# UNKNOWN TERRITORY — candidates.md (U001–U020) *Twenty candidates for the genuinely unknown. None of them is a claim: a claim would already be inside the space of things that can be imagined. Each names a place where the machinery of knowing is structurally blind — an automatic correction, a meeting of conventions, a proxy standing in for a mechanism, an effect labelled an error, an absent incentive — and each says why it is probably nothing. Read the honest-status lines first.* **The five blind spots.** (1) Where a correction is applied automatically. (2) Where two conventions meet. (3) Where a measurement is a proxy for a mechanism. (4) Where an effect is defined as an error. (5) Where the incentive is absent. --- ### U001 — The dead time that removes the correlations **The suspicion.** Counting detectors — photomultipliers, avalanche diodes, proportional counters, silicon photomultipliers — deliver a number of counts per interval. Dead time and pile-up distort that number; a correction restores the mean rate. The correction is a function of the rate alone. If the arriving photons are correlated — bunched from a thermal source, anti-bunched from a single emitter, or ordered by a driven mode — the dead time filters precisely those correlations, and the corrected rate is indistinguishable from a Poisson stream of the same mean. Sources may be far more, or less, ordered than they are recorded as being, because the only quantity that leaves the instrument is a rate. **Why it is invisible.** The timestamp stream is binned or discarded at acquisition, and the dead-time correction is the formal licence for discarding it: if the process is Poisson, the mean is sufficient. That assumption is in the instrument manual and is inherited by every downstream analysis. The higher moments are never computed, so no one notices that they are missing. **A way to glimpse it.** Two detectors with deliberately different dead times viewing the same source through a split fibre. The difference in their corrected rates as brightness rises measures the second moment the correction removed. A time-tagging module with resolution much finer than the dead time gives the same information, provided afterpulsing can be separated. **Honest status.** For dim thermal sources the correction is exact and the effect is unmeasurably small. Almost everything a dead time removes really is dead time; any residual would most likely be afterpulsing — known detector physics, not a new phenomenon. --- ### U002 — The background that is the sample twice **The suspicion.** In total scattering from disordered materials a photon or neutron can scatter twice before leaving the sample. The multiple-scattering term is smooth, grows with sample thickness, and is subtracted as background. But it is a self-convolution of the structure factor: a broadened, shifted copy of the same correlations. In exactly the regime where medium-range order is weak and diffuse, it may carry the only measurable weight — and it is removed before anyone looks at it. **Why it is invisible.** The subtraction sits in the standard reduction chain, applied with the vanadium normalisation and the container subtraction, before a physicist sees a structure factor. The assumption that multiple scattering is a smooth additive constant is written into the reduction manual and is never revisited per sample. **A way to glimpse it.** A thickness series on one material. Single scattering scales linearly with thickness; multiple scattering, to first order, quadratically. Fit the intensity at each wavevector as a quadratic in thickness: the curvature is the multiple-scattering term, and its shape can be compared with the self-convolution of the single-scattering structure factor. A shape that matches is a recovered signal; a shape that is flat is the assumed background. **Honest status.** Multiple scattering is usually a small percentage and broad enough that it is unlikely to imitate a sharp feature. In most materials the quadratic term will turn out to be smooth, and the exercise will confirm the standard correction. --- ### U003 — The equilibration that never happened **The suspicion.** Almost every slow measurement has a stopping rule: wait until the signal changes by less than some rate, then record. The rule defines equilibrium operationally. Any process slower than the threshold is therefore declared absent — not measured and found small, but never allowed to finish. Creep, slow adsorption, a second-stage phase transformation, slow ion redistribution and long-time relaxation all sit below such thresholds, and a sample that is still evolving is recorded as a sample at equilibrium. **Why it is invisible.** The criterion is a line in a methods section, not a result. Nothing in the data product records how long the sample was held or what fraction of the signal was still moving when the measurement began. "Equilibrated" is a state asserted by the operator, not a quantity with an error bar. **A way to glimpse it.