{"id":"6d5c8766-7ec6-4d5f-966a-9c7901f54507","arxiv_id":"2501.02296","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Higher-viscosity water-glycerol floods enhance tertiary oil recovery mainly by fragmenting trapped oil ganglia into smaller, transportable clusters, not by moving whole ganglia.","lead":"In microfluidic experiments, injecting increasingly viscous glycerol-water mixtures after a waterflood mobilizes trapped oil by breaking large oil ganglia into smaller droplets and blobs that can be transported out of the chip. The finding gives a pore-scale mechanism for tertiary enhanced oil recovery and points to how injection-fluid viscosity might be tuned to improve residual oil removal.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The quantitative 'ganglia number vs volume' claim rests on hand-set, non-exhaustive cluster classification rules with no sensitivity analysis; threshold shifts could alter the reported trend.","rationale":"The reader's weakest assumption identified the hand-set cluster classification thresholds as a load-bearing assumption. I agree that this is the key quantitative vulnerability, but I would sharpen it further: the classification rules are not merely threshold-sensitive; they are non-exhaustive, so a meaningful fraction of oil clusters may be excluded from all three categories. This makes the reported counts and volumes of ganglia, blobs, and droplets potentially incomplete in a way that is not addressed anywhere in the manuscript. The central claim depends on the comparison of ganglia number versus ganglia volume across ηR, and that comparison is defined by an arbitrary area cutoff. A breakup event that produces fragments smaller than 0.004 mm² automatically reduces the 'ganglia' volume without necessarily reducing the 'ganglia' number, so the observed pattern could be partly baked into the classification rather than discovered in the physics. However, the visual evidence in Fig. 9(d) and the accompanying video show genuine breakup of a large oil cluster into smaller fragments, so I do not think the qualitative mechanism should be rejected. The appropriate response is to require a sensitivity analysis and, ideally, release of the segmented images or raw data. Since the reader already issued a CONDITIONAL verdict, my read does not move that verdict; it strengthens the justification for the condition. I therefore mark verdict_should_be as UNCHANGED, meaning no change to the reader's CONDITIONAL recommendation.","tokens_in":19859,"tokens_out":4836,"duration_ms":47903,"concrete_test":"Recompute Figs. 9(a,b) and 11(b,c) from the same raw segmented images using a grid of alternative classification thresholds, for example area boundaries of 0.001, 0.002, 0.003, 0.005, and 0.008 mm² and circularity boundaries of 0.5 and 0.8, with all objects explicitly assigned to exactly one class by a stated priority rule. If, for every threshold within a factor of two of the stated values, the number of ganglia remains nearly flat in ηR while their total volume drops by the reported factor, the central claim survives. If the number trend or the 'slight reduction' statement flips at any such threshold, the quantitative claim is an artifact of the hand-set classification.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim is that with increasing viscosity ratio, the number of ganglia changes only slightly while their total volume decreases sharply, and this is interpreted as ganglia breakup feeding the medium with blobs and droplets. The categories in Section 3.2.2 are defined by hand-set intervals: droplets are 0-0.002 mm² with circularity 0.7-1, blobs are 0.002-0.004 mm² with circularity 0.3-1, and ganglia are >0.004 mm² with circularity 0-0.7. These rules are not exhaustive: an object with area 0.001 mm² and circularity 0.5, or an object with area 0.003 mm² and circularity 0.2, or a large object with area >0.004 mm² and circularity 0.8 falls into no category. The manuscript does not state how such objects were treated. More importantly, because 'ganglia' is defined by a lower area cutoff, any breakup that produces fragments just below 0.004 mm² automatically moves both number and volume out of the ganglia class. Thus the statement that the number of ganglia changes only slightly is not independent of the chosen cutoff; shifting the boundary by a factor of two could turn a 'slight change' into a substantial decrease or even reverse the reported trend. No sensitivity analysis, replicate counts for the tertiary experiments, or raw data are provided (the data availability statement says 'No data was used'), so the quantitative support for the central mechanism is weaker than the text implies. The direct image evidence in Fig. 9(d) does show breakup, so the qualitative mechanism remains plausible; the concern is specifically the quantitative cluster statistics that carry the central claim.