Hand-considered hypotheses, 2301-2400
Ore formation, rock mechanics, comminution, the physics of mineral processing, drilling, ground control. Each entry was written on its own and given its own logical form, printed above it, so that a run of similar shapes is visible rather than hidden. The forms differ structurally, not only topically, and several of the entries assert that a quantity standard practice treats as a property of the rock or the ore belongs partly to the apparatus, the sampling, or the bookkeeping. None has been tested; an estimate is marked as one.
Ore formation and hydrothermal systems
Misattribution
H2301. Banded quartz veins are conventionally read as crack-seal records of repeated seismic pumping, but much of the banding in boiling-zone veins may be rhythmic silica precipitation driven by phase separation, which requires no earthquake at all. On that reading, band spacing should track the change in CO2 content across the boiling front rather than the recurrence interval of local seismicity. A vein cutting a fault with a documented aseismic slip history, banded as strongly as its seismic neighbour, separates the two.
Hidden term
H2302. Grade in a vein is modelled as a function of fluid composition and temperature, and it carries an unacknowledged term for the host rock's capacity to consume H2S by sulfidation. The same fluid should deposit more gold in an iron-rich host than in an iron-poor one at identical pressure and temperature, with the excess tracking the pyrite produced per unit vein volume. A vein crossing a lithological contact at constant depth and dip is the clean test, because fluid path and thermal structure do not change across it.
Bound
H2303. Hydrothermal veins cannot exceed the metal concentration set by the ore mineral's solubility at the inlet divided by the fraction precipitated per pass; beyond that, more fluid flow dilutes rather than enriches. This gives an upper bound on grade that is independent of how much metal the source region contains. Grade profiles from inlet to outlet along a single vein should decay monotonically, and an interior shoot rising above the inlet-equilibrium value needs either a second fluid source or a second precipitation mechanism.
Threshold
H2304. Fault-valve behaviour requires a permeability contrast between the sealed and open states, and below a threshold contrast the fault leaks continuously and never accumulates the pressure needed to re-open. The threshold is the ratio at which the sealed fault's leakage equals the recharge rate, so it depends on wall-rock permeability rather than on fault-rock strength. A fault hosted in impermeable wall rock should be banded at a depth where the same fault in permeable wall rock is massive and unbanded.
Reversal
H2305. Cooling is the default explanation for sulfide precipitation, yet several metal complexes have retrograde solubility, so precipitation should occur where the fluid is hottest along its path. Ore shoots should then sit at the thermal maximum, near the intrusion or at the deepest part of the conduit, rather than at the distal cool tip. Homogenization temperatures measured along one continuous vein settle it: a positive correlation between temperature and grade reverses the standard distal-cooling picture.
Count
H2306. Alteration selvages on either side of a vein are normally read as a diffusion profile, but their count may simply equal the number of distinct fluid pulses, each depositing one monomineralic band. On that reading the selvage count should match the number of fluid-inclusion populations and the number of distinct sulfur isotope modes, and should not scale with the square root of time. A vein formed from a single pulse under a long-lived thermal gradient is the discriminator, since it should show one selvage and a smooth gradient.
Artefact-mimics-effect
H2307. Low-salinity secondary inclusions are routinely cited as evidence of boiling, but polishing and local heating during thin-section preparation can stretch inclusions and lower their apparent salinity by necking. The artefact would produce exactly the salinity-homogenization correlation that boiling predicts, so the correlation alone cannot be evidence for it. Comparing inclusions in a polished section with those exposed on a freshly broken, ion-milled surface from the same vein decides the matter: if the low-salinity population vanishes, it was preparation.
Scale argument
H2308. Metal endowment per unit area is sometimes claimed to be scale-invariant across three orders of magnitude of fault-network size, which would make deposit density useless as a test of any metallogenic model. The invariance may instead be an artefact of counting deposits above a fixed cutoff: as the area grows, smaller deposits fall below the cutoff and the density is held constant by construction. Recomputing endowment with a cutoff that scales with area, or taking a complete census in a small window, breaks the artefact; if density then falls with area, the invariance was definitional.
Inversion of standard practice
H2309. Exploration for orogenic gold commonly ranks ground by proximity to intrusions, on the reasoning that heat drives fluid flow. Where magmatic volatiles are sulfur-rich, the intrusions may instead have titrated the ore fluid, making the productive structures the ones that stayed cool and far from the pluton. The inversion predicts that the highest-grade shoots sit in the lowest-heat-flow part of a camp, which thermochronology along a single fault set can test.
Selection effect
H2310. Gold-silver ratios quoted for epithermal ores are usually computed only on samples above an economic gold cutoff, which systematically excludes silver-rich, gold-poor material and biases the ratio. The bias is not random: it grows as the cutoff approaches the median grade, so the quoted ratio drifts with whatever cutoff each study happened to use. Assaying a continuous suite across the full grade range with no cutoff applied shows whether the ratio is a mineralogical fact or a bookkeeping one.
