{"id":"ca51701f-2abc-42e9-86ed-efea83df0506","arxiv_id":"1908.02874","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":6,"one_line_summary":"In mixed-wet pore networks, incremental oil from low-salinity flooding can only come from oil-wet pores, and improving microscopic sweep in those pores is necessary but not sufficient for extra recovery.","lead":"This paper uses computer simulations of oil and brine moving through pore-scale networks to study why low-salinity water flooding sometimes recovers extra oil from rocks with mixed oil-wet and water-wet pores. It finds that the oil-wet pores are the only possible source of extra oil, and that triggering wettability changes there at the right time is critical.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (1)'s heuristic tracer time sets when OW pores are wettability-modified; the paper's positive-LSE mechanism and its Swi/Z sensitivity may be artifacts of that functional form.","rationale":"I read the paper as establishing a qualitative pore-scale mechanism, not a quantitative prediction. Within the model's displacement rules, the claim that WW pores are largely exhausted by imbibition and that incremental oil must come from originally OW pores is internally coherent and matches the occupancy plots. The necessary-but-not-sufficient statement about microscopic sweep efficiency in the OW pore fraction is also defensible given the assumption that LS only changes the strength, not the sign, of wetting. The reader's conditional verdict is well calibrated. My stress-test pass converges on the same weakest point: the tracer timing equation is the one place where a parameter-free mechanistic argument is replaced by a fitted heuristic. Every result that depends on when OW pores are modified — the contrast between post-breakthrough and secondary LS, the Swi and Z sensitivities, and the 4.9%/1.9%/1.6% incremental numbers — passes through Eq. (1). That does not make the paper wrong, but it makes the central positive-LSE result unverified in a way that only a direct sensitivity test or comparison with an unsteady-state model can settle. I therefore do not recommend changing the verdict; I recommend making the requested conditional acceptance depend on that test and on reporting ensemble statistics.","tokens_in":26788,"tokens_out":6159,"duration_ms":65629,"concrete_test":"Rerun the Section III.B secondary LS base case (MWL, α=0.5, Swi=0.12, Z=3.5) with Eq. (1) replaced by its lower bound TΔV=ΔV/Q and by its cap TΔV=TPV, and also with an alternative monotone heuristic such as TΔV = (ΔV/Q)·(ΔPold/ΔPnew). Keep all capillary displacement rules, Pc stepping and contact-angle updates fixed. If the 4.9% incremental oil, and the ranking of Swi=6/12/18% and Z=3.5/4.5 in Figures 9–10, are not qualitatively preserved across all three tracer timings, then Eq. (1) is the proximate cause of the paper's positive-LSE result rather than a neutral detail.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that OW pores are the only viable source of incremental oil is well supported by the displacement logic and does not depend on the tracer heuristic. The load-bearing issue is Eq. (1) in Section II.C.1: TΔV = (ΔV/Q)[ΔPold/(ΔPold−ΔPnew)], which the paper itself describes as 'chosen to reflect qualitative observations from experimental coreflooding studies' rather than derived. This time controls how much LS brine advances through spanning water clusters at each Pc step, and through Eq. (2) determines whether neighbouring oil-filled OW pores have their contact angles reduced before or during the drainage cycle. The entire positive-LSE mechanism in Section III.B is a delayed modification of OW pores: the connate HS brine and reduced Z must keep OW pores unmodified until drainage, so that sweeping the OW cluster becomes more efficient (62.2% vs 58.1% pores displaced). If a defensible transport calculation gave a shorter TΔV, the LS front would modify OW pores earlier and the 4.9% incremental oil could disappear; if it gave a longer TΔV, even the Swi=6% case might show an LSE. Thus the stated dependence of the LSE on Swi and Z, and the specific interpretation of secondary LS injection, rest on an unvalidated scalar factor in the tracer timing, not on the displacement physics alone. The omission of ensemble statistics and several 'results not shown' comparisons compounds this, because a single realization cannot reveal whether the timing effect is robust.