{"id":"2eda6514-9616-4562-bb7a-97fc53385ddd","arxiv_id":"2608.00039","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":7,"one_line_summary":"Bridgmanite and post-perovskite hold up to 0.55 wt% and 1.22 wt% water respectively at deep lower-mantle conditions, implying subducted slabs do not dehydrate at the core-mantle boundary and supporting a deep low-D/H water reservoir.","lead":"Experiments squeezing wet mantle rock to pressures of the core-mantle boundary show the minerals bridgmanite and post-perovskite can absorb far more water than previously thought, up to 1.22 weight percent in post-perovskite. This challenges the idea that sinking tectonic plates release their water at the bottom of the mantle and points to a hidden deep water reservoir that may explain puzzling isotope signatures in volcanic plumes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The D/H fractionation factor α may be a SIMS matrix-effect artifact; the Methods' argument against matrix effects is incomplete, and the lattice-residence and deep low-D/H conclusions depend on it.","rationale":"The reader's weakest_assumption identified the same load-bearing concern: the D/H fractionation measurements rely on an uncalibrated assumption about SIMS matrix effects between Bdg/PPv and silicate melt. Close reading of the Methods confirms a logical gap: the spectral argument that 'differences in the relative sensitivity coefficient between hydrogen and deuterium do not affect α' only holds if that coefficient is phase-independent, and the paper provides no evidence for that. The 'identical ion energy' point is red herring—ion energy uniformity does not equal matrix-effect uniformity. This is the single most load-bearing issue because the D/H difference is used twice: first to demonstrate that the measured water is in the lattice (ruling out melt inclusions), and second to predict a deep low-D/H reservoir. If α is an artifact, both pillars weaken, even though the raw solubility values may still be reproducible. A concrete test—varying SIMS offset voltage or primary beam and observing whether α changes—can arbitrate this. The reader's verdict of CONDITIONAL is appropriate; no change is needed. No other concern (e.g., pressure/temperature uncertainties, post-hoc fit selections) is as directly decisive for the paper's headline implications.","tokens_in":22534,"tokens_out":6183,"duration_ms":67200,"concrete_test":"Re-analyze one recovered sample (e.g., run #5 or #9) by SIMS on the same Bdg and melt regions at multiple offset voltages (e.g., -50 V, -100 V, -150 V) with the energy slit fixed, and optionally with a different primary beam (e.g., Cs+). If the apparent α = (D/H)_Bdg/(D/H)_melt stays constant within 1σ across these conditions, the matrix-effect concern is largely resolved. If α changes by more than ~0.05 (5%) across the offset range, the claim that α is free of matrix effects is falsified, because any genuine isotopic fractionation must be independent of instrumental settings. A complementary test would be to synthesize a hydrous Bdg/PPv standard with known D/H via large-volume press, measure its bulk D/H, and calibrate the relative sensitivity factor R against a glass of the same nominal water content.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central interpretive claims—that water in Bdg/PPv resides in the crystal lattice (not in melt inclusions) and that a deep low-D/H reservoir exists—both rest on the measured hydrogen isotopic fractionation α = (D/H)_Bdg/(D/H)_melt = 1.15–1.30. In the Methods section 'Matrix effect in hydrogen isotope analysis by SIMS', the authors argue that α is free of matrix effects because (i) simultaneous SCAPS imaging ensures identical ion energy for Bdg and melt, and (ii) α = (D_Bdg/D_melt)/(H_Bdg/H_melt) cancels relative H/D sensitivity. However, (ii) is only valid if the relative sensitivity factor R = (D/H)_measured/(D/H)_true is identical for Bdg and melt. No such equality is demonstrated. R depends on ionization efficiency, which varies with matrix composition and crystal structure; identical ion energy does not make R equal across different phases. Their D/H calibration was performed on a single rhyolitic glass standard, not on Bdg/PPv. If the true α is 1, the observed 150–300‰ apparent fractionation could be entirely a matrix artifact. This would invalidate the claim that Bdg/PPv water is lattice-bound and distinct from melt, and with it the deep low-D/H reservoir and plume-source explanations. The absolute water solubility numbers (e.g., PPv >1 wt%) would remain as observations, but their interpretation as true lattice solubility would lose a key supporting argument.