Unit trap
H2311. Accumulation, the product of grade and thickness, is used as if it were a quantity of metal, but it is a mass per unit area only after multiplying by density, and comparing two deposits in different host rocks without that factor can reverse their ranking. A compact magnetite-hosted ore and a low-density volcanic-hosted ore with the same accumulation value differ in contained metal by the density ratio. Recomputing both as contained metal per cubic metre, rather than per square metre, tests whether the ranking survives.
Degenerate pair
H2312. Fluid-to-rock ratios derived from oxygen isotope shifts are computed by assuming a temperature, so the ratio and the temperature are one equation with two unknowns rather than two independent outputs. A single isotopic system therefore cannot yield both, and published pairs of ratio and temperature trace a curve rather than filling a plane. Adding a temperature-sensitive system, clumped isotopes or a mineral pair, collapses the degeneracy, and the two estimates should agree only if the assumed temperature was right.
Sign error
H2313. Skarn zonation is conventionally drawn with garnet proximal and pyroxene distal to the intrusive contact, but the sense of the zonation is fixed by the sign of the coupled calcium-iron flux, which reverses in reduced intrusions. A reduced skarn should therefore carry the iron-poor mineral proximally, making the textbook pattern a redox indicator rather than a universal sequence. Mapping one skarn from marble to pluton in both an oxidized and a reduced system tests the prediction directly.
Competing rates
H2314. Whether sulfides are disseminated or massive is usually attributed to the total sulfur budget, but it is more likely set by the ratio of nucleation rate to fluid flux: fast nucleation at many sites gives disseminated grains, slow nucleation gives a few large ones that clog the conduit. A dimensionless group, nucleation rate times residence time divided by seed density, should order the textures across deposits. Growth rates read from sector zoning, combined with residence times estimated from inclusion fluxes, give the group a value that can be checked against the texture it predicts.
Percolation
H2315. Disseminated sulfide ore becomes electrically connected, and therefore leachable without fine comminution, at a percolation threshold that should sit near 15 to 20 per cent modal sulfide (est.) for equant grains. Below the threshold the grains are isolated and the rock conducts through its silicate matrix; above it, conductivity rises steeply. A resistivity-versus-grade series on one mineralized horizon, with grade from modal analysis rather than assay, would locate the threshold and show whether it is a geometric constant or a mineral-specific one.
Path dependence
H2316. Redox mineralogy in a sediment-hosted deposit records the order in which oxidant and metal arrived, not merely their final amounts, so two deposits of identical bulk composition can differ in mineralogy. Early oxidant followed by metal produces disseminated pyrite rims, while the reverse produces a single generation of complex sulfides. Spatially resolved sulfur isotopes across one zoned specimen, read as a sequence rather than a bulk average, distinguish the two histories.
Boundary layer
H2317. Vein width is generally taken as a proxy for fracture aperture, but precipitation is limited by the diffusive boundary layer at the wall, so a wide vein can grow in a narrow fracture if flow is slow. The width then records the ratio of wall reaction rate to flow velocity rather than the aperture, and the two diverge most at low flow. A vein in a fracture whose aperture is independently known, from vein-margin geometry or from a later cross-cutting vein, would test the relation.
Feedback loop
H2318. Mineral precipitation lowers permeability, which raises fluid pressure until the fracture re-opens, and the cycle's period is set by the poroelastic diffusivity of the wall rock rather than by tectonic forcing. If so, band spacing multiplied by the number of bands should give a time that scales with the square of the diffusivity length. Measuring band spacing in veins hosted by wall rocks of contrasting diffusivity, at the same depth, tests the scaling without needing an absolute date.
Aliasing
H2319. Band spacing in some laminated veins is described as random, but if the true spacing is periodic and the outcrop is sampled at an interval close to a rational fraction of it, the apparent spacing becomes a beat pattern indistinguishable from noise. Aliasing of this kind produces long-range correlation without any long-range process. Sampling one vein at several intervals, including one much finer than the suspected period, either resolves the beat or shows the spacing to be genuinely irregular.
Null result
H2320. Vertical metal zoning is often asserted from a handful of deep holes, but once the depth to the boiling point is accounted for, some districts show no residual zoning at all. The apparent zoning would then be a temperature map rather than an ore-forming process, and the district's deep potential would be misjudged. Recomputing grade as a residual against the boiling-depth curve, rather than against depth, tests whether any zoning survives.
Non-uniqueness
H2321. Two fluids of different origin can produce identical isotope and trace-element signatures if one is the other after reaction with the wall rock, so mixing and wall-rock reaction are indistinguishable in a single-tracer survey. The degeneracy breaks only with a component the wall rock cannot supply and that mixing dilutes monotonically, that is, a conservative tracer with a known end-member. A district where that tracer varies independently of the isotope ratio is the place to test which process dominates.
Hysteresis
H2322. In a roll-type uranium deposit the redox front advances and retreats along different pH-Eh paths, so its position depends on the direction of the last fluid event rather than on the present chemistry. A front that retreated should sit inside its own alteration halo, leaving an outer reduced zone with no ore. Mapping the redox boundary and the ore boundary separately, in a deposit with a documented flow reversal, tests the prediction.