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper extends a steady-state pore-network model, previously developed for uniformly wetted media, to simulate low-salinity (LS) waterflooding in heterogeneously wetted networks that contain spanning clusters of both water-wet (WW) and oil-wet (OW) pores. The model couples capillary-controlled displacement with a tracer algorithm that estimates the spatial evolution of brine salinity, and it applies a Heaviside-type rule for local contact-angle modification when salinity falls below a threshold. The main claims are: (i) in such networks the OW pores are the only viable source of incremental oil; (ii) a LS-induced increase in microscopic sweep efficiency in the OW pore fraction is a necessary but not sufficient condition for additional oil production; and (iii) the OW fraction alpha, network connectivity Zbar, and initial water saturation Swi are key controls on the magnitude of the low-salinity effect (LSE). The central reported result is a secondary-mode simulation that produces roughly 5% more oil with LS brine than with HS brine, attributed to delayed wettability modification of OW pores until the drainage cycle. Sensitivity runs for Swi and Zbar, plus unreported runs for alpha, are used to support the parameter-dependence conclusions.","tokens_in":27076,"tokens_out":7428,"duration_ms":84546,"significance":"If the results hold, the paper provides a falsifiable mechanistic explanation for why LS waterflooding may be ineffective or effective depending on the initial wettability distribution, and it highlights a specific pore-scale mechanism (delayed modification of OW pores during drainage) that could guide experimental coreflood design. A notable strength is that the central logical argument that OW pores are the only incremental oil source follows directly from the stated displacement rules and is not fitted to data; the paper also explicitly lists its idealizations and limitations. The contact-angle change is tied to an experimental measurement, and the tracer algorithm is described in enough detail to be re-implemented. However, the quantitative LSE magnitudes and the Swi/Z/alpha sensitivities rest on the heuristic timing in Eq. (1) and on several unreported simulation sets, so the significance is tempered by the need for validation and more complete reporting.","major_comments":[{"comment":"The tracer timing equation T_DeltaV = (DeltaV/Q)[DeltaP_old/(DeltaP_old - DeltaP_new)] is explicitly described as 'chosen to reflect qualitative observations' rather than derived from transport physics. This time controls how far the LS front advances before each pressure step, and via Eq. (2) it determines whether neighbouring OW pores are wettability-modified before or during drainage. The entire positive-LSE mechanism in Section III.B (delayed OW modification producing 4.9% incremental oil) and the reported Swi and Zbar sensitivities are contingent on this functional form. If a defensible transport calculation gave a shorter T_DeltaV, the LS front would modify OW pores earlier and the incremental oil could vanish; if it gave a longer T_DeltaV, even the Swi=6% case could show a larger LSE. I request either a derivation of Eq. (1) from a stated transport model, a validation against the unsteady-state model of Boujelben et al., or a sensitivity analysis showing that the qualitative conclusions are unchanged under alternative plausible forms for T_DeltaV.","section":"II.C.1, Eq. (1)"},{"comment":"All reported results appear to come from a single network realization for each parameter set. Pore-network simulations with randomly assigned radii, contact angles, and wettability are stochastic, and the claimed differences (62.2% vs. 58.1% pores displaced; 4.9% vs. 1.9% incremental oil for Swi=6%; 4.9% vs. 1.6% for Z=4.5) could be within realization-to-realization variability. No error bars, ensemble averages, or statistical tests are provided. The paper should either report statistics over multiple independent realizations (even a modest number) or explicitly state that the quantitative trends are single-realization observations; without this, the quantitative sensitivity claims in the abstract and conclusions are not established.","section":"III.B, Figures 6-10"},{"comment":"Several load-bearing supporting results are reported only as 'results not shown' or with no explicit simulation output. In particular, the alpha trend in Section IV is given as 2.5% incremental oil for alpha=0.465 and 9.1% for alpha=0.535, and this trend is used in the conclusions to assert that alpha is a critical control on the LSE, but no figure, simulation details, or error information are provided. Similarly, the claims