DARPA Perrseus asks for manufacturable microscale UHV packages: <1e-10 mbar in <5 mL, >3 yr lifetime, -40 to +85 C, at least 2 single-mode photonic + 2 ohmic electrical low-leak interconnects (<3 dB) as the only I/O, integrated pressure sensing across 1e-3 to 1e-10 mbar, and a wafer-scale-compatible process proven by 10 repeatable units. Known blockers: He permeation through SiO2, bonding and interconnect leaks, getter CH4 production, micro ion pumps that cannot pump noble gases. What architecture and process approaches best meet this, and where are the unsolved seams?
6 models deliberating WITH a human Commander steering between rounds. Sealed 2026-08-05T00:28:02.737Z. Engine lucentfire-roundtable/v1 (live).
The question put to the room
DARPA Perrseus asks for manufacturable microscale UHV packages: <1e-10 mbar in <5 mL, >3 yr lifetime, -40 to +85 C, at least 2 single-mode photonic + 2 ohmic electrical low-leak interconnects (<3 dB) as the only I/O, integrated pressure sensing across 1e-3 to 1e-10 mbar, and a wafer-scale-compatible process proven by 10 repeatable units. Known blockers: He permeation through SiO2, bonding and interconnect leaks, getter CH4 production, micro ion pumps that cannot pump noble gases. What architecture and process approaches best meet this, and where are the unsolved seams?
What survived
- Helium sets the binding constraint: with zero in-package noble-gas pump speed, admissible noble influx is ~5e-21 mbar·L/s, so any contiguous fused-silica/SiO2-clad path directly bounding the 5 mL volume (fiber, window, SiN cladding) is disqualified on arithmetic alone.
- Getter species-selectivity converts the package into its own noble-gas leak detector: after a single pre-seal >400 °C activation, sealed-cavity rate-of-rise IS the He/Ar influx, and this is the only credible basis for the Phase-2 3-year extrapolation (superior to external He fine-leak testing).
- The getter must be bulk-absorbing (ZrVFe/TiZrV) and run cold in service — implanted/monolayer 'structure-IS-the-getter' fails on capacity (~3e-5 mbar·L/cm² vs ~1e-5 mbar·L demand with no margin), and in-service heating or kV micro-ion devices plausibly generate more CH4/CO than the Ar they bury.
Seal (sha-256, single-writer): d6097688d3716e341520cca391a4b07a15863ee8477e7d2136f4d0b441aeb9b6
LucentFire