Rock mechanics and fracture
Misattribution
H2323. Rockburst damage near an advancing face is often attributed to the dynamic stress wave from a seismic event, but the wave carries only a small fraction of the event's energy and the static stress concentration after excavation can do more work. If the static mechanism dominates, damage should correlate with the local stress path and with face advance, not with event magnitude or distance from the hypocentre. Strain-rate sensors at the damage site, recording whether the rock accelerated before or after the wave arrived, separate the two.
Hidden term
H2324. Uniaxial compressive strength measured in the laboratory contains a term from the loading system's stiffness, because the machine's compliance adds to the specimen's measured displacement and alters the post-peak behaviour from which the peak is read. The same rock tested in a stiff frame and a soft one should differ, and the difference should scale with the machine-to-specimen stiffness ratio. Testing one homogeneous block in two machines, with the platens instrumented separately, quantifies the term directly.
Identity
H2325. Spalling and slabbing are described as distinct instabilities, but they may be one process observed at two confining pressures, with the boundary at the pressure where the crack-normal stress becomes compressive. Above that pressure the same microcrack population grows in shear rather than in extension, changing the surface morphology but not the mechanism underneath. A series of tests at confining pressures bracketing the transition, with each fracture surface mapped, should show a continuous change in roughness rather than a discontinuity.
Bound
H2326. Stable vertical cut height in jointed rock is usually calculated from intact strength, but in a blocky mass it is bounded by the joint spacing, because the critical block is defined by the spacing and fails by sliding or toppling at a height set by it. Intact strength becomes irrelevant once the spacing is small relative to the cut. Cutting two walls in one rock mass at different joint spacings, with intact strength held constant, tests which variable controls the height.
Threshold
H2327. Jointed rock behaves as a continuum when the excavation span greatly exceeds the joint spacing and as an assembly of discrete blocks when it does not, and the transition should occur at a span-to-spacing ratio near five (est.). Below the threshold, continuum models should over-predict stability because they average away the blocks that actually fail. Testing one numerical model against monitored convergence across a range of span-to-spacing ratios locates the transition and shows whether it is a constant.
Non-monotonic
H2328. Confining pressure usually suppresses dilatancy, but in high-porosity rock it should first enhance and then suppress it, because low pressure closes pores while moderate pressure crushes them and creates fines that dilate. The peak in dilatancy with pressure is the signature of pore collapse competing with frictional sliding. A triaxial series on a porous sandstone, with porosity measured before and after each test, would show whether the dilatancy peak coincides with the onset of grain crushing.
Count
H2329. Acoustic emission events recorded during loading are counted at a sensitivity set by the sensor's threshold, not by the amount of damage, so the b-value computed from them is a threshold artefact unless the threshold lies below the smallest event of interest. Lowering the threshold should raise the count and shift the b-value if the catalogue is incomplete. Running one test at two thresholds and comparing the b-values, rather than the counts, shows whether the reported value belongs to the rock.
Artefact-mimics-effect
H2330. Time-dependent strain in a triaxial cell is often read as creep, but a saturated specimen keeps compacting as pore fluid drains from the dead volume of the platens, and the resulting displacement looks exactly like primary creep. The artefact should scale with the system's storage volume and vanish for a dry specimen at the same stress. Testing a dry and a saturated specimen of the same rock, with the drainage path instrumented, separates the two.
Scale argument
H2331. Strength scaling from laboratory to field is often fitted as a single power law, but the exponent changes at the crossover between flaw-controlled and fabric-controlled failure, so an exponent fitted over two decades mispredicts at four. The crossover should appear as a break in slope when strength is plotted against size for specimens spanning the transition. Testing specimens from centimetre to metre scale in one rock mass, rather than extrapolating, locates the break.
Inversion of standard practice
H2332. Support design commonly targets the highest measured stress, on the assumption that the most stressed rock is the most damaged. Where damage is caused by unloading during excavation rather than by the final stress, the most damaged rock can be the least stressed, because it followed the steepest stress path. Comparing damage mapping with the modelled stress path, rather than with the final stress, tests whether the correlation runs the other way.
Selection effect
H2333. Intact-rock properties in the literature are biased toward specimens that survived extraction and preparation, so the mean of a published suite overestimates the intact strength of the mass. The bias is largest in weak, foliated rocks, where specimens that fail during coring never reach the machine. Comparing a cored suite with a suite cut from an impregnated block, where weak specimens are preserved, quantifies the truncation.
Unit trap
H2334. Creep rates are variously reported per second and per minute, and a factor of sixty is easily read as a difference between rock types rather than between conventions. The error is invisible within a single study and systematic across a compilation, so a literature trend can be manufactured from it. Recomputing every rate in one unit, with the original test duration checked against the reported strain, removes the artefact and shows whether any real trend remains.
Derived quantity
H2335. GSI is estimated from discontinuity observations and then used to predict modulus, but both the index and the prediction rest on the same observations, so the reported correlation is largely circular. An independent modulus measurement, seismic or plate-loading, should reproduce the correlation only if the index carries information the modulus does not already contain. Comparing predicted with independently measured modulus in a mass with well-characterized joints tests whether the index adds anything.