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports microfluidic experiments on tertiary oil recovery: after drainage and waterflooding of a water-wet micromodel, glycerol/water mixtures of increasing viscosity are injected, and the mobilization of residual oil clusters is imaged and quantified. Oil clusters are classified into droplets, blobs, and ganglia using size and circularity thresholds, and the authors report that increasing the viscosity ratio mainly reduces the total volume of ganglia while their number changes only slightly, which they interpret as ganglia breakup feeding the medium with blobs and droplets that are subsequently transported. The paper also compares the results with literature on cluster-size scaling and constructs a capillary desaturation curve. The central qualitative claim is supported by direct time-lapse images of ganglia breakup at pore throats.","tokens_in":20201,"tokens_out":3055,"duration_ms":31127,"significance":"If the quantitative support were robust, this would be a valuable pore-scale demonstration that tertiary recovery by viscosity increase operates through a two-step mechanism: rapid ganglia breakage followed by slower transport of small clusters. The direct visual evidence of breakup at pore throats (Fig. 9(d), Fig. 10) and the careful control of fluid properties and flow conditions are genuine strengths. The paper also connects its observations to earlier mechanistic hypotheses and to the capillary-desaturation-curve framework. However, the quantitative claims about cluster number and volume trends rest on hand-set classification rules without sensitivity analysis and on experiments with no reported replicates or error bars; these gaps currently weaken the evidential weight of the central new result.","major_comments":[{"comment":"","section":"3.2.2"},{"comment":"","section":"2.4.1 and 3.2"},{"comment":"","section":"Data availability"},{"comment":"","section":"3.3"}],"minor_comments":[{"comment":"","section":"Introduction"},{"comment":"","section":"2.4.2 and Fig. 9(d)"},{"comment":"","section":"3.2.3"},{"comment":"","section":"3.3"}],"recommendation":"major_revision","confidential_remarks":"The paper has a genuine experimental result with direct visual evidence, but the quantitative backbone needs substantial revision: the classification-threshold sensitivity, replicate strategy, and data-availability statement must be addressed before the central claim can be accepted. The scope fits physics of fluids / microfluidics journals well. I would not recommend rejection because the qualitative mechanism is directly observed and is not merely an artifact of the classification; however, the current manuscript is not ready in its quantitative claims."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe paper is worth knowing about: it is one of the few microfluidic studies that systematically varies the viscosity of the injected fluid after waterflooding, covering ηR from 1 to 80, and the imaging directly shows large ganglia snapping at pore throats and the resulting fragments moving downstream. That qualitative result—tertiary recovery working through ganglia breakage feeding blobs and droplets, then transport—looks real and is well illustrated. The authors also do a careful job with the drainage and waterflood baselines, including repeatability for those steps, which gives the setup credibility.\n\nThe soft spots are in the quantitative layer. The classification of oil clusters into droplets/blobs/ganglia uses hand-set area and circularity intervals (Section 3.2.2), and those rules are not exhaustive: an object of 0.001 mm² with circularity 0.5 falls in no class, and the paper does not say how such objects were handled. More importantly, because \"ganglia\" is defined by a lower area cutoff, the statement that their number changes only slightly while their volume drops is partly a consequence of the cutoff—any breakup that pushes fragments below 0.004 mm² automatically moves them out of the ganglia class. Without a sensitivity analysis on these thresholds, the quantitative support for the central claim is weaker than the text implies. The tertiary experiments also have no error bars or stated replicate counts, and the data availability statement says \"No data was used,\" which is at odds with the experiments described. The CDC curve combining viscosity and flow-rate changes is fine as a trend but not a precise quantitative fit.\n\nNone of this kills the paper. The time-lapse evidence in Fig. 9(d) is direct and convincing, and the mechanism is independent of the classification details. The fix is straightforward: report thresholds sensitivity, give replicate statistics for the tertiary runs, and correct the data availability statement.\n\nThis deserves peer review and likely publication after revision. I'd send it to someone working on pore-scale displacement or microfluidic EOR; they'll get a useful experimental reference and a reminder about how easily cluster statistics can be threshold-driven.\n\nBest.","headline":"A visually convincing tertiary-recovery mechanism, with cluster statistics that need threshold sensitivity and replication before the quantitative claims can be trusted.","tokens_in":20689,"tokens_out":2737,"would_cite":true,"duration_ms":27203,"reading_group":"yes","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":"Injecting a more viscous fluid after waterflooding raises oil recovery primarily by breaking large trapped oil clusters into smaller fragments that are then transported away, not by mobilizing whole clusters.","keywords":["oil clusters mobilization","breakage","viscosity ratio","tertiary oil recovery","microfluidic chip","pore-scale dynamics","ganglia breakup","capillary desaturation curve"],"falsifier":"Image one isolated trapped ganglion continuously while the injected fluid viscosity is stepped up: the breakup-first