Viscoelastic polymer flooding · Criteria-screened evidence synthesis
Does polymer viscoelasticity reduce residual oil beyond the viscous contribution?
The screened record does not resolve it. Thirty-nine published studies reduce, through logged admission criteria, to seven holding a matched-viscosity inelastic reference, yielding 49 elastic-versus-reference contrasts on one declared velocity and capillary-number basis. The primary quantity is the paired within-study high-minus-low velocity difference across the four studies carrying both regimes: +0.047, +0.005, +0.024 and −0.037 saturation units, mean +0.010 s.u., paired interval at three degrees of freedom [−0.046, +0.066]. Only an effect larger than approximately 0.06 saturation units in magnitude would have been resolvable at this corpus size, so the record measures the literature’s resolving power rather than the effect.
—
Paired within-study mean ΔSor
—
Contrasts / admitted studies
—
Stage-level high−low difference
—
Lowest measured carbonate onset
Interactive companion to the manuscript Viscoelastic Polymer Flooding and
Residual-Oil Desaturation: A Criteria-Screened Evidence Synthesis, Carbonate Injectivity Bounds, and
Experimental Design — prepared for submission to
Petroleum Science (KeAi).
Manuscript PDF
Data Workbook
How to use
The classifier and paired estimates recompute from the 49-contrast dataset. Seeded bootstrap results are reported from the manuscript calculation.
Open the Velocity classifier pane to move the velocity cut and the boundary rule and watch the stage
cells, the paired within-study estimator and the leave-one-out range respond. Open Injectivity bounds
to set q/h and porosity and read the onset radius r*. Values that depend on a random
seed — the study-resampling bootstrap — are quoted as computed at seed 20260731 and are not
re-derived here. Contrast dataset carries every record behind the statistics, and
Model & equations carries the conventions.
The central result — paired within-study contrasts
velocity-only classifier · vi > 1 ft/D strict
Screened recordbasis of every number on this site
What the screened record can and cannot sayconclusions, as stated in the manuscript
- Of thirty-nine inventoried published studies, seven satisfy the matched-inelastic-reference screen, yielding 49 elastic-versus-reference contrasts. The screen, the per-value provenance, the exclusion log and the statistical guards against non-independence are the methodological contribution; the extensional capillary number used for replotting is prior art (Azad and Trivedi, 2021, 2023).
- The corpus does not resolve an independent velocity effect attributable to polymer viscoelasticity; this is the central finding. At the stage level the velocity-only tertiary cells do not separate (+0.052 against +0.059, n = 15 and 17); an ordering appears only under boundary reassignment (+0.068 against +0.036) or under the superseded classifier; and every salinity-unconfounded tertiary increment of at least 0.10 saturation units lies at 1 ft/D (0.3048 m/d) or above. The record likewise contains no matched-reference support for a secondary-mode elastic advantage.
- Two quantified channels inflate apparent scatter in compiled desaturation curves: stage-wise salinity change (exceeding the unconfounded elasticity channel) and the endpoint-krw convention (a 1.3–2.9-fold capillary-number bias, saturation-dependent and therefore systematic rather than random).
- New in-situ measurements in six carbonate reservoir cores place the apparent onset of increasing porous-medium resistance between at or below 0.93 ft/D (0.283 m/d) and 6.6–13.1 ft/D (2.01–3.99 m/d), history-dependent and absent to 88 ft/D (26.8 m/d) in one run. Retention, pore plugging and progressive permeability reduction are not excluded by ascending-rate data. The constructive output is a proposed minimum reporting checklist and a fully specified matched-reference carbonate experiment.
What is in this companionclick any card to open
Section 02.1 · Source screening and evidence tiers
Corpus and screening funnel
Thirty-nine published studies were inventoried. Five same-study merges reduced these to 34
evidence units; 15 were excluded with logged reasons, leaving 19 viscoelastic coreflood sources, of which
seven were Admitted and 12 Supportive. The comparison-level mapping yields the 49
elastic-versus-reference contrasts carried through the rest of this companion.
Screening flowcounts from the deposited dataset
Admission rule: a study reporting endpoint saturations for a
viscoelastic stage and a matched-viscosity inelastic reference (glycerol, xanthan, scleroglucan, or
low-molecular-weight polymer at matched apparent viscosity or matched pressure gradient), with extractable
velocity or pressure-gradient information. Supportive: a study missing one element — no inelastic
reference, recovery-basis saturations, or unreported interfacial tension or geometry; these inform
interpretation but never enter the quantitative comparisons.