Limiting case
H2336. Any failure criterion for rock must reduce to the tensile strength at zero confining pressure, and several widely used empirical criteria do not, which predicts exactly where they will fail. The discrepancy grows as a criterion is extrapolated toward tension, so the error is largest in shallow, low-confinement excavations. Direct tensile tests on the same rock, compared with the criterion's zero-pressure intercept, quantify the mismatch without a full triaxial series.
Memory effect
H2337. Rock retains a memory of the maximum stress it has experienced, expressed as a Kaiser effect in acoustic emission, and this is used to bound palaeostress. The memory is erased by pore fluid at elevated pressure, so a wet specimen should show no Kaiser effect above a threshold saturation. Testing one rock dry and at several saturations, with the stress memory imposed under controlled conditions, establishes the saturation limit beyond which the palaeostress bound is invalid.
Symmetry breaking
H2338. In a layered sequence the direction of fracture propagation flips at a stiffness contrast ratio near two (est.), so the same tectonic stress produces orthogonal fracture sets in alternating beds. The flip occurs because the stiffer bed stores more strain energy and releases it along the interface rather than across it. Mapping fracture orientation bed by bed in a sequence of known elastic contrast, rather than in bulk, tests whether the flip coincides with the predicted ratio.
Feedback loop
H2339. Shear heating during slip raises pore pressure, which lowers effective stress and promotes further slip, giving a positive feedback whose onset is a critical slip rate. The feedback should produce a strain-rate threshold for localization that is independent of total displacement. Experiments at controlled slip rates bracketing the threshold, with pore pressure measured during slip, show whether localization begins at a rate rather than at a strain.
Comminution
Misattribution
H2340. Bond work index is treated as a property of the ore, but a measurable part of it belongs to the mill's internal classification, since recirculating load changes the effective energy per unit of new surface. The same ore ground in an open circuit and in a closed circuit with a sharper classifier should give different indices, and the difference should scale with classification efficiency. A locked-cycle test on one ore, with the classifier cut varied, separates the ore's contribution from the circuit's.
Hidden term
H2341. Drop-weight tests record the energy of the striker, but part of that energy returns to the anvil as recoil, and the fraction returned depends on the anvil-to-striker mass ratio rather than on the rock. The fragmentation energy attributed to the particle is therefore an overestimate that varies between apparatuses. Instrumenting the anvil with a force plate, and running two anvil masses with the same striker, measures the recoil term directly.
Identity
H2342. Abrasion and attrition are treated as separate comminution mechanisms, but they may be one process observed at two particle sizes, with the transition where the particle's inertia becomes comparable to the contact forces in the bed. Below that size the particle follows the flow and is rounded; above it, it is chipped. Grinding one material in narrow size fractions and measuring shape change rather than mass loss should show a continuous transition rather than two regimes.
Bound
H2343. Finest product achievable in a stirred mill is bounded by the classifier's cut size, not by the grinding energy, because particles below the cut are removed before they can be ground further. Energy added beyond the point where the classifier limits produces heat, not fineness. Running one mill with two classifier cut sizes at identical specific energy tests whether the product size follows the classifier or the energy.
Threshold
H2344. Below a critical particle size, fracture cannot propagate because the flaw population is exhausted, so grinding ceases to be a fracture process and becomes one of plastic wear. The threshold should scale with the largest intrinsic flaw in the material, which is a microstructural length rather than a grindability number. Grinding a series of materials with known flaw sizes to successively finer sizes, looking for the size at which the energy-fineness curve flattens, tests the scaling.
Reversal
H2345. Adding water to a dry mill first increases throughput by reducing coating and then decreases it by promoting ball coating and viscous damping, so throughput has a maximum at an intermediate moisture. The maximum should shift with mill diameter, because the damping term scales with bed depth while the coating term does not. A moisture series at two mill diameters would show whether the reversal is a property of the material or of the machine.
Count
H2346. Progeny particles per breakage event are set by the material's grain size rather than by the energy applied, so the count jumps at the energies just sufficient to break through a grain boundary and is otherwise flat. This makes progeny count a step function of energy, which no continuous breakage model reproduces. Single-particle tests with progeny counted by size, and grain boundaries mapped in the fragments, show whether the steps coincide with grain size.
Artefact-mimics-effect
H2347. Grinding aids are reported to raise mill throughput, but part of the gain may come from improved powder flow out of the mill rather than from faster breakage, since a mill that discharges more freely holds less inventory and reports more throughput at the same breakage rate. A locked-cycle test, which removes the discharge effect, should show no change in breakage rate if the aid acts only on flow. Running one aid in open and in locked-cycle configuration separates the two contributions.
Scale argument
H2348. Specific energy against size reduction follows a power law whose exponent depends on the dominant breakage mode, so extrapolating a laboratory exponent across the ductile-to-brittle transition is invalid, and the error grows with the extrapolation ratio. The exponent should change at the particle size where the plastic zone becomes comparable to the particle. Measuring the exponent in narrow size bands across the transition, rather than fitting one line to the whole curve, tests the change.
