kb/literature/ gets its first entry: TRITC's lifetime, and G20 stops refusing

[kyu-softmatter]

Depends on #1. The before/after below cannot be reproduced without it — python -m photo.cli currently raises TypeError on every invocation.

data/fluorophores.yaml > TRITC > lifetime_ns was null, so photo/gate.py emitted missing.lifetime and lens 5 returned BLOCKED for every TRITC setting — no margin, no bottleneck, no difficulty grade.

BEFORE  TRITC @ 555 nm  153.0 W/cm^2  ->  BLOCKED    [FAIL] missing.lifetime
AFTER   TRITC @ 555 nm  153.0 W/cm^2  ->  PASS       10.00  perturbation.saturation
                                                     evidence: assumed   advances: NO

Filed: 2.2 ns, free dye in PBS — Nanguneri et al. 2014, Microsc Res Tech 77(7):510-516, 10.1002/jemt.22349.

Two papers, and the pair is what makes it filable

Source	Form	Solvent	τ
Nanguneri 2014	free dye	PBS	2.2 ns
Nanguneri 2014	phalloidin conjugate	PBS	2.3 ns
Nanguneri 2014	free dye + 30 mM tryptophan	PBS	1.7 ns
Brismar 1995	free dye	ethanol	1.58 ns (SD 10 ps)
Brismar 1995	IgG conjugate	not restated	3.32 + 1.4 ns

Brismar et al. 1995 (10.1177/43.7.7608524) measured the same quantity on a confocal laser scanning microscope — this instrument's class — but dissolved free TRITC in ethanol while putting its other three dyes in distilled water. Samples here are aqueous, so the PBS value transfers and the ethanol value becomes an independent cross-check rather than a mismatch: two instruments, 572 and 575 nm, landing 1.4× apart across a solvent change that would be expected to move it.

confidence: low stands. They agree in magnitude, not to a figure.

The two disagree about what conjugation does, which is worth more than either number

Brismar's IgG conjugate rises to 3.32 ns and turns double-exponential. Nanguneri's phalloidin conjugate stays flat at 2.3 ns, and that paper attributes the quenching to a static dye–tryptophan complex rather than to binding.

So "conjugated" is not one condition — and TRITC here is neither form. data/particles.yaml records it as an explicit stand-in for a rhodamine/TRITC-class bead dye whose vendor excitation and emission specs were both measured wrong on this instrument. One registry name spans values the sources measure as different, by different amounts, in different directions. The entry says so, and the registry note: repeats the warning where someone reading the row will see it.

How much the spread matters here: it does not

G20 crosses f = 0.1 at k_ex·τ = 1/9:

candidate	G20's boundary
1.58 ns (ethanol)	7.75 × 10⁴ W/cm²
2.2 ns (filed)	5.57 × 10⁴ W/cm²
2.3 ns (phalloidin)	5.33 × 10⁴ W/cm²
3.32 ns (IgG)	3.69 × 10⁴ W/cm²

The measured power at the sample for the band that excites TRITC — Spectra > Green, 15.3 mW at 20× — is 153 W/cm² over a 100×100 µm field, 241× below the strictest of the four, and 1.5 × 10⁴ W/cm² even over a 10×10 µm field. The whole 2.1× span clamps the saturation margin at its ceiling.

Which value is filed starts to matter only near a diffraction-limited focus, which is the tweezers regime and not this gate's. That is in the entry as the reason the remaining condition mismatches are tolerable rather than disqualifying — the entry states both what it buys and how coarse it is.

The tier's rule, working as written

evidence: assumed, so advances stays NO beside a margin of 10.00.

The falsifier is a saturation power series on this instrument, and the entry sizes it: the departure from linear sits near 10⁵ W/cm², so a widefield series will not reach it and the measurement needs a deliberately tight focus. data/light_sources.yaml already carries the measured power_at_sample_mw for both usable bands, which is what makes the falsifier something someone can actually run.

Housekeeping

• kb/INDEX.md regenerated with python -m knowledge.cli write — 45 entries. The suite caught it before I did.
• Suite: 1210 passed, 11 skipped.
• filed_by is left unfilled. It attests that a person checked the citation against the paper, and the draft this came from was produced by a tool that cannot give that signature. Please sign it or tell me to.

🤖 Generated with Claude Code