that FW and MWS networks show 'little or no additional oil recovery' (Section III.A) and that delayed LS injection supports the proposed theory (Section III.A) are unverifiable. Given that these results are central to the stated parameter-dependence conclusions, the authors should either present the supporting simulations or explicitly downgrade these statements to preliminary hypotheses.","section":"IV and III.A"},{"comment":"The abstract and conclusions state that increased microscopic sweep efficiency in the OW pores is necessary but not sufficient to guarantee additional oil production. The 'necessary' part is well supported by the displacement logic and by the reported simulations. The 'not sufficient' part, however, rests on schematic arguments in Fig. 11 and on 'observations from an extensive model sensitivity analysis' for which no explicit numerical results are reported, including cases where OW pores become WW or WW pores become OW. Because this statement appears as a central, abstract-level claim, the authors should provide at least one documented simulation showing increased microscopic sweep efficiency without a net increase in oil production, or explicitly rephrase the claim as a conjecture from the authors' unpublished sensitivity work rather than a demonstrated result.","section":"IV, Fig. 11"}],"minor_comments":[{"comment":"Equation (1) is undefined when DeltaP_old equals DeltaP_new, and the paper does not specify what happens if the pressure drop increases or stays constant between steps. A short comment on the domain of validity would prevent ambiguity in re-implementation.","section":"II.C.1, Eq. (1)"},{"comment":"The Heaviside modification rule treats C_N < C* and C_N > C* but does not specify the boundary case C_N = C*. Since this equality determines whether a pore is modified, the strict inequality should be stated explicitly.","section":"II.D, Eq. (2)"},{"comment":"The term 'early tertiary' is used for LS injection commencing at water breakthrough following HS injection; in conventional terminology, tertiary flooding typically follows a secondary flood to near its economic limit. The authors should clarify this definition to avoid confusion with standard experimental protocols.","section":"III.A"},{"comment":"The pore-size occupancy histograms would be easier to interpret if the y-axis were normalized to the total number of pores in each bin or if the bin totals were stated, since the raw counts differ between the two network configurations and make cross-figure comparison difficult.","section":"Figures 5 and 7"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the journal's scope and the central mechanistic insight is valuable, but the revision must address the heuristic basis of Eq. (1), the absence of ensemble statistics, and the reliance on 'results not shown' for several statements presented as conclusions. I would support publication after these points are resolved; if the authors can only provide single-realization results and cannot validate or bound the tracer timing, the quantitative LSE magnitudes should be substantially softened. The manuscript would also benefit from a data/code availability statement for reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know two things before reading. First, the central claim is genuinely new and defensible: in networks with spanning clusters of both water-wet and oil-wet pores, the oil-wet pores are the only viable source of incremental oil, and a low-salinity-induced increase in microscopic sweep efficiency in that fraction is necessary but not sufficient. That claim follows directly from the capillary displacement rules, not from fitting to data, so there is no circularity in the strong sense. Second, the quantitative side is softer. The magnitude of the low-salinity effect and its dependences on initial water saturation and connectivity are controlled by an openly heuristic tracer-timing factor, Eq. (1), and a few supporting simulations are only described as \"results not shown.\"\n\nWhat the paper does well: it extends the authors' own uniformly wetted steady-state model to fractionally wet, mixed-wet-large, and mixed-wet-small networks, and it gives a clean mechanistic story for why post-breakthrough low-salinity injection often fails while secondary injection can work. The distinction between pore-sequence effects and microscopic sweep efficiency is useful, and the secondary-mode result—connate high-salinity brine delays wettability modification so that more oil-wet pores are swept during drainage—is a plausible explanation for the patchy experimental literature. The paper is honest about its qualitative scope and about the model's simplifications. The self-citations are appropriate because the framework is carried over from prior work; the choice of Δθ = 20° is at least consistent with the Khishvand et al. contact-angle measurements.