Inversion of standard practice
H2349. Finer grinding before flotation is assumed to improve liberation and therefore recovery, but in ores that produce slimes the finest fraction is the least floatable, so recovery can peak at an intermediate grind and fall on both sides. The peak's position should depend on the ore's slime-generating mineralogy rather than on its liberation spectrum. A grind series with recovery measured by size class, not in bulk, locates the peak and shows whether slimes cause it.
Selection effect
H2350. Ore hardness values are often taken from a single sample chosen because it survived blasting and handling, which selects the toughest material and overstates the hardness the plant will meet. The mill receives a distribution weighted toward the friable, so a design based on the survivor is optimistic for throughput and pessimistic for energy. Sampling the full run-of-mine stream, fines included, gives the distribution the mill actually experiences.
Unit trap
H2351. Work indices are quoted in kilowatt-hours per short ton and in kilowatt-hours per metric tonne, a difference of about ten per cent, and in compilations the two are mixed without conversion, so a real difference between ores can be manufactured from the unit alone. The discrepancy is systematic and small enough to look like geological variation. Recomputing every value to one tonnage, with the original unit traced to its source, removes the artefact.
Derived quantity
H2352. Specific surface area from gas adsorption is used as the comminution metric, but the gas accesses internal porosity as well as fracture surface, so for a porous ore the measured area is dominated by pores that comminution never created. True fracture area is then a small difference between two large numbers, and its relative error is large. Comparing gas-adsorption area with the area measured on polished sections of the same particles, where internal porosity is invisible, tests the claim.
Non-monotonic
H2353. Grinding rate against slurry density has a maximum because transport to the charge improves with density while damping of the ball charge worsens, and the position of that maximum depends on mill diameter rather than on the ore. A plant scaling up from a laboratory mill should therefore see its optimum shift to a lower density. A density series at two mill diameters, with power draw measured, locates the maximum and shows whether it moves with diameter.
Conservation constraint
H2354. Energy in a mill splits between creating new surface and sliding existing surfaces, and the split is fixed by the ratio of fresh fracture area to total contact area, so grinding efficiency cannot be improved without changing particle shape. A circuit reporting higher efficiency must therefore be producing a different shape, not a different energy. Measuring particle shape alongside surface area in two circuits of different reported efficiency tests the constraint.
Mineral processing physics
Misattribution
H2355. Poor flotation recovery is often attributed to the ore's mineralogy, but residual reagent from the previous campaign can persist in the process water and depress or activate the current feed, producing a recovery loss that follows the water rather than the ore. Such a loss should appear when water is reused and vanish when it is replaced with fresh water at the same feed. A plant trial that swaps only the water, holding the ore constant, distinguishes the two.
Hidden term
H2356. Contact angles measured on polished mineral sections carry a contribution from the polishing-induced surface layer, which can be more hydrophilic or more hydrophobic than the bulk mineral. The measured angle is therefore a property of the preparation as much as of the mineral, and the error is largest for soft, easily deformed sulfides. Comparing a polished surface with a freshly fractured one from the same crystal, in the same solution, isolates the preparation term.
Identity
H2357. Slime coating and heterocoagulation are described as separate problems, but they are one electrostatic phenomenon at different length scales, with the controlling variable the ratio of the two particles' zeta potentials rather than either particle's size. When that ratio passes a threshold, fine particles coat coarse ones regardless of the size ratio. Measuring coating extent across a series of zeta-potential ratios, at fixed size ratio, tests whether the threshold is electrostatic.
Bound
H2358. Flotation recovery is bounded by the bubble surface area flux, not by collector dose, because attachment requires a bubble surface to attach to and the dose cannot create area. Beyond the dose at which the mineral is fully hydrophobic, additional collector changes nothing, and the recovery plateau is set by gas rate and bubble size. A dose series at two gas rates shows whether the plateau moves with gas rate, which would confirm the area bound.
Threshold
H2359. Attachment fails below a critical hydrophobic force range even when the contact angle is large, because the thin film must drain to a critical thickness before the meniscus can rupture. The threshold is kinetic rather than thermodynamic, so it should depend on approach velocity and ionic strength rather than on the angle alone. A flow-cell experiment varying approach velocity at fixed contact angle locates the threshold and shows whether the controlling quantity is kinetic.
Reversal
H2360. Frother concentration usually increases recovery by stabilising bubbles, but beyond a point it decreases recovery because smaller bubbles raise the detachment probability in the froth and reduce each bubble's carrying capacity. The reversal should occur at a lower concentration in a deep froth than in a shallow one, since residence time in the froth is longer. A concentration series at two froth depths tests whether the reversal moves as predicted.
Count
H2361. Rougher concentrate mineral phases reflect the number of liberated particle classes, not the number of minerals in the feed, because a composite reports according to its dominant surface. A feed with five minerals but two dominant composite classes should show two behavioural groups, not five. Partition curves measured per particle class by automated mineralogy, rather than per mineral, test the claim.