** Hold one sample far past the criterion — ten times, a hundred times — and log the signal. A second, slower stage appears as a change of slope in a log-time plot. Alternatively, vary the threshold deliberately and plot the reported value against it: a value that depends on when you stopped is a value that was still moving. **Honest status.** For most systems the fast stage is the whole story and the slow stage carries a negligible fraction of the signal. The exercise is likely to confirm that the standard criterion is generous, and the slow component, where it exists, is usually known and dull diffusion. --- ### U004 — The charge that was put back in the wrong pixel **The suspicion.** In pixelated semiconductor detectors a photon's charge cloud usually spans several pixels. The standard correction sums the charge into the pixel with the largest signal and reconstructs the photon energy. It thereby discards the sharing pattern, which encodes the depth at which the photon interacted, the local electric field, and the carrier drift length. The detector is a depth-sensitive device that reports only energy and position. **Why it is invisible.** The correction exists to recover the energy spectrum, and it works. The sharing fraction appears only as a quality diagnostic; events with too much sharing are rejected outright. The data product is a count and an energy. Depth never enters it. **A way to glimpse it.** The sharing fraction against bias voltage and incidence angle. Grazing incidence shifts the interaction-depth distribution; lower bias shortens the drift length. Both change the pattern predictably, so the pattern can be inverted for the drift length — a material parameter otherwise needing a separate, often destructive, measurement. **Honest status.** For imaging the correction is adequate, and the depth information is a bonus rather than a hidden phenomenon. In well-behaved sensors the pattern is dominated by diffusion, which is already modelled, and the inversion may add nothing. --- ### U005 — The carbon line that moves **The suspicion.** X-ray photoelectron spectra are charge-corrected by shifting the whole spectrum until the adventitious carbon 1s line sits at a tabulated binding energy. That shift absorbs not only static charging but any sample-dependent surface potential, band bending or final-state effect that moves the carbon line along with everything else. A real, systematic variation in the sample's own electrostatics — with coverage, illumination, annealing — is reported as "referenced to C 1s," and the referencing value is itself a convention that differs between laboratories. **Why it is invisible.** The shift is applied in software before peak fitting, and its magnitude is not a reported result. The assumption that the carbon line has a fixed energy is the definition of the calibration, so a violation of it cannot appear as a residual. **A way to glimpse it.** The Auger parameter — a core-level binding energy plus the corresponding Auger kinetic energy — is independent of charging. Plot it against the nominal binding energy across a series: variation that survives is chemistry, variation that vanishes was charging. The secondary-electron cutoff or a work-function measurement gives an independent check. **Honest status.** The Auger parameter is routine in careful surface science, and the carbon convention is widely acknowledged as crude. In most cases the shift really is uniform charging, and the residual is small next to the chemical shifts being measured. --- ### U006 — The area of a molecule **The suspicion.** Specific surface area is reported as one number from nitrogen adsorption, using the BET equation and an assumed cross-sectional area for the nitrogen molecule. That area is a convention, not a measurement, and it is applied to surfaces whose chemistry, roughness and pore geometry differ by orders of magnitude. Measured with argon, krypton or carbon dioxide at their own temperatures, the same material gives systematically different areas. The differences are dismissed as "probe-dependent" — a label, not an explanation. They may be a real titration of how a surface packs different adsorbates. **Why it is invisible.** Surface area is quoted without the probe, the pressure range or the cross-section used. The BET fitting range is chosen by a consistency criterion that is itself conventional. The number then enters every derived quantity — uptake per area, turnover frequency, coating thickness — as though it were a property of the solid alone. **A way to glimpse it.** One well-characterised series measured with two probes of very different size and polarity, across a range of coverages, plotting the apparent area ratio against the BET constant. A ratio that varies smoothly with surface polarity is a real surface chemistry; a ratio that is constant is only geometric scaling. **Honest status.