Admitted studiesAppendix A · one row per study
Values are in the first-listed unit of each column header.
Reference grades: A — xanthan, scleroglucan or a matched low-elasticity polymer; B — glycerol.
Laboratory lineagebounds independence
Laboratory lineage is carried as an explicit dataset column
because it bounds independence: Qi, Erincik, Jin, Ehrenfried and Koh share the UT Austin ecosystem
(overlapping protocols, polymers and outcrop cores), against Vermolen et al. (2014, Shell), Clarke
et al. (2016, Schlumberger), Irfan et al. (2021, UTP), Cottin et al. (2014, TotalEnergies) and the
Daqing study. Four of the seven admitted studies are one lineage, and a fifth study of the same
lineage sits in the Supportive tier; the statistical treatment exists because of this.
Evidence-unit classification and reasonsAppendix B · exclusion and support log
Classification rule, applied
in order. First, does the unit contain or preserve an interpretable observation bearing on polymer
elasticity, residual-oil displacement, porous-medium resistance or field accessibility? If not, it is
Excluded. Second, does it satisfy the full matched-reference admission criteria? If so, it is Admitted.
Third, can the observation still inform interpretation, risk of bias, carbonate applicability, mode
contrast or experimental design without entering pooled or cell statistics? If so, it is Supportive; if
not, it is Excluded. A unit reporting the same experiments as a more complete source is labelled a Merged
duplicate and is never counted twice.
Domain-resolved design and reporting assessmentTable B.1 · seven admitted studies
L low concern, M some concern, H high concern, U unclear. A
criterion is rated low concern only where it is satisfied for all admitted contrasts used in the relevant
analysis; some concern where it holds for a subset, through an indirect proxy, or with incomplete
reporting; high concern where it fails for the contrasts used; and unclear where the source does not
provide sufficient information. No composite score is computed; the ten domains are not commensurable.
Section 03 · Elastic-versus-reference contrasts
Contrast dataset
All 49 elastic-versus-reference contrasts on one declared velocity basis (interstitial,
frontal advance). Positive ΔSor means the elastic fluid desaturated further than
its matched inelastic reference. Flood identifiers are dataset keys and are carried unaltered.
ΔSor against interstitial velocity
log velocity axis · dotted line = 1 ft/D cut · dashed line = 0.10 s.u.
Within the salinity-unconfounded tertiary set, every increment of
at least 0.10 saturation units lies at 1 ft/D (0.3048 m/d) or above, five of the eight at exactly the
boundary velocity. Across all 49 contrasts, thirteen increments reach 0.10 s.u., eight of them at the
boundary.
The 49 contrastsclick a column header to sort
tertiary, salinity-unconfounded — the primary 32
at exactly 1.000 ft/D — boundary record
salinity-confounded — excluded from the cell statistics
Two structural features bound the reading of any single record.
In-core contrasts are chained: each elastic stage starts from its reference’s endpoint, so the
recorded contrast is the elastic increment from a saturation the reference already set.
Injection order is not constant across studies: in Jin coreflood 1 the elastic stage preceded its
reference, is credited +0.110, and the glycerol stage injected after it removed a further 0.168 (0.238 to
0.070). Four Jin comparisons (J3, J6, J7, J11) carry no inelastic stage in their own cores and are
admitted on the study-level reference; the dataset flags both features per comparison.
IRFAN21 contributes a single high-velocity comparison, and
its admitted increment rests on 0.7 mL of recovered oil against a 62.33-mL pore volume, that is
ΔSor = 0.011. The source does not report the collection-system accuracy or
repeatability, so the significance of this increment cannot be established from the published record.
It is retained at face value and flagged in the domain-resolved assessment rather than adjusted.
Section 03.3 · Mode and flow intensity
Velocity classifier
Each comparison is classified on interstitial velocity alone: high-intensity above the cut,
low-intensity at or below it, applied as a strict inequality. Stage-level cells are descriptive context;
the paired within-study contrast is the primary quantity. Both are recomputed here as the cut and the
boundary rule are moved.