Artefact-mimics-effect
H2362. Apparent activation of sphalerite by copper is sometimes inferred from surface assays, but a trace of copper already in the ore's lattice can report at the surface during polishing and be mistaken for an adsorbed activator. The artefact would produce activation in ores that were never copper-activated, and would vanish if the surface were etched before analysis. Comparing etched and unetched surfaces from one specimen, with depth profiling, separates lattice copper from adsorbed copper.
Scale argument
H2363. Laboratory flotation cells cannot reproduce plant hydrodynamics because bubble size and turbulence scale differently from cell size, so a scale-up factor derived from residence time alone is not geometric and will mispredict recovery. The mismatch should grow with cell size and be largest for coarse particles, whose detachment depends on turbulence. Running one ore in cells of three sizes at matched bubble size, rather than matched impeller speed, tests which variable governs the scale-up.
Inversion of standard practice
H2364. Gravity separation is usually improved by finer feed, on the assumption that liberation increases, but the sharpness of a shaking-table separation depends on the ratio of particle size to riffle spacing, so a coarser feed can give a sharper cut. Finer particles then follow the water film rather than the riffle, and separation degrades. A size series at fixed riffle geometry, with the cut measured by partition curves, shows whether sharpness peaks at an intermediate size.
Selection effect
H2365. Liberation measured on two-dimensional sections misrepresents the true three-dimensional liberation, because a section through a composite can show it as free while a section through a free particle can show it as composite, and the net bias depends on texture. The stereological correction is exact for simple textures and can be tested directly against X-ray tomography of the same particles. Comparing section-based and tomography-based liberation on one sample quantifies the bias without a modelling assumption.
Unit trap
H2366. Magnetic susceptibility is reported in SI and in cgs, differing by a factor of four pi, and plant surveys sometimes mix the two, so a real difference between ore types can be manufactured from the convention alone. The factor is close enough to other plausible differences that the error survives review. Recomputing every value to SI, with the original convention traced and checked against a standard of known susceptibility, removes it.
Derived quantity
H2367. Separation efficiency is derived from two assays, so it inherits both errors and can exceed one hundred per cent when both assays sit near their detection limits. The excess is not a physical result but an artefact of dividing two small, noisy numbers. Computing efficiency from a mass balance that includes an independent measurement of the feed, rather than from the two product assays alone, shows whether the excess disappears.
Percolation
H2368. Electrostatic separation of a granular bed requires a percolating path of conductive particles, so below the percolation threshold the separator acts as a dielectric and the observed separation is by size rather than by conductivity. The threshold should appear as a sharp change in separation efficiency at a fixed conductive fraction, independent of applied voltage above a minimum. A series of artificial mixtures with known conductive fractions, separated at several voltages, locates the threshold.
Boundary layer
H2369. Thickener performance is set by the permeability of the hindered-settling bed, which is fixed by floc size rather than by feed rate, so a thickener can be stable over a range of fluxes bounded by two permeabilities. Within that range, changing the feed rate should change bed height but not underflow concentration. A flux series with bed height and underflow density measured together shows whether the operating window has the predicted bounds.
Symmetry breaking
H2370. Shaking tables separate by density only because their stroke is asymmetric; a symmetric stroke produces a symmetric spread and no density separation. The asymmetry is therefore not an engineering detail but the mechanism, and separation should vanish as the fore-and-aft ratio of the stroke approaches one. Running one feed with a drive that can be set to symmetric and to asymmetric strokes tests the claim directly.
Hysteresis
H2371. Froth stability has hysteresis: a froth that has collapsed needs more frother to re-form than was needed to maintain it, because the collapsed film's contact line has receded and must re-advance over a hydrophobic surface. This predicts that plant upsets are self-sustaining beyond the point where the frother dose was originally sufficient. A dose ramp up and down in one cell, with froth height recorded, shows the loop and measures its width.
Competing rates
H2372. Depressant selectivity is usually discussed in terms of equilibrium adsorption constants, but in a plant the competition is kinetic: a fast collector can occupy a surface before a thermodynamically stronger depressant arrives, so selectivity depends on the ratio of the two rates. Conditioning time should therefore change selectivity even when doses are unchanged. A conditioning-time series at fixed dose, with adsorption measured on the particles, tests the kinetic ordering.
Drilling and blasting
Misattribution
H2373. Bit wear is usually attributed to abrasion by quartz, but the dominant loss in hard-rock drilling may be impact fatigue of the carbide buttons, which scales with the fourth power of impact velocity rather than with the rock's quartz content. If so, wear rate should correlate with drilling dynamics, bit bounce and impact frequency, better than with mineralogy. Drilling two rocks of similar abrasivity but different hardness at matched weight on bit, with bounce monitored, separates the two.
Hidden term
H2374. Rate of penetration contains a term from the flushing fluid's pressure, because below a critical cleaning pressure the cuttings are re-ground and the bit drills its own debris. The resulting decline is usually attributed to bit dulling, but it should appear earlier in a larger hole where annular velocity is lower. Drilling one rock at two hole diameters with the same bit type, and measuring cuttings size at the return, separates cleaning from wear.