** Most of the discrepancy is genuine micropore accessibility and is already understood. On mesoporous, chemically homogeneous materials the probes agree well, and the exercise would mostly confirm that the convention is adequate. --- ### U007 — The potential of the electrode you did not measure **The suspicion.** Every electrochemical potential is reported against a reference electrode whose own potential is assumed constant. In practice the reference drifts, leaks chloride and develops a liquid-junction potential that depends on the sample's ionic composition. The reported overpotential is a difference between two partly unknown quantities, and a real, slow change at the working electrode — reconstruction, adsorbate ordering, space-charge relaxation — is degenerate with reference drift. **Why it is invisible.** The reference is treated as a constant of the apparatus, not as a sample. Its drift is handled by calibration against a standard, and that calibration assumes the junction potential is the same in the standard and in the sample, which it is not. The correction is performed once and never reported. **A way to glimpse it.** Two references of different chemistry — silver/silver chloride against mercury/mercury oxide, or a single against a double junction — measured simultaneously against the same working electrode. A component common to both belongs to the working electrode; a component that differs belongs to the reference. Alternatively, measure the junction potential directly in a concentration cell across the sample's ionic-strength range. **Honest status.** Junction potentials are calculable and usually small, and reference drift is dominated by mundane chloride leakage. The effect is likely to be a known nuisance rather than a hidden phenomenon. --- ### U008 — Three thicknesses **The suspicion.** Film thickness is quoted without the method that produced it. Ellipsometry gives an optical thickness convolved with a dispersion model; X-ray reflectivity gives a mass thickness and a roughness; photoelectron spectroscopy gives an attenuation length that assumes a density and an inelastic mean free path; stylus profilometry gives a mechanical step height. For a film with an interfacial layer, a density gradient or a native oxide these are four different quantities. Their disagreement is routinely reported as "method-dependent" and left there. The disagreement is a measurement of the interface. **Why it is invisible.** Each community reports its own number to its own precision and treats cross-method comparison as a metrology chore. The dispersion model, the density and the roughness are all fitted parameters, and the fit absorbs exactly the structure that would make the methods differ. **A way to glimpse it.** One film series spanning a range of roughness and interfacial chemistry, measured by all four methods on the same spots. Plot the pairwise differences against annealing time or roughness. A difference that grows is a real interfacial layer, and its slope gives the layer's thickness and optical density. **Honest status.** Each method is well characterised, and their discrepancies are largely understood in terms of precisely these effects. A careful study would probably reproduce known corrections rather than find anything new. --- ### U009 — The shift with two causes **The suspicion.** Raman and photoluminescence peak positions are used as local stress gauges: a shift of so many wavenumbers per unit strain, calibrated on a reference. The same shift is produced by local heating from the probe laser, by alloy composition, by carrier concentration and by confinement. The calibration fixes one variable and assumes the rest are constant. In a working device several vary together, and the extracted stress map is really a map of whatever combination the calibration happened to remove. **Why it is invisible.** The conversion coefficient is a constant in the analysis software, traceable to a calibration measurement, and the local temperature is either assumed equal to the stage temperature or corrected with a nominal value. The instrument reports one number — the peak position — so the two causes are never separated. **A way to glimpse it.** The Stokes/anti-Stokes intensity ratio measured on the same spot at the same time gives the local temperature independently, because it depends on the Bose population and not on stress. Subtracting the thermal part of the shift leaves the mechanical part. Sweeping laser power at fixed stage temperature separates them further. **Honest status.** The thermal shift is usually small at low power, and careful practitioners already make the anti-Stokes measurement. For most samples the correction is adequate and the residual sits inside the stated uncertainty. --- ### U010 — The pressure of a gas you did not identify **The suspicion.