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports laser-heated diamond-anvil cell melting experiments on a hydrous pyrolite composition across 36–141 GPa and 3,340–4,860 K, using synchrotron XRD to identify liquidus phases and Cryo-SIMS to measure H2O/D2O abundances and D/H ratios in coexisting Bdg/PPv and melt. It reports high water solubilities in bridgmanite (up to 5,510 ppm) and post-perovskite (up to 1.22 wt%), a positive pressure dependence of the Bdg/melt H2O partition coefficient (Eq. 1), and D/H fractionation between crystals and melt (α = 1.15–1.30, Eq. 3). From these data the paper argues that subducting slabs do not release water at the CMB, and that hydrous Bdg/PPv cumulates in LLSVPs/ULVZs could constitute a deep low-D/H reservoir that may explain negative δD values reported in some plume-derived lavas.","tokens_in":23018,"tokens_out":8588,"duration_ms":73434,"significance":"If the solubility measurements hold, this is a field-changing result: it challenges the prevailing view that Bdg and PPv are nearly dry at deep lower-mantle conditions and implies that the lowermost mantle can store water at the wt% level, with direct consequences for slab dehydration, ULVZ/LLSVP origins, and the deep volatile cycle. The cryogenic sample handling, two-dimensional SIMS imaging of H and D in crystal layers away from melt pools, and the explicit reporting of raw data are clear strengths. However, the absolute concentration calibration and the D/H fractionation factor rest on assumptions that are not fully validated, and the BMO/geochemical 'predictions' are scenario calculations that depend on the paper's own fitted parameters. The solubility data and the geochemical narrative should be evaluated separately.","major_comments":[{"comment":"Absolute H2O and D2O concentrations in Bdg/PPv are derived from a calibration curve (1H+/28Si+ versus H/Si mass ratio) built on three silicate glasses (Extended Data Fig. 8a). No Bdg/PPv standard of known water content is used. The relative sensitivity factor for H/Si in a (Mg,Fe,Al)SiO3 perovskite or post-perovskite matrix may differ from that in rhyolitic glass; if so, the reported solubilities (e.g., 1.22 wt% in PPv) could be systematically biased. At minimum, the authors should justify the transfer of the glass calibration to crystalline Bdg/PPv, or provide a cross-check with an independent method (e.g., FTIR or atom-probe quantification).","section":"Methods: SIMS analyses"},{"comment":"The argument that α = (D/H)Bdg/(D/H)melt is free of matrix effects is incomplete. Cancellation of the D/H relative sensitivity requires the relative sensitivity factor R = (D/H)measured/(D/H)true to be identical in Bdg and melt; identical ion energy does not guarantee this. The calibration was performed on a single rhyolitic glass, not on Bdg/PPv. Therefore α = 1.15–1.30 may be an artifact of differing ionization efficiencies between crystalline silicate and melt. This affects the lattice-residence argument (main text, 'The difference in the D/H ratio...') and all deuterium modeling in Fig. 4. A crystalline Bdg/PPv D/H standard with independently known D/H is needed.","section":"Methods: Matrix effect in hydrogen isotope analysis by SIMS"},{"comment":"The deep low-D/H reservoir story is not an independent prediction: the δD values in Fig. 4 are computed using the paper's own fitted α (Eq. 3) and D(H2O) (Eq. 1), combined with assumed model choices (F=0.5, Bdg:Fp=6:4, PPv-in at ~114 GPa, and an initial H2O of 1,075 ppm derived by mass balance with an assumed complement reservoir). The apparent match to Koolau/Baffin Island δD values is therefore a consequence of the assumptions, not a test. The authors should provide a sensitivity analysis to these choices or reframe the results as a scenario calculation rather than a prediction.","section":"Crystallization of hydrous Bdg and PPv in a basal magma ocean; Fig. 4"},{"comment":"The conclusion that subducting slabs do not release water at the CMB extrapolates the experiments to conditions not covered by the data: the present runs are at 3,340–4,860 K with 2–3 wt% Al2O3 in Bdg/PPv, whereas slabs are cold and Al-poor. The estimate of ~1,000 ppm H2O transported, and the assumed Al3+:H+ = 1:1 capacity limit for Al-poor Bdg, are reasonable but unverified. This should be stated explicitly as an extrapolation; the abstract's phrasing ('subducting slabs do not release water') is stronger than the evidence supports.","section":"Fate of water transported by subducting slabs into the deep lower mantle"}],"minor_comments":[{"comment":"The caption notes that low-temperature uncertainty may be larger than illustrated, yet the main text uses Eq. (3) at 2,000 K (α = 1.7) to give 'as low as δD = –400 to –510‰' without propagating this uncertainty. Please quantify or soften the statement.","section":"Fig. 3 caption / main text"},{"comment":"Typo: 'Extended Datra Fig. 3' should read 'Extended Data Fig. 3'.","section":"Methods, 'Electron microprobe analyses'"},{"comment":"The notation in Eq. (5) is hard to follow: the integration limits and the meaning of m are not fully defined. Please rewrite with explicit definitions.","section":"Methods, 'Modelling the evolution of a basal magma ocean', Eq. (5)"},{"comment":"Many melt D2O cells are blank (e.g., runs 2, 4–8, 10–13). Please add placeholders or a note explaining which samples were excluded from D2O analysis and why, to avoid the appearance of missing data.","section":"Extended