Before carbon-paper logbooks, keepers of remote lighthouses often jotted entries on whatever surface lay to hand. In 2029 curators dismantling the 1874 lantern at Fastnet Rock found that nightly washings with seawater had drawn brine into the seam of the copper oil-pan; evaporation left halite veins whose width tracked the humidity swings produced by the keeper’s breath and speech. Using synchrotron Laue micro-diffraction and a crystal-growth inverse model, researchers decoded six months of spoken entries—weather, lamp-wick trims, even the birth of a keeper’s child—turning a verdigrised tray into the world’s saltiest voice recorder.
Copper pans were hand-raised from 1.5 mm sheet, then sweated with tin at the joints. When hot paraffin oil (180 °C) met Atlantic spray, thermal shock opened micro-fissures 2–10 µm wide. Capillary action wicked brine inward; each exhaled word (60–75 dB at 10 cm) changed local relative humidity by ±3 %, modulating halite nucleation rate. The resulting vein thickness r is inversely proportional to vapour pressure, encoding amplitude as a spatial grating sampled at speech frequencies.
Reading starts by excising a 10 × 5 mm coupon under argon to prevent hydration. A 2 µm polychromatic X-ray beam maps Laue spot broadening every 1 µm; broader spots = thicker veins = lower humidity. Rotating the coupon 180° yields a 2-D vein-thickness field; projecting along the pan’s radial axis gives a 1-D humidity trace sampled at 40 kHz—sufficient for 4 kHz voice after equalising for thermal drift.
Clock recovery exploits the light-room routine. The lamp was trimmed at 20:00 GMT; vein spacing shows a 24 h periodicity. Cross-correlation with meteorological archives (kept at Valentia Observatory) aligns the trace to the calendar day. One anomalous 18 h gap coincides with a documented keeper’s influenza bout in March 1875, confirming temporal accuracy to ±30 min.
Error correction uses maritime redundancy. Each weather report is spoken twice; stacking suppresses nucleation noise, boosting SNR by 9 dB. Weak signals—such as the 800 Hz whistle used to test the fog-horn—emerge after median stacking, revealing code-words consistent with 1874 Light-House Service instructions.
Storage capacity is modest but historically priceless. A 40 cm pan stores ~300 kB of humidity data—across the estimated 1,200 pre-1900 copper pans still extant in lighthouses worldwide, the potential archive is 360 MB of daily keepers’ voices, enough to chart the human side of automation.
Restoration is nominally destructive—the coupon is removed—but the bore is filled with electrolytically grown halite crystals tinted to match patina, returning the pan to display-ready condition. Legal title follows UK National Archives policy: the pan is Crown property; the audio, being immaterial, is released under Open Government Licence after 75 years.
For coastal historians the lesson is clear: every verdigrised oil-pan is a disk. Beneath the salt blooms and copper carbonate lies a crystalline vein where the keeper’s breath still counts the waves, waiting for the right X-ray pulse and the right nucleation kernel to step out of the salt and back into the lantern room.