** An ionisation gauge reads pressure using a sensitivity factor calibrated for nitrogen. In any system whose residual gas differs from nitrogen — after a bake, with a hydrocarbon background, with helium from a leak, with water desorbing from a wall — the reading is wrong by a species-dependent factor, and the error is absorbed into "the pressure." A real change in composition therefore appears as a pressure change and is interpreted as one. **Why it is invisible.** The controller has a single sensitivity setting and a certificate for nitrogen. Composition is measured, if at all, by a residual gas analyser with its own calibration, in a different place, and the two numbers are never multiplied into a true pressure. The correction factor is a table entry, not a measurement. **A way to glimpse it.** The ratio of a nude ionisation gauge to a spinning-rotor gauge, which is composition-independent, during a controlled bake or a controlled leak. The ratio's time dependence is the composition's time dependence — a partial-pressure measurement made from two total-pressure readings. **Honest status.** Vacuum engineers know this well and the factors are tabulated. The composition is usually dominated by water and hydrogen, whose factors are known, so the true pressure is recoverable and the unknown is mostly a bookkeeping failure. --- ### U011 — The lifetime that is a diffusion time **The suspicion.** In time-resolved photoluminescence the decay is fitted with exponentials and the time constant is reported as the carrier lifetime. In a small excitation spot, carriers also diffuse out of the collection volume, and the decay is faster than the recombination time. The fit absorbs diffusion into "surface recombination velocity" or into a fast component, and the reported lifetime becomes a mixture of bulk lifetime, surface condition and spot geometry. **Why it is invisible.** A sum of exponentials fits a diffusion tail well. The spot size is a property of the microscope, not of the sample, so it is neither varied nor reported beside the lifetime, and the fitting software has no parameter for it. **A way to glimpse it.** The same sample at several spot sizes and excitation densities. Diffusion scales with spot radius and is independent of density at low excitation; bimolecular and Auger recombination scale with density. A lifetime that moves with spot size at fixed density is a diffusion time. **Honest status.** The effect is well known in the semiconductor community and is normally avoided by using a spot much larger than the diffusion length. Where it is not avoided it is a known artefact, not a hidden phenomenon. --- ### U012 — The mass that is not rigid **The suspicion.** A quartz crystal microbalance reports a frequency shift converted to mass by the Sauerbrey equation, which assumes the film is thin, rigid and impedance-matched. Soft films — polymers, hydrogels, adsorbed protein layers, liquid-like monolayers — are viscoelastic: they slip, dissipate and give a frequency shift that is not proportional to mass. The instrument also measures dissipation, but that channel is normally read as a quality check, "the film is rigid enough," rather than as a measurement of the film's shear modulus and its evolution. **Why it is invisible.** Frequency is the headline number and Sauerbrey is the default conversion. Dissipation is a second output treated as a diagnostic, not a measurand. No standard data product reports the film's rheology, so the viscoelastic state is measured and discarded in the same step. **A way to glimpse it.** The dissipation-to-frequency ratio against film thickness, swept through a phase transition — swelling, melting, a glass transition. The ratio's shape gives the viscoelastic moduli without a separate rheometer, and its discontinuity locates the transition. **Honest status.** For thin rigid films Sauerbrey is excellent, and the viscoelastic correction is standard in soft-matter work. The quantity is hidden mainly from people who do not want it. --- ### U013 — The spectrum called noise **The suspicion.** A measurement's noise is characterised by a single slope: white, one-over-f, or one-over-f-squared. But a one-over-f spectrum is, in most physical models, a superposition of Lorentzian relaxations with a broad distribution of corner frequencies. Fitting one exponent collapses that distribution into two numbers and discards its structure — the corner frequencies, their temperature dependence, their dependence on bias or strain. That distribution is a fingerprint of the defect population. **Why it is invisible.** Noise is characterised by its slope because the slope is what matters for a signal-to-noise budget. It is the error term in someone else's measurement, not the subject of one. The label "one-over-f noise" ends the inquiry exactly where the spectrum could be decomposed. **A way to glimpse it.** Spectra over many decades of frequency at several temperatures, fitted with a distribution of Lorentzians rather than a power law. The Arrhenius plot of the corner frequencies gives activation energies; a distribution with peaks rather than a smooth spread names specific defects. **Honest status.