Classifier controls
tertiary, salinity-unconfounded · n = 32
How to use
① Set the velocity cut with the slider. The manuscript cut is 1.0 ft/D; 0.5 and 2.0 ft/D
are the reported regroupings. ② Choose the boundary rule: strict assigns
vi > cut to the high cell (the primary rule), inclusive assigns
vi ≥ cut. Records sitting at exactly the cut are counted in the amber flag; at
1.0 ft/D five of the 32 comparisons sit there, so the stage-level ordering is a property of the boundary
treatment as much as of the floods. ③ Read the paired estimator below the cells: only
studies carrying both cells enter it, and the count changes with the cut.
Velocity cut vi1.0 ft/D
Boundary rulestrict
Boundary records—
—
Paired within-study estimator — primary
—
Each row is one laboratory compared against itself at two flow
rates, holding laboratory, protocol, fluid systems and measurement practice fixed. Studies carrying only
one cell form no paired difference and are excluded from the estimator.
Paired differencesstudy means, high minus low
Paired inference
t interval and leave-one-study-out
A study-resampled interval for the paired estimator is not derived:
four clusters admit only 35 distinct resample combinations and cannot support a calibrated percentile
interval. The paired inference rests on the t interval and the leave-one-out table. At the
manuscript cut the interval does not constrain the effect at four studies, and the sign reverses only
when QI17 is removed.
Study-resampling bootstrapstage-level high−low difference
Unpaired study-level comparisonmean of study means
The unpaired statistic mixes one-sided studies into the contrast and is retained as a
confounded sensitivity. At the manuscript cut, five studies contribute a high cell and five a low cell,
and only four contribute both.
Subset sensitivitiescomputed live at the selected cut
The ΔNc-matched subset reproduces the
absence of separation, so restricting to capillary-number-matched pairs does not restore an ordering.
The Grade-A-referenced subset holds a single comparison above the boundary at the manuscript cut, and
no conclusion is drawn from it.
Stage-level leave-one-study-outTable 3 · at the 1 ft/D cut
The stage-level difference moves between −0.025 and
+0.009 and reaches approximate zero only when Erincik, the study holding the four boundary stages, is
removed.
Reference-grade splitTable 4 · static
The large tertiary increments are glycerol-referenced (Erincik,
Qi, Ehrenfried). Where the in-situ offset is recoverable from stage-wise pressure gradients and
velocities it runs against the elastic stages: the in-core resistance ratio of the elastic stage to its
same-core glycerol reference spans 1.16–2.11 across the six Erincik pairs, 0.44–5.0 across
Qi and 1.00–1.07 across Jin. Part of the large glycerol-referenced increments can therefore be
ordinary viscous displacement at an unmatched in-situ viscosity.
Superseded velocity-or-gradient classifierTable S2 · labelled sensitivity
The earlier classification assigned a comparison to the
high cell when the interstitial velocity exceeded 1 ft/D or the pressure gradient reached
10 psi/ft (226.2 kPa/m). It is retained only as a sensitivity, because the pressure gradient is in part
a response to the elastic resistance under test rather than an imposed condition, and because it groups
stages that differ in velocity by an order of magnitude. It does not feed any statistic elsewhere on
this site.
Section 03.6 · Carbonate injectivity measurements
Carbonate onset bounds
In-situ resistance factors were measured in six injectivity corefloods on carbonate
reservoir cores: permeability 49–193 md, φ = 0.23–0.31, 42,500–243,000 ppm TDS,
50–60 °C; five tests at residual oil saturation (Sor = 0.19–0.27) and one
single-phase test run in two cycles. The apparent onset of increasing porous-medium resistance — the
velocity at which the resistance factor turns upward — is test-dependent.
Onset ladder
interstitial ft/D · log axis · brackets, one-sided limits and run splits
bracketed onset
one-sided limit (open below)
cycle- or run-dependent state
sandstone-derived 1 ft/D threshold
Apparent onset per testTable 5 · 0.3048 m/d per ft/D
None of the tests separates rate response from progressive
retention: every test used an ascending-only rate ladder.
In-situ resistance factor against interstitial velocity
Fig. 6 basis · log–log · shaded band = 66-md onset bracket
SAV10 ATBS-based powder at 1,000 ppm and emulsion at 800 ppm;
second commercial sulfonated polymer, 10–12 MDa, at 1,000 ppm. Permeability effective at
Sor for CF-1/CF-2 and absolute for the Buhasa cores. Unstable low-rate points are shown
as open grey markers and enter no onset statement.