Data Table 1"},{"comment":"The phrase 'subducting slabs do not release water when they reach the bottom of the mantle' is a strong conclusion. Consider 'are unlikely to release water' or 'may retain water' to match the extrapolative nature of the argument.","section":"Abstract / main text"}],"recommendation":"major_revision","confidential_remarks":"This is a potentially high-impact paper, but the central solubility numbers and the D/H fractionation factor need independent validation. I recommend that the editor seek a reviewer with SIMS expertise to assess the matrix-effect issues, particularly the use of glass standards for crystalline Bdg/PPv and the D/H calibration. The BMO modeling is clearly presented but should be treated as a scenario calculation rather than a prediction; the authors should provide sensitivity analyses. The solubility data and the geochemical narrative should be separated in revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. The solubility numbers are the real news: Bdg water up to 5,510 ppm at 36–117 GPa and PPv up to 1.22 wt% at 117–141 GPa, with a positive pressure dependence that fits the Lu et al. data. That resolves a long controversy and is a big deal. The D/H fractionation is the soft spot: α = 1.15–1.30 between Bdg and melt is presented as the key evidence that water is in the lattice and as the basis for the deep low-D/H reservoir, but the methods argument against matrix effects is incomplete. Identical ion energy does not equal identical relative sensitivity for H vs D across phases, and they calibrate on a rhyolitic glass, not Bdg/PPv. So the observed 150–300‰ enrichment could be a matrix artifact. The stress-test note lands.\n\nThat said, the solubility data are carefully acquired: cryo-SIMS preserves H, images show H in crystal layers away from melt, and the partition coefficients are internally consistent. The claim that slabs don't dehydrate at the CMB does not depend on α; it only needs high Bdg/PPv capacity, which the direct measurements support. So the central water-storage conclusion survives even if α is wrong.\n\nThe BMO modeling is clearly an extrapolation: they feed their fitted D and α into a model with several assumptions (F = 0.5, Bdg:Fp ratio, PPv transition pressure, initial H2O). Calling the low-D/H reservoir a 'prediction' overstates it; it is a scenario consistent with the data, not a test. Also, α for PPv is never measured; run #9 only gives a PPv/Bdg comparison with large uncertainty.\n\nBottom line: this paper deserves a serious referee. The solubility dataset is important and will be cited regardless of the isotopic story. The referee should push for a matrix-matched D/H calibration or at least a quantitative bound on the matrix effect before the deep-reservoir claims are taken as established. I'd bring it to reading group and cite it in my own work.","headline":"New Bdg/PPv water solubility data to CMB conditions are a solid advance, but the D/H fractionation evidence is not yet quantitative and the deep low-D/H reservoir story leans on it.","tokens_in":23548,"tokens_out":2893,"would_cite":true,"duration_ms":35202,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Bridgmanite and post-perovskite can hold far more water than previously thought—up to 5,500 ppm and 1.2 wt% respectively—meaning subducting slabs carry water deep into the lower mantle.","keywords":["bridgmanite","post-perovskite","water solubility","lower mantle","core-mantle boundary","hydrogen isotopes","D/H fractionation","basal magma ocean"],"falsifier":"Synthesize a bridgmanite crystal with a known D/H ratio, measure it by the same Cryo-SIMS method alongside a melt glass of identical D/H; if the measured D/H ratio differs between the two phases, the reported fractionation is a matrix artifact rather than a real effect.","tokens_in":22459,"feed_emoji":"💧","tokens_out":7651,"duration_ms":65643,"temperature":0.7,"pith_summary":"This paper reports laser-heated diamond-anvil-cell melting experiments on a hydrous, natural mantle composition covering the entire pressure range of the lower mantle (36–141 GPa). Cryogenic secondary-ion mass spectrometry on the recovered samples shows that bridgmanite, the main lower-mantle mineral, dissolves up to 5,510 ppm water, and post-perovskite, the mineral of the lowermost mantle, dissolves more than 1 wt% (up to 1.22 wt%). The crystals also strongly prefer deuterium over hydrogen relative to coexisting melt, with a fractionation factor of 1.15–1.30, which the authors take as proof that the water sits in the mineral lattice rather than in melt inclusions. From this they argue that subducting slabs do not dehydrate when they reach the core-mantle boundary, and that water-rich bridgmanite and post-perovskite cumulates in large low-shear-velocity provinces and ultralow-velocity zones could hold a substantial, D-depleted deep water reservoir inherited from a basal magma ocean. If these findings hold, the lowermost mantle may be one of Earth's largest water reservoirs.","feed_headline":"Post-perovskite stores over 1% water at