** This is standard practice in semiconductor defect spectroscopy, and in many systems the distribution really is featureless. The decomposition is also ill-posed, so a structured distribution can always be manufactured by fitting. --- ### U014 — The strain that is a fault **The suspicion.** X-ray line broadening is separated into crystallite size and microstrain by Williamson-Hall analysis and its variants, which assume strain broadening is isotropic and size broadening has a known shape. Stacking faults, twin boundaries and antiphase boundaries broaden reflections with a specific, anisotropic dependence on indices — and that broadening is absorbed into "microstrain." A real, process-dependent defect density is reported as a strain and then interpreted mechanically. **Why it is invisible.** The analysis is a plot of width against d-spacing, and the fault contribution lies along the same trend. Size and strain emerge from a linear fit, and the fit is the report. There is no residual left to inspect. **A way to glimpse it.** Compare reflections sensitive to faulting with those that are not, for the same structure. Broadening that differs between the two families and does not scale with d-spacing is the fault probability, and its dependence on processing is a microstructural variable otherwise invisible. **Honest status.** Fault analysis is well established in fcc metals and layered oxides, and strain is often shorthand for a mixture that everyone knows is a mixture. Where faults are not expected, the two-term decomposition may be entirely correct. --- ### U015 — The events in the rejected bin **The suspicion.** Imaging and counting pipelines reject events by shape: cosmic-ray hits in a CCD, muon tracks, hot pixels, glitches in an interferometer. The criteria are tuned to known backgrounds. Any real population with similar morphology — a short, sharp, spatially compact event of physical origin — is rejected with them. The rejection rate is kept as a monitoring statistic, never as a data set. **Why it is invisible.** The mask is applied before the science image is formed, and the masked pixels are discarded rather than stored. "Cosmic ray" is a label that ends the question of what the event was; the pipeline is designed so that no one ever has to ask. **A way to glimpse it.** The same exposure with and without the sample, or with the shutter closed, comparing the rejected-event rate and its spatial pattern. Events correlated with the sample's position or mass are not cosmic. Their rate, spectrum and clustering measure the instrument's own radioactivity, its Cherenkov light, or the sample's luminescence under background radiation. **Honest status.** The rejected population is tiny and overwhelmingly genuine background. Any sample-correlated excess is more likely to be a known effect — fluorescence, phosphorescence, radioactivity in the optics — than anything new. --- ### U016 — The stress that is not hydrostatic **The suspicion.** In a diamond-anvil cell, pressure is read from the ruby fluorescence shift, calibrated under hydrostatic conditions. Non-hydrostatic stress broadens the line and shifts it differently, and the standard practice is to load a pressure-transmitting medium and treat residual broadening as an imperfection. But the deviatoric stress is a real variable: it moves phase boundaries, drives preferred orientation and generates defects. The reported pressure is a mean stress, and the difference between that mean and the hydrostatic equivalent is discarded as an artefact. **Why it is invisible.** The gauge is calibrated to report one scalar, and the broadening is used only to certify hydrostaticity. A sample that fails the check is re-loaded, not studied. The stress tensor has no place in the data format, which has one pressure column. **A way to glimpse it.** The ruby line width and shape mapped across the cell at several loads and temperatures, or two gauges with different orientation dependence. The broadening's anisotropy maps the deviatoric part; its evolution with load is the yield behaviour of the medium. **Honest status.** The effect is well known in diamond-cell practice, and most experiments use a medium precisely to make it negligible. In soft media at moderate pressure the deviatoric part is small, and the scalar pressure is a good description. --- ### U017 — The constant nobody has re-measured **The suspicion.** Many numbers in handbooks — refractive indices, thermal conductivities, diffusion coefficients, solubilities, dielectric constants of common solvents — trace to one measurement, often decades old, made on an instrument that no longer exists. The value is quoted with an uncertainty and never repeated, because repetition is not a result. The true uncertainty may be larger than the quoted one and may be systematic: a calibration error, a sample-purity issue, a temperature scale since revised. **Why it is invisible.