History dependencewhat the ascending ladders cannot separate
Retention, pore plugging, progressive permeability reduction and the extrapolated
brine baselines are not excluded by these data, and for two tests they are the leading reading. The
BU464-6B repeat runs return resistance factors of 9 and 3 near 1 ft/D as the brine permeability at
Sor falls from 116 to 51 to 27 md, a decline reproduced within the stated precision by
the permeability ratio alone (18.75 × 51/116 = 8.2), and the single-phase test’s cycle-2
curve sits a factor of 7–9 above cycle 1 below onset.
No bulk viscosity in the make-up brine and no residual
resistance factor were recorded, so the resistance factor cannot be decomposed into a viscosity part
and a permeability-reduction part; published decompositions in comparable carbonates attribute a factor
of 2.2–4.8 to permeability reduction (Masalmeh et al., 2019; Alfazazi et al., 2021). The rise is
reported as an apparent increase in porous-medium resistance rather than as a demonstrated elastic
transition.
The onset is a system propertyspread across nominally similar tests
The order-of-magnitude spread across nominally similar tests indicates that the onset
is a system property — polymer, brine, temperature, retention history and core — and must be
measured for the system of interest rather than imported. The second polymer, at 10–12 MDa, showed
no thickening to 88 ft/D (26.8 m/d) in its final run, so the behaviour is not generalized to a
polymer class and molecular weight alone does not order the onsets.
Scaling a nominal 1 ft/D onset for an 18-MDa HPAM in a
2,000-md, φ = 0.22 sandstone by the geometric term of Eq. 3 alone places the onset near
0.18–0.37 ft/D (0.055–0.113 m/d) in these cores, and the krw and
oil-saturation terms of the published at-residual shift factor lower it further, whereas the measured
onsets for the 4.3–8-MDa ATBS-based polymer lie a factor of 5–30 higher — the
direction expected from the shorter extensional relaxation time of the lower-molecular-weight backbone.
Relaxation times were not measured for either fluid.
Section 04.2 · Field-velocity application of the measured onset range
Injectivity bounds
Field relevance is bounded with the radial velocity around an unfractured vertical injector.
Eq. 7 gives the radius r* inside which a given interstitial onset velocity is exceeded at porosity
φ. This is a conservative single-layer bound; fractures, thin high-flux layers and horizontal-well
geometries extend it.
Onset-radius calculator
r* = q/(2πhφvonset)
How to use
① Set q/h, the injection rate per metre of completed interval. The
manuscript case is 10 m³/d per metre. ② Set the porosity; the representative value
is 0.27 and the measured range across the six carbonate cores is 0.23–0.31. Radii scale as
1/φ. ③ Read the three onset radii for the measured carbonate onsets of 1, 7
and 9 ft/D. The chart traces r* against q/h for the same three onsets, with the open
marker at the selected rate. Hover any curve for values; drag to zoom; double-click to reset the view.
Rate per interval, q/h10.0 m³/d per m
Porosity, φ0.270
Onset radius against rate per intervallog–log · three measured onsets
Onset radii—
At φ = 0.270 this table reproduces Table 6 of the
manuscript. Radii scale as 1/φ across the measured 0.23–0.31 porosity range.
Screening statementwhat the bound does and does not do
At a representative q/h of 10 m³/d per metre and φ = 0.27,
the measured carbonate onsets are exceeded only close to an unfractured vertical injector: within
approximately 19–24 m for the at-or-below-1-ft/D onset (0.283–0.3048 m/d; the lowest onset is
a one-sided limit, so larger radii are not excluded) and within approximately 2–3 m for the
7–9 ft/D onsets. Applied with per-test porosities the measured range places the regime of measured
resistance onset within approximately 2.4 m to 24 m of the injector.
Viscoelastic Sor credit, where claimed at all, belongs only to
the reservoir volume that the radial bound or a fracture / thin-bed equivalent places above onset.
Resistance factors of the magnitude measured here (15–75) are themselves difficult to sustain in an
unfractured vertical well; comparable factors have been argued to force open fractures regardless of the
assumed rheology. The unfractured single-layer geometry is therefore a screening idealization in this
respect as well. The bound delimits where elastic flow resistance is possible; it does not locate
desaturation.