mechanism predicts the ganglion thins at a pore throat, snaps into a downstream fragment that moves and an upstream remnant that remains trapped, and the small fragments reach the outlet only after a delay. Observing the whole ganglion translate without snapping, or seeing small-fragment counts rise without a transport lag, would contradict the paper's central claim.","tokens_in":19688,"feed_emoji":"🛢️","tokens_out":11145,"duration_ms":100257,"temperature":0.7,"pith_summary":"After a waterflood leaves oil trapped as large connected clusters, this paper asks where the extra oil comes from when a more viscous fluid is injected afterward. Using a transparent water-wet micromodel, the authors show that raising the injected viscosity from that of water to eighty times that of water substantially lowers residual oil saturation, yet the number of large ganglia barely changes while their total volume drops sharply. The missing volume reappears as blobs and droplets produced by ganglia breakup at pore throats, and these small fragments are what get transported out of the porous medium. The paper concludes that tertiary recovery by viscosity contrast is a two-step process: fast fragmentation that feeds the medium with small oil clusters, followed by slower transport of those clusters. If correct, this shifts the design question for such floods from how to move whole clusters to how to make and keep small fragments mobile.","feed_headline":"Breakage, not transport, is the first step in tertiary oil recovery","feed_subtitle":"Microfluidic images show viscous floods break trapped oil into droplets and blobs that then flow away","key_machinery":"The objects that carry the analysis are the three cluster classes: droplets (area 0-0.002 mm², circularity 0.7-1), blobs (0.002-0.004 mm², circularity 0.3-1), and ganglia (area above 0.004 mm², circularity 0-0.7), where ganglia are the large clusters spanning more than one pore. Tracking the number, total volume, and average volume of each class as a function of injected pore volume is what exposes the asymmetry between a nearly constant ganglia count and a strongly decreasing ganglia volume. The physical mechanism is pore-throat breakage: the more viscous injected fluid builds pressure along pre-existing water pathways, a trapped ganglion thins and elongates at a throat, and when local pressure exceeds the critical capillary pressure the ganglion snaps, releasing a downstream fragment while the upstream part remains trapped. This two-step picture — fragmentation feeding the medium with blobs and droplets, then transport removing them — is the paper's central explanatory claim.","core_discovery":"In a borosilicate-glass micromodel with a rock-like pore network, after oil drainage and waterflooding leave residual oil saturations around 0.4-0.5, the authors inject glycerol/water mixtures with viscosity ratios $\\eta_R = 1, 4, 8, 20$, and $80$ at a fixed flow rate. Recovery rises with $\\eta_R$, with the steepest gains between $\\eta_R = 1$ and $20$, and the remaining oil becomes smaller and more uniformly distributed. Classifying clusters into droplets, blobs, and ganglia by area and circularity thresholds, the authors find that ganglia, although fewest in number, hold most of the oil volume; as $\\eta_R$ increases the number of ganglia changes only slightly while their total volume falls sharply. The volume loss is accounted for by an initial increase and then a gradual decline in blobs and droplets, which are born from ganglia rupture, mobilized because they are weakly held by capillary forces, and carried out of the chip. Supporting images show a ganglion thinning and snapping at a pore throat within about a second, producing a downstream fragment that moves while the upstream remnant stays trapped for further breakup, and capillary-pressure estimates place the throat pressure above the pressures in the two new segments at the moment of rupture. The paper therefore proposes a breakup-first, transport-second mechanism, with the characteristic time for transport longer than that for breakage.","pith_inferences":["If the breakup-first picture generalizes, tertiary flooding formulations should be screened for two separate abilities: creating fragments at pore throats through local viscosity contrast, and keeping those fragments from re-trapping during the slower transport step.","A direct experimental follow-up would be to stop injection immediately after a short viscous slug and image whether fragment counts continue to change; the breakup-first picture predicts they should not once flow stops.","The same fragmentation-then-transport logic could apply to other three-fluid subsurface operations, such as CO2 storage or aquifer remediation, where a viscous chase fluid follows a waterflood.","Because the reported numbers depend on fixed area and circularity thresholds, a natural robustness check is to sweep the 0.004 mm² ganglia/blob boundary and see whether the 'nearly constant ganglia count, sharply falling ganglia volume' statement survives; the qualitative mechanism does not depend on the exact boundary."],"forward_implications":["Oil recovery from a viscosity-contrast tertiary flood should be tracked by cluster-volume transfer between size classes, because total saturation alone can hide a state where ganglia counts stay flat while their volume collapses.","The steepest