δ²⁰²Hg⁎0.2520, Δ²⁰⁰Hg = δ²⁰⁰Hg-δ²⁰²Hg⁎0.5024, Δ²⁰¹Hg = δ²⁰¹Hg-δ²⁰²Hg⁎0.7520, Δ²⁰⁴Hg = δ²⁰⁴Hg-δ²⁰²Hg⁎1.4930. The δ²⁰²Hg, Δ¹⁹⁹Hg, Δ²⁰⁰Hg, Δ²⁰¹Hg, and Δ²⁰⁴Hg values achieved a precision of 0.09‰ and 0.08‰, 0.09‰, 0.07‰, and 0.11‰ (2 SD, n = 11), respectively. Method reliability was verified using the standard material NIST SRM 1944. Relative to NIST 3133, we obtained δ¹⁹⁹Hg = −0.12 ± 0.04‰, δ²⁰⁰Hg = −0.24 ± 0.05‰, δ²⁰¹Hg = −0.38 ± 0.08‰, and δ²⁰²Hg = −0.47 ± 0.04‰, consistent with previously reported values⁶⁸.

Author contributions

Conceptualization, X.L. and Z.Z.; investigation, G.W., Y.W. and X.L.; methodology, X.L. and Z.Z.; data curation and analysis, G.W., X.L., W.Z., H.Z., P.X. and Z.Z.; supervision, X.L. and Z.Z.; project administration, X.L. and Z.Z.; resources, X.L., H.Z. and Z.Z.; writing—original draft, X.L. and Z.Z.; and writing—review and editing, all authors.

Peer review

Peer review information

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

Funding

This work was financially supported by the financial support from the Natural Science Foundation of Hubei Province (2025AFA048), National Natural Science Foundation of China (42003018, 22274145), the National Key R&D Program Youth Scientist Project (2022YFF0504600), and the Space Application System of China Manned Space Program (KJZ-YYNCL0602).

Data availability

The isotope data generated and analysed during this study have been deposited in the Figshare database under 10.6084/m9.figshare.32207079. The dataset is described in ref. ⁶⁹.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75079-5.