** Re-measuring a handbook value is unpublishable in both directions. Agreement is not news; disagreement is a dispute about someone's old apparatus. The incentive is absent at both ends, and the value enters thousands of calculations as an exact input. **A way to glimpse it.** A consistency check that does not require repeating the measurement: compare the handbook value with one derived from an independent relation — a thermodynamic identity, a sum rule, a correlation with a related quantity. A systematic offset across a family of related constants points at a shared calibration rather than at random error. **Honest status.** Most handbook values are good, and the commercially important ones have been re-measured. The exercise is likely to find a long tail of small, inconsequential discrepancies. --- ### U018 — The twenty-year measurement **The suspicion.** Some quantities reveal themselves only over decades: the creep of a loaded material, the dimensional drift of a standard, the slow corrosion of a container, the aging of a detector's gain, the relaxation of a glass. No funding cycle, no thesis and no paper covers a measurement whose result arrives after the instrument's designer has retired. Such quantities are known only by extrapolation from short experiments, and the extrapolation is an assumption, not a measurement. **Why it is invisible.** The incentive is absent by construction: the measurement cannot finish inside any career or grant horizon. What exists instead is a patchwork of short tests, each extrapolated with a model, and the models are never checked against a real long baseline. **A way to glimpse it.** An accidental long baseline: an instrument that has run for other purposes — a seismic station, a telescope, a standards laboratory's reference resistor — with a documented calibration history. The drift of its calibration over decades is exactly this measurement, made for free. **Honest status.** Most such drifts are dominated by well-understood processes — oxidation, relaxation, thermal cycling — and are modelled adequately already. The exercise is likely to confirm the extrapolations rather than overturn them. --- ### U019 — The measurement that would settle nothing **The suspicion.** Some discrepancies are stable, reproducible and inconclusive: two methods agree within their stated uncertainties but not within their central values, and neither is precise enough to decide. Resolving one requires a better experiment whose result matters only if it disagrees — if it agrees, the paper is "we confirmed the old value." These discrepancies persist for decades, and the quantity in question is used with an uncertainty that is really a disagreement. **Why it is invisible.** The discrepancy is known but not actionable. A null result is unpublishable and a positive one is a controversy, so the incentive is absent for precisely the measurement that would settle it. The disagreement is absorbed into an inflated error bar. **A way to glimpse it.** A meta-analysis that treats the disagreement as a signal rather than as scatter: plot the difference between methods against every parameter that differs between them — sample preparation, temperature, calibration chain. A difference that correlates with a parameter is a systematic, and the correlation is the result. **Honest status.** Most such discrepancies are eventually resolved by a better measurement or turn out to be underestimated uncertainties. The exercise may find nothing beyond the expected. --- ### U020 — The container as a reagent **The suspicion.** Every experiment happens inside something — a beaker, a crucible, a cell, a window, a substrate, a vial. The container is classified as apparatus, not sample, so its chemistry is outside the study: its leaching, its adsorption, its catalysis, its permeability, its own contribution to the signal. Yet the container contacts the sample for the whole experiment and its surface is often the most chemically active thing present. A real, systematic container contribution is absorbed into "the blank" and subtracted. **Why it is invisible.** The blank is measured once and subtracted, which converts a material into a correction. Nobody studies beakers as materials, because a paper about container leaching has no venue and no product. The container's properties are assumed from a supplier's specification rather than measured at the conditions of use. **A way to glimpse it.** Run the identical experiment in two container materials, or in the same material with two surface treatments, and difference the results. The difference is the container's contribution; its dependence on time, temperature and sample composition is a surface chemistry that no one has a reason to map. **Honest status.** Container effects are well known in trace analysis — blanks, leaching, carry-over — and are controlled if not studied. For most bulk experiments the container's contribution is far below the effect being measured.