the core-mantle boundary","feed_subtitle":"Bridgmanite and post-perovskite can store 5,510 ppm and 1.22 wt% water, reshaping Earth's water budget.","key_machinery":"The argument is carried by the combination of laser-heated diamond-anvil-cell melting experiments (36–141 GPa, 3,340–4,860 K) with fully cryogenic sample recovery and Cryo-SIMS imaging of 1H+ and 2D+ secondary ions, which lets the authors measure water concentrations and D/H ratios in individual crystal and melt regions while avoiding hydrogen contamination and loss. The key physical quantity is the hydrogen isotope fractionation factor α=(D/H)_crystal/(D/H)_melt=1.15–1.30, which both proves the water is lattice-bound and (via the temperature dependence α≈1+a/T² with a=2.95×10^6) seeds the prediction of a deep, D-depleted water reservoir. Water incorporation is attributed mainly to the coupl","core_discovery":"The paper's central claim is that water solubility in the two dominant minerals of the deep lower mantle is far higher than previously recognized—bridgmanite hosts up to 5,510 ppm (H2O+D2O) and post-perovskite up to 1.22 wt% at core-mantle-boundary pressures and temperatures. The authors further report strong hydrogen isotope fractionation, α=(D/H)_Bdg/(D/H)_melt=1.15–1.30, with crystals enriched in deuterium relative to coexisting melt, which they use to show the water is truly in the crystal lattice. They conclude that dehydration of subducting slabs is not a viable source of melt or hydrogen at the base of the mantle, and that the large low-shear-velocity provinces and ultralow-velocity z","pith_inferences":["If deep mantle minerals really hold percent-level water, other volatile elements such as carbon or nitrogen may be similarly hosted in Bdg/PPv, implying a much larger deep-mantle volatile cycle than current models assume.","The strong D/H fractionation between crystals and melt means mantle convection and partial melting should produce measurable isotopic heterogeneity in the deep mantle—plume lavas with low D/H may be a tracer of unfractionated, ancient mantle rather than recycled surface water.","The measured positive pressure dependence of water solubility suggests that with slightly higher temperatures at the CMB, the capacity could exceed the observed values, making a fully hydrated D″ layer plausible if enough water is transported there.","A direct experimental extension would be to dope the starting material with variable Al2O3 to test whether PPv water capacity scales with Al content as the authors assume for natural pyrolite."],"forward_implications":["Subducting slabs that reach the core-mantle boundary will retain their water rather than release it; dehydration melting cannot explain ULVZ-like seismic structures or a hydrogen-rich layer at the top of the outer core.","Hydrous post-perovskite and bridgmanite forming in a crystallizing basal magma ocean can store roughly a third of the water present in the initial magma ocean, sequestering it in LLSVPs and ULVZs.","Water stabilizes post-perovskite relative to bridgmanite, so hydrous cumulates could explain post-perovskite-like seismic discontinuities inside LLSVPs.","Plumes entraining this material could deliver D-depleted water (δD as low as –400 to –510‰ at 2,000 K), matching negative D/H anomalies in some ocean island basalts.","Combined reservoirs imply the core may contain at least 0.3 wt% hydrogen."],"fun_headline_variants":["Post-perovskite absorbs over 1% water at Earth's core boundary","Water-rich mantle minerals prevent slab dehydration at core boundary","Post-perovskite holds 1.2wt% water, hints at deep-water reservoir"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The deep low-D/H reservoir story depends on the measured deuterium enrichment in bridgmanite and post-perovskite relative to coexisting melt being a real chemical effect, not an artifact of measuring two different materials (crystalline mineral versus glassy melt) with the same ion microprobe—no matrix-matched D/H standard is available to rule that out.","fun_headline_variants_meta":{"raw":{"variants":["Post-perovskite absorbs over 1% water at Earth's core boundary","Water-rich mantle minerals prevent slab dehydration at core boundary","Post-perovskite holds 1.2wt% water, hints at deep-water reservoir"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001119,"raw_usage":{"total_tokens":4578,"prompt_tokens":912,"completion_tokens":3666,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":656,"completion_tokens_details":{"reasoning_tokens":3603}},"tokens_in":656,"tokens_out":3666,"duration_ms":23505,"temperature":1.0,"reasoning_tokens":3603,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T01:35:59.869669+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Synthesize a bridgmanite crystal with a known D/H ratio, measure it by the same Cryo-SIMS method alongside a melt glass of identical D/H; if the measured D/H ratio differs between the two phases, the reported fractionation is a matrix artifact rather than a real effect.","supporting_citations":[],"review_version":1}