Identity
H2375. Stick-slip and bit whirl are described as separate dysfunctions, but they may be one torsional instability observed at different rotary speeds, with the boundary set by the ratio of the string's torsional natural frequency to the rotation rate. Below that ratio the string oscillates as a whole; above it the bit rotates about a displaced axis. A speed sweep with torque and lateral acceleration recorded together should show a continuous transition rather than two regimes.
Bound
H2376. Maximum rate of penetration in a given rock is bounded by the rate at which cuttings can be removed from the face, not by the weight on bit, so beyond the cleaning limit additional weight only re-grinds debris. The bound should scale with annular flow area and the fluid's carrying capacity, not with the bit's cutting structure. Drilling at a series of flow rates and weights, with cuttings concentration measured at the return, locates the bound.
Threshold
H2377. No chip forms below a minimum weight on bit, because the bit's contact stress stays below the rock's ductile-to-brittle transition, and the threshold is a pressure rather than a force, so it scales with bit area. A larger bit should need proportionally more weight, not the same weight, to start chipping. Testing two bit sizes on one rock, with the threshold identified from cuttings morphology, shows whether the threshold is a stress.
Reversal
H2378. Heavier weight on bit normally drills faster, but in a deep, small-diameter hole it can drill slower, because the added load buckles the string, misaligns the bit, and converts axial load into lateral friction. The reversal should occur at a weight set by hole diameter and string stiffness, and should be absent in a large vertical hole. A weight series at two diameters, with survey data showing string curvature, tests it.
Count
H2379. Blast fragment counts are set by the in-situ block size, not by the explosive energy, because the energy propagates existing discontinuities and comminutes only the intact bridges between blocks. A massively jointed rock should give a fragment count close to its block count regardless of powder factor. Counts from two blasts of different powder factor in the same blocky mass, with the block size surveyed, test the claim.
Artefact-mimics-effect
H2380. Apparent sympathetic detonation in a blast can be an artefact of initiation scatter, because two charges fired a few milliseconds apart produce a vibration record indistinguishable from one charge triggering the other. The artefact should disappear when the initiation system's scatter is reduced below the travel time between holes. High-speed photography of the face, or downhole pressure records on each charge, distinguishes true sympathetic propagation from a timing coincidence.
Scale argument
H2381. Blast fragmentation curves are often fitted with a single exponent and extrapolated to larger blasts, but the breakage mechanism changes from crushing near the charge to spalling far from it, so the curve has two slopes and the single-exponent fit fails at scale. The break should appear at the distance where the stress wave's amplitude falls below the rock's tensile strength. Fragment counts by distance from a single charge, rather than by blast, locate the two slopes and the break between them.
Inversion of standard practice
H2382. Presplit blasting is designed with the closest spacing the drills can achieve, on the assumption that closer holes give a cleaner face. Beyond a point, however, the charges couple and the face is crushed rather than split, so there is an optimum spacing and not a minimum. A spacing series in one rock, with the resulting face surveyed for half-barrel count and backbreak, locates the optimum.
Selection effect
H2383. Powder factors are biased when the blasted volume is taken from the design rather than surveyed, because overbreak is omitted from the denominator and the factor is inflated. The bias is largest where the geology departs most from the design, which is exactly where comparisons between blasts matter. Surveying the blasted volume with a cavity scanner and recomputing the factor measures the bias.
Unit trap
H2384. Explosive energy is quoted in kilocalories per kilogram and in megajoules per kilogram, differing by a factor near four, and blast designs occasionally mix the two, so a change of explosive can appear as a change of rock blastability. The error is large enough to reverse a conclusion about which explosive suits a rock. Recomputing every energy to one unit, with the manufacturer's original basis traced, removes it.
Derived quantity
H2385. Fragmentation energy derived from a blast model is not independent of the fragment size distribution the same model predicts, so agreement between the two is a restatement rather than a validation. The circularity is hidden because the model's energy term and its size term are fitted separately. Predicting the size distribution from measured near-field strain, and comparing it with the model's output, breaks the circle.
Limiting case
H2386. Two adjacent charges do not produce twice the crater of one, and the additivity assumption fails below a spacing-to-depth ratio near one (est.), where each crater interacts with its neighbour and the combined volume is less than the sum. Treating craters as additive therefore overestimates volume at close spacing and underestimates it at wide spacing. A spacing series with crater volumes measured by cavity survey, rather than inferred, maps the non-additivity.
Path dependence
H2387. Hole deviation depends on the order in which rods are added, not only on the geology, because the first rod's orientation biases every later one and the error accumulates in a direction set by that initial misalignment. Re-drilling the same hole in the same ground should therefore give a different path unless the first rod is oriented identically. Two holes drilled with and without a measured initial orientation, in one rock, test whether deviation is geological or procedural.
Ground control and support
Misattribution
H2388. Roof sag in a supported opening is often attributed to the rock mass, but the measured displacement may be the compliance of the mesh and shotcrete skin, which moves before the rock does. The sag should then be largest where the skin is least stiff, independently of the rock's modulus. Extensometers anchored in the rock, read alongside surface convergence at the same station, separate skin deflection from rock deformation.
