The pairing. The Earth-and-planetary project proved that surface gravity cannot identify a depth profile: every radial density distribution with the same mass gives the same exterior field. The failure-forensics project needs to infer a thermal history from a solder joint's intermetallic layer — also a single integrated quantity.
These are the same problem. A closure-based "age" is an integral of a rate over a history; so is an intermetallic thickness. Two histories that integrate to the same marker are indistinguishable from the marker, however different they were.
With layer thickness x = √(k·t) and k = k₀·exp(−Q/RT), each history below was chosen to produce the same intermetallic thickness:
| History | Time for equal marker (h) | Relative damage (Q = 1.2 eV) |
|---|---|---|
| 150 °C | 1000 | 1.00 |
| 175 °C | 252.8 | 1.88 |
| 200 °C | 73.9 | 2.90 |
| 225 °C | 24.5 | 3.80 |
Identical marker, damage differing by 3.8×. The marker records the integral; the damage records a differently weighted integral; one does not determine the other.
The transferable statement: any ageing marker that grows as a single integrated rate can be matched by infinitely many histories, so "the layer is X thick therefore the joint has seen Y" is not an inference — it is a choice of history.
A second marker is the whole trick. Two markers with different activation energies give two equations for two unknowns (peak temperature and time), exactly as two spot sizes resolve κ and G in the thermal-metrology work.
Practical use: reflow and burn-in histories are routinely reconstructed from a single intermetallic thickness. The calculation says that reconstruction is underdetermined by a factor of nearly four in damage, and that the fix is a second, differently-activated marker — a design change, not a better measurement.
The pairing. The repair-and-recovery project's verification step is a mass balance. The extreme-metrology project catalogues the floors that limit any measurement. The floor that matters for a mass balance is not the instrument — it is how much material was taken.
Gy's relation: the relative variance of a sample from a particulate stream is σ² ≈ C·d³/M, with d the particle size and M the sample mass. So the mass needed for a target precision grows as the cube of the particle size:
| Particle size | Mass for 5% | Mass for 1% |
|---|---|---|
| 0.1 mm | <1 µg | 1 µg |
| 0.5 mm | 5 µg | 125 µg |
| 1 mm | 40 µg | 1 mg |
| 5 mm | 5 mg | 125 mg |
| 10 mm | 40 mg | 1 g |
The consequence for recovery claims. A 1% mass-balance closure on a 10 mm shredded stream needs a 1 g sample at minimum — and that is the theoretical floor with perfect mixing. A recovery test that assays 100 mg of a coarse stream cannot close to 1%, whatever the chemistry, because the sampling variance alone exceeds the target.
The transferable statement: before attributing a mass-balance gap to an unaccounted stream, compare the gap with the sampling variance at the particle size actually handled. If the gap is smaller than the sampling error, no chemistry is missing — the sample is.
Practical use: this is a ten-minute test that decides whether a recovery discrepancy is chemistry or sample size, and it needs no new equipment.
| Hybrid A | Hybrid B | |
|---|---|---|
| Marker | intermetallic thickness | mass-balance gap |
| What it integrates | a rate over a thermal history | a composition over a sample |
| Why it fails | many histories give the same integral | many sub-samples give the same assay |
| The fix | a second marker with a different activation energy | a larger sample, or a smaller particle size |
| Cost of the fix | one extra measurement channel | grinding or a bigger split |
Both are the gravity theorem in disguise: a single scalar integral cannot determine the distribution that produced it. The remedy is always a second, differently-weighted integral.
| # | Lead | |
|---|---|---|
| H2-1 | Any single ageing marker underdetermines thermal history by at least a factor of three in damage. | test |
| H2-2 | Two markers with activation energies differing by 0.3 eV determine both peak temperature and duration. | test |
| H2-3 | Published burn-in reconstructions that use one marker are wrong by a factor traceable to the missing second equation. | note |
| H2-4 | The closure-temperature geometry constant is the same class of error as a single-marker reconstruction. | test |
| H2-5 | Mass-balance gaps smaller than the sampling variance are routinely attributed to chemistry. | test |
| H2-6 | A sample-split test would resolve most recovery discrepancies in ten minutes. | test |
| H2-7 | Reported recovery yields should state particle size and assayed mass, or they are not comparable. | note |
| H2-8 | Grinding to one tenth the particle size reduces the required sample mass by 1000×, which is cheaper than assaying more. | test |
| H2-9 | Any "integral marker" in any field admits the same non-uniqueness argument, so the gravity proof transfers without change. | test |
| H2-10 | The two-hybrid structure (integral plus differently-weighted integral) is the general fix for non-identifiability. | test |
| H2-11 | Solder-joint lifetime predictions based on intermetallic thickness alone will scatter by 3–4×, matching the damage range computed here. | note |
| H2-12 | Recovery processes on fine powders are verifiable at milligram samples; coarse streams need grams. | test |
If one sentence survives: a single integrated marker — a layer thickness, a mass balance, a surface gravity — cannot determine the distribution that produced it, and the universal fix is a second integral weighted differently.