The field-velocity argument was raised in Seright and Wang (2023)
and developed by Azad and Seright (2025), whose deep-pattern analysis (0.01–0.2 ft/D typical, under
one percent of pattern volume above onset) and exceptions (close spacing, ≈ 1.7 ft/D at
Pelican-Lake-type geometry; Daqing throughput of 0.14–0.20 PV/yr at 150–250-m spacing) are
reproduced here as an application of the measured onsets rather than as a new result.
Section 04.3 · Proposed minimum reporting checklist
Proposed minimum reporting checklist
The screen that reduced 39 published studies to seven usable studies is itself the clearest
statement of what the literature lacks. It is restated here as a prospective checklist: a future coreflood
that satisfies it enters this framework directly; one that does not cannot be placed on a common basis with
the admitted set.
R1–R8each item with its motivating defect in the screened corpus
A coreflood satisfying R1–R8 supplies a contrast the present record cannot: a
matched-viscosity inelastic reference in the same core, at matched capillary number, at constant salinity,
on a declared velocity basis, with endpoint saturations measured twice and the saturation-dependent water
relative permeability reported. One compliant flood pair would add evidence the existing record cannot supply
under any re-analysis.
Section 04.4 · The carbonate gap and the designed experiment
Designed two-core carbonate experiment
No carbonate study satisfying the admission criteria was identified in the screened corpus at
the 3 August 2026 search cutoff; the admitted set is entirely sandstone and sandpack, and the closest
near-miss fails on in-situ reference matching and constant-salinity grounds. The matched-reference
experiment is specified by the checklist of Section 4.3 and uses the measured onset range in place of the
sandstone threshold.
Elastic flow resistance is a necessary condition for any viscoelastic desaturation
mechanism, not a demonstration of it; whether desaturation accompanies the measured onset is the question the
designed experiment answers.
Core 1 — matched-reference sequencechecklist item per element
Velocity schedule
ascending then descending, bracketing the measured onset range
The ladder brackets the measured onset range of the six
carbonate injectivity tests (0.93–8.5 ft/D interstitial; 0.283–2.59 m/d, open below). The
descending arm is the control absent from all six tests: without it, progressive retention and rate
response are not separable. Anchor velocities shown are the measured onsets of Table 5.
Core 2 — order-reversal controlthe discriminating control
Pre-registered outcomesread against the competing hypotheses
Operating boundsquantified constraints on execution
A single core pair resolves the mechanism only for the system tested and does not establish a
population effect; the pre-registered outcomes are stated so the result enters the checklist framework as one
study, and the checklist is written so further studies accumulate on a common basis.
Reference · Model and equations
Model & equations
Every convention, constant and estimator behind the numbers on this site. Two velocity bases
appear: the interstitial basis of Eq. 1, Eq. 2 and Eq. 5, and the superficial (Darcy) flux
u = φvi on which the shear-rate coefficient of Eq. 3 and the radial bound of
Eq. 7 are defined. Every reported value states its basis.
Manuscript equationsEq. 1 – Eq. 7
Estimators computed in this pagesingle source of truth
Assumptions and input parameterscarried from the deposited dataset
Extensional capillary number — implementation verification
Table S1 · against Azad and Trivedi (2021), Table 3
The extensional capillary number is prior art and is applied
unchanged. The reimplementation reproduces the source pore-scale viscosities to better than 0.4% on each
source’s own velocity-reporting convention, which verifies implementation consistency only. It does
not validate the physical model, and it does not establish that
Nce resolves the study-level contradiction.
Worked example (Azad and Trivedi 2021, Experiment 2;
source inputs). k = 2,100 md, φ = 0.22, σi = 17.3 mN/m,
u = 0.2 ft/D as reported (vi = 0.909 ft/D), 1,800-ppm HPAM
(μp0 = 110 cp, λ = 0.1 s, n = 0.6,
τext = 0.352 s, μmax = 560,000 cp,
n2 = 3.57). Eq. 3 with C = 6 gives
γ̇ = 4.98 s−1; Eq. 4 gives
μapp,pore = 25,559 cp (shear branch 105 cp; strain-hardening branch 25,454 cp)
against 25,504 cp reported; Eq. 5 on the reported velocity gives
Nce = 1.04×10−3 against 1.0×10−3. The
corresponding corpus flood is CF3-HPAM1800-lo.
Reference · Nomenclature
Nomenclature
Symbols, Greek symbols and abbreviations used across this companion and the manuscript, with
the reporting unit of each quantity and its SI equivalent where the two differ.