Hidden term
H2389. Support pressure of a grouted rockbolt includes the plate's bearing capacity, so a bolt in soft rock delivers less than its steel capacity because the plate punches into the surface. Delivered load should therefore scale with plate area and the rock's bearing strength, not with the bolt's diameter. Pull tests with two plate sizes on one rock and bolt, measuring the load at which the plate begins to embed, quantify the term.
Identity
H2390. Squeezing and swelling ground are usually treated as different hazards, but they can be one convergence observed with different mechanisms, distinguishable by whether movement continues when the face stops advancing. Swelling continues because it is driven by moisture uptake, while squeezing stops because it is driven by stress redistribution. Convergence monitored through a face stoppage, with water content measured over the same interval, separates them without a mineralogical analysis.
Bound
H2391. Stand-up time of an unsupported span is bounded by the time for the fracture process zone to propagate across the span, not by the rock's strength, because a strong rock with a fast-propagating process zone can fail sooner than a weak one with a slow zone. The bound should scale with span divided by the subcritical crack growth velocity at the ambient stress. Measuring stand-up time across spans in one rock mass, and computing the velocity from double-torsion tests, tests the scaling.
Threshold
H2392. Roof failure changes mode at a critical span-to-joint-spacing ratio, failing as a beam above it and as a block below it, and the two modes need different support because a beam needs mid-span reinforcement while a block needs anchoring at its corners. The threshold should be identifiable from failure geometry rather than from load. Mapping failure modes across a range of span-to-spacing ratios in one mine, and noting which support worked, locates it.
Reversal
H2393. Denser bolting is assumed to improve stability monotonically, but in a highly fractured rock it can reduce it by clamping the surface and preventing a natural arch from forming, so load transfers to a thinner zone. The reversal should occur at a density set by block size rather than by span. A density series at a blocky test site, with convergence and bolt load measured together, shows whether there is an optimum.
Count
H2394. Bolt counts for a roof are fixed by the number of kinematically admissible blocks, not by the roof area, so a larger span can need fewer bolts if its blocks are larger. The count should then be predictable from the joint survey alone, without a stress analysis. Comparing bolt counts that performed well across openings of different span in one jointed mass, against surveyed block counts, tests that.
Artefact-mimics-effect
H2395. Damage near a tunnel wall is often read as rockburst damage, but blast-induced overbreak produces fractures in the same orientation and at the same distance from the wall, so the two are easily confused. The artefact should be identifiable by the fracture's orientation relative to the blast holes rather than to the principal stress. Mapping fracture orientation against the blast-hole layout and the modelled stress field, in a development heading, separates them.
Scale argument
H2396. Excavation damaged zone thickness scales with the tunnel radius rather than with the rock's grain size, because the zone is stress-driven and the stress field scales geometrically. A small tunnel's damaged zone therefore cannot be used to predict a large one by simple proportion if the grain size is comparable to the small tunnel's radius. Measuring the zone in two tunnels of different radius in one rock, one small enough for grain size to matter, tests the scaling.
Inversion of standard practice
H2397. Pillar design conventionally uses the highest expected stress as the conservative case, but a pillar that yields sheds load to its neighbours, so designing the strongest pillar can leave the system less stable than designing a weaker, more compliant one. The inversion predicts that the highest-stress pillar in a yielding layout is not the one that fails first. Monitoring load and displacement across a pillar array during extraction tests which pillar actually fails.
Selection effect
H2398. Convergence records are biased toward instruments that survived, because extensometers in the most converged ground are the ones destroyed by the movement. The distribution is truncated exactly where the interesting behaviour is, so the mean understates true convergence. Comparing instrument survival with the convergence recorded before loss, and replacing lost instruments with redundant ones, measures the truncation.
Unit trap
H2399. Convergence is reported in millimetres while strain is the quantity that matters, and a ten-millimetre closure across a five-metre span is two thousand microstrain, well into the range where rock begins to fail, yet it reads as a small number. Reporting millimetres hides the severity and makes spans of different size incomparable. Recomputing every closure as a strain, and plotting it against the failure threshold, changes which openings appear critical.
Derived quantity
H2400. Rock mass classification is derived from the same discontinuity observations that are then used to design support, so the design restates its input rather than predicting independently. The circularity is invisible when classification and design both agree with observed performance, because both were calibrated on that performance. Measuring block size independently by photogrammetry, and using it to predict support demand without the classification, breaks the circle.
Status
One hundred entries, H2301 to H2400. Every one states a claim, a mechanism, and a test that could in principle falsify it; the forms are labelled so that structural repetition can be seen and counted rather than inferred. Where an entry needs a number it carries an estimate marked as one.
If one of these is worth an experiment, it is H2334 — that a factor of sixty between two time conventions, not any difference between rocks, accounts for part of the reported trend in creep rates. It is checkable from published tables in an afternoon, without a laboratory.
Incoprea Research · hand-considered hypotheses 2301-2400 · 3 November 2026 · ~/Desktop/Nov3/