\n\nThe main soft spot is Eq. (1). The paper says that formula was \"chosen to reflect qualitative observations from experimental coreflooding studies,\" not derived. That timing controls when oil-wet pores see low-salinity brine and hence whether their contact angles are modified before or during the drainage cycle. If a defensible transport calculation gave a shorter residence time, the 4.9% incremental oil could shrink or disappear; if longer, even the low-Swi case might show an effect. So the sensitivity to Swi and Z is real within the model, but the specific magnitudes rest on an unvalidated scalar factor. This is a moderate concern, not a fatal one, because the central conclusion about where incremental oil must come from does not depend on that factor.\n\nI also count at least three places where quantitative support is asserted rather than shown: the claimed equivalence of FW/MWS networks in the post-breakthrough section, the delayed-injection simulations, and the alpha sensitivity in the Discussion (where numbers 2.5% and 9.1% are given but no curves or realizations). There are no error bars or ensemble averages, so single-realization effects cannot be separated from trends. The paper's own limitation statements are candid, which helps, but they don't fix the reproducibility gap.\n\nWho is this for? Pore-network modelers and EOR experimentalists who want a mechanistic way to interpret inconsistent low-salinity corefloods. I would send it to peer review rather than desk-reject, and I would make the revision ask concrete: release the code and data, or at minimum show the missing simulations and rerun the key cases over an ensemble of network realizations. I would also ask for a sensitivity analysis of Eq. (1) itself, or a derivation from a more transparent transport model. If the authors can satisfy that, the paper would be a solid contribution.","headline":"A credible steady-state pore-network study whose central mechanistic claim holds up, but whose quantitative LSE magnitudes rest on an openly heuristic tracer-timing equation and several 'results not shown' claims.","tokens_in":27659,"tokens_out":2194,"would_cite":true,"duration_ms":27831,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["47.56.+r"],"model":"deepseek-v4-flash","headline":"In mixed-wet porous media, low-salinity waterflooding adds oil only from the oil-wet pore fraction, and only when it increases the sweep of those pores.","keywords":["low-salinity waterflooding","pore network model","wettability alteration","mixed-wet porous media","enhanced oil recovery","tracer algorithm","capillary displacement","percolation"],"falsifier":"Time-resolved micro-CT or micromodel experiments on a mixed-wet rock that image both the salinity front and oil-water contact angles while oil is displaced would settle the core claim: if extra oil appears from pores that were water-wet before the flood, or if oil-wet pores yield incremental oil without an increase in the fraction of oil-wet pores invaded, the mechanism fails. A model-level check is to replace Eq. (1) with a full unsteady-state front calculation and observe whether the predicted connectivity trend and the non-monotonic $S_{wi}$ dependence survive.","tokens_in":26476,"feed_emoji":"🛢️","tokens_out":11125,"duration_ms":120831,"temperature":0.7,"pith_summary":"The paper asks when low-salinity (LS) waterflooding, a method that often improves oil recovery but sometimes fails, can work in rocks containing both water-wet (WW) and oil-wet (OW) pores. It uses a 3D steady-state pore network model coupled to a tracer algorithm that follows injected LS brine as it mixes with resident high-salinity (HS) brine. The central claim is that when both wettability classes form connected paths through the network, the WW pores are already swept by ordinary imbibition, so LS brine can only add oil from the OW fraction. Extra oil appears only when LS-induced contact-angle reduction increases the microscopic sweep efficiency of the OW pores; that condition is necessary but not sufficient. The simulations identify the OW fraction, the network connectivity, and the initial HS-brine saturation as the controls, and they reproduce positive, neutral, and negative outcomes.","feed_headline":"Low-salinity