additional recovery gains occur for viscosity ratios between 1 and 20; beyond that, further viscosity increase buys less extra oil under these conditions.","Residual oil after an effective viscous flood should be concentrated in dead-end pores and small pore sizes, since connected ganglia in swept pathways are preferentially fragmented and removed.","Because transport, not breakage, is the slower step, the injection volume or time needed to finish a tertiary flood is set by how long small fragments take to exit the medium, not by how fast they are created."],"supporting_citations":[{"why":"Shows from velocity measurements that trapped ganglia thin and break at pore throats where internal velocity is nearly zero; this is the breakage mechanism the paper invokes.","marker":"[24]"},{"why":"Supplies the capillary-number/viscosity-ratio flow-regime map used to place the drainage and waterflooding data.","marker":"[35]"},{"why":"Gives the micromodel-specific capillary number definition used for all the reported Ca values and for the desaturation curve.","marker":"[43]"},{"why":"Proposes that oil ganglia are mobilized by fragmentation into droplets rather than whole-cluster movement, the hypothesis tested here in tertiary conditions.","marker":"[76]"},{"why":"Documents an initial-rise-then-fall cluster count from snap-off and droplet formation, the same temporal pattern seen during mixture flooding.","marker":"[44]"},{"why":"Provides the droplet, blob, and ganglion classification by size and shape on which the quantitative cluster analysis is built.","marker":"[77]"},{"why":"Supplies the capillary-pressure estimation method used to compare throat pressure with upstream and downstream pressures at breakage.","marker":"[74]"},{"why":"Supports the claim that small blobs and droplets are weakly held by capillary forces and are the entities transported after ganglia breakage.","marker":"[27]"}],"fun_headline_variants":["Viscous floods break oil ganglia before moving them","Ganglia rupture precedes transport in tertiary oil recovery","Breakage first, then flow: how viscous floods free trapped oil","Oil recovery: ganglia split, then small droplets escape","Viscosity-driven breakup unlocks trapped oil in pore networks"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The conclusion that ganglia number changes little while their volume drops sharply rests on the hand-set thresholds that define the three cluster classes, especially the 0.004 mm² boundary between blobs and ganglia; a breakup that pushes fragments below that boundary automatically moves volume out of the ganglia class by definition.","fun_headline_variants_meta":{"raw":{"variants":["Viscous floods break oil ganglia before moving them","Ganglia rupture precedes transport in tertiary oil recovery","Breakage first, then flow: how viscous floods free trapped oil","Oil recovery: ganglia split, then small droplets escape","Viscosity-driven breakup unlocks trapped oil in pore networks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000187,"raw_usage":{"total_tokens":1411,"prompt_tokens":1112,"completion_tokens":299,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":728,"completion_tokens_details":{"reasoning_tokens":219}},"tokens_in":728,"tokens_out":299,"duration_ms":3402,"temperature":1.0,"reasoning_tokens":219,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T22:13:35.349244+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Image one isolated trapped ganglion continuously while the injected fluid viscosity is stepped up: the breakup-first mechanism predicts the ganglion thins at a pore throat, snaps into a downstream fragment that moves and an upstream remnant that remains trapped, and the small fragments reach the outlet only after a delay. Observing the whole ganglion translate without snapping, or seeing small-fragment counts rise without a transport lag, would contradict the paper's central claim.","supporting_citations":[{"cited_title":"Zarikos, A","cited_arxiv_id":null,"evidence_quote":"Shows from velocity measurements that trapped ganglia thin and break at pore throats where internal velocity is nearly zero; this is the breakage mechanism the paper invokes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the micromodel-specific capillary number definition used for all the reported Ca values and for the desaturation curve."},{"cited_title":"Haney, T","cited_arxiv_id":null,"evidence_quote":"Proposes that oil ganglia are mobilized by fragmentation into droplets rather than whole-cluster movement, the hypothesis tested here in tertiary conditions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents an initial-rise-then-fall cluster count from snap-off and droplet formation, the same temporal pattern seen during mixture flooding."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the droplet, blob, and ganglion classification by size and shape on which the quantitative cluster analysis is built."},{"cited_title":"Alzahid, P","cited_arxiv_id":null,"evidence_quote":"Supplies the capillary-pressure estimation method used to compare throat pressure with upstream and downstream pressures at breakage."},{"cited_title":"Anastasiou, I","cited_arxiv_id":null,"evidence_quote":"Supports the claim that small blobs and droplets are weakly held by capillary forces and are the entities transported after ganglia breakage."}],"review_version":1}