oil gains come only from oil-wet pores","feed_subtitle":"Pore-network simulations show when those pores are swept late in the flood extra oil appears—and sometimes not even then.","key_machinery":"The machinery is a capillary-dominated steady-state displacement model on a 3D cubic pore network, augmented by a tracer algorithm that estimates salinity after each saturation step. Each pore carries a radius $R$ and a contact angle $\\theta$, and water invades in order of capillary entry pressure $2\\sigma\\cos\\theta/R$. After every step the algorithm rewinds the flow and uses Eq. (1) to estimate how long the step took, advects a salinity tracer through connected water, and applies a Heaviside rule (Eq. (2)): if neighbouring water salinity falls below a critical value $C^*$, the contact angle of an oil-occupied pore is reduced by $\\Delta\\theta$. This dynamic contact-angle map changes the sequence of pore filling, and the timing of the contact-angle reduction relative to the drainage half of the flood is what determines whether microscopic sweep efficiency improves.","core_discovery":"The paper's central claim is that in a network where both water-wet and oil-wet pores form spanning clusters before flooding, the oil-wet pores are the only viable source of incremental oil during low-salinity injection. Making water-wet pores more water-wet does not help, because those pores are already invaded by imbibition in the corresponding high-salinity flood; the two simulations converge to nearly the same water occupancy in the WW fraction. In the OW fraction, the standard drainage sequence fills pores from largest to smallest, so the only route to extra oil is to invade a larger fraction of the OW pores, the 'microscopic sweep efficiency effect'. A necessary but not sufficient condition is that OW contact angles are reduced during the drainage cycle rather than before it; whether this happens is set by how long connate HS brine protects OW pores from freshening. Low network connectivity and an intermediate initial HS-brine saturation delay the salinity front, while extreme $S_{wi}$ values weaken or erase the gain, and the fraction $\\alpha$ of OW pores sets the ceiling on what can be recovered.","pith_inferences":["As an extension of the paper's logic, replacing the heuristic step-time formula with a fully dynamic flow simulation would test whether the predicted $S_{wi}$ and connectivity trends survive in degree or only in direction.","A practical screening inference follows: low-salinity candidates should be characterised by connate-water saturation and pore connectivity, not only by oil-wet fraction, because those hidden variables can flip a pilot between positive and neutral.","The delay-until-drainage principle suggests a design rule the paper does not simulate: a short high-salinity pre-flush or a viscosity-moderated front could hold the salinity boundary back and enlarge the incremental-oil window.","Because the geometric constraint is general, the same two-percolating-cluster argument may apply to other wettability-altering enhanced-oil-recovery agents, not just low-salinity brine."],"forward_implications":["In a mixed-wet reservoir where both wettabilities form through-going clusters, the water-wet pores cannot yield incremental oil to a low-salinity flood; the upper bound on the gain is fixed by the oil-wet fraction.","A low-salinity flood can weaken oil-wet contact angles and still produce no extra oil, because the required increase in oil-wet sweep efficiency is necessary but not sufficient.","Poorly connected pore networks should show larger low-salinity benefits, since circuitous flow delays the arrival of fresh brine at oil-wet pores.","The initial connate-brine saturation acts nonlinearly: too little lets the front freshen too early, too much causes early breakthrough that flushes the protective HS brine out.","The specific wettability arrangement (mixed-wet large, mixed-wet small, or fractionally wet) becomes decisive only when LS brine changes the wetting class of pores, not when it merely weakens or strengthens the existing wettability."],"supporting_citations":[{"why":"This prior study supplies the steady-state LS waterflooding model and the pore-sequence and sweep-efficiency vocabulary that the paper extends to mixed-wet networks.","marker":"[37]"},{"why":"The unsteady-state counterpart provides the contrasting injection-timing picture used to justify the tracer algorithm's role in freshening oil-wet pores.","marker":"[38]"},{"why":"Experimental micro-CT evidence that LS brine lowers oil-water contact angles is the direct support for the model's uniform contact-angle reduction of 20 degrees.","marker":"[32]"},{"why":"The pore-volume and conductance scaling laws in this reference determine how realistic the initial HS-brine saturation and connectivity sensitivity runs are.","marker":"[41]"},{"why":"This pore-scale mechanism proposal motivates the modelling choice of representing the low-salinity effect as direct wettability alteration rather than explicit chemistry.","marker":"[36]"},{"why":"The review of experimental corefloods showing that secondary LS injection usually outperforms HS injection drives the paper's turn to secondary-mode simulations.","marker":"[45]"},{"why":"This coreflood comparison linking connate brine to LS recovery provides the experimental benchmark for the paper's $S_{wi}$ sensitivity findings.","marker":"[18]"}],"fun_headline_variants":["Only oil-wet pores produce low-salinity oil gains","Low-salinity extra oil demands oil-wet pore sweep","Low-salinity oil? Only oil-wet pores deliver","Low-salinity recovery limited by oil-wet pore fraction"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The simulations assume the heuristic time estimate in Eq. (1) correctly predicts how long each saturation step takes, and that timing decides whether oil-wet pores have already been freshened before the drainage cycle begins; if the timing is wrong, the claimed roles of initial brine saturation and connectivity, and the specific incremental-oil figures, would be artifacts of that assumption.","fun_headline_variants_meta":{"raw":{"variants":["Only oil-wet pores produce low-salinity oil gains","Low-salinity extra oil demands oil-wet pore sweep","Low-salinity oil? Only oil-wet pores deliver","Low-salinity recovery limited by oil-wet pore fraction"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000462,"raw_usage":{"total_tokens":2394,"prompt_tokens":1115,"completion_tokens":1279,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":731,"completion_tokens_details":{"reasoning_tokens":1212}},"tokens_in":731,"tokens_out":1279,"duration_ms":12845,"temperature":1.0,"reasoning_tokens":1212,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:31:29.097210+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Time-resolved micro-CT or micromodel experiments on a mixed-wet rock that image both the salinity front and oil-water contact angles while oil is displaced would settle the core claim: if extra oil appears from pores that were water-wet before the flood, or if oil-wet pores yield incremental oil without an increase in the fraction of oil-wet pores invaded, the mechanism fails. A model-level check is to replace Eq. (1) with a full unsteady-state front calculation and observe whether the predicted connectivity trend and the non-monotonic $S_{wi}$ dependence survive.","supporting_citations":[{"cited_title":"Sorbie and I","cited_arxiv_id":null,"evidence_quote":"This prior study supplies the steady-state LS waterflooding model and the pore-sequence and sweep-efficiency vocabulary that the paper extends to mixed-wet networks."},{"cited_title":"Watson, I","cited_arxiv_id":null,"evidence_quote":"The unsteady-state counterpart provides the contrasting injection-timing picture used to justify the tracer algorithm's role in freshening oil-wet pores."},{"cited_title":"Nasralla, M","cited_arxiv_id":null,"evidence_quote":"Experimental micro-CT evidence that LS brine lowers oil-water contact angles is the direct support for the model's uniform contact-angle reduction of 20 degrees."},{"cited_title":"Shaker Shiran and A","cited_arxiv_id":null,"evidence_quote":"The pore-volume and conductance scaling laws in this reference determine how realistic the initial HS-brine saturation and connectivity sensitivity runs are."},{"cited_title":"Blunt, Multiphase flow in permeable media — a pore-scale perspective (Cambridge Univer- sity Press, Cambridge, 2017)","cited_arxiv_id":null,"evidence_quote":"This pore-scale mechanism proposal motivates the modelling choice of representing the low-salinity effect as direct wettability alteration rather than explicit chemistry."},{"cited_title":"Dixit, S","cited_arxiv_id":null,"evidence_quote":"The review of experimental corefloods showing that secondary LS injection usually outperforms HS injection drives the paper's turn to secondary-mode simulations."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"This coreflood comparison linking connate brine to LS recovery provides the experimental benchmark for the paper's $S_{wi}$ sensitivity findings."}],"review_version":1}