An a-Si:H Photoconductor-on-Active-Pixel (POAP) Image Sensor on Self-Aligned p+-Sidewall-Isolated CMOS (RHDS+) for Fusion Remote Handling and Nuclear Decommissioning

Taking Photodetection out of Crystalline Silicon

抄録

Two remote-operation fields face the same limit in the radiation tolerance of cameras: the in-vessel maintenance of fusion devices and the decommissioning of the Fukushima Daiichi Nuclear Power Station. In fusion, Japan supplies the blanket remote-handling system of ITER, and a total-dose target of 1 MGy has been set for the electronics of ITER remote handling. Maintenance takes place after plasma shutdown, when neutrons are absent and the field is dominated by gamma-rays from activated structures - the same class of environment as decommissioning, where spare cameras are already unobtainable and work continues by replacing them as they degrade. Meanwhile, attempts to harden crystalline-silicon photodetectors have reached a limit. The advantage of the pinned photodiode (PPD) rests on its surface pinning layer, and what total ionizing dose (TID) modulates is precisely that surface potential. Reported countermeasures compensate this modulation with an external field, and the required potential deepens with dose. This paper proposes to take the photodetector out of silicon. Applying the electrical shielding structure (RHDS) proposed in the author's Report 1 to an imager, we move the photodetector to a hydrogenated amorphous silicon (a-Si:H) p-i-n film stacked on the multilevel interconnect, so that only a three-transistor active pixel (3T-APS) and peripheral circuits remain in silicon - the photoconductor-on-active-pixel (POAP) structure. This configuration has no pinned photodiode, no transfer gate and no pinning layer, and the main part of the degradation paths identified in prior work disappears by construction. For isolation we use a simplified form of RHDS without the bottom layer and without bonding (RHDS+): mesa isolation with self-aligned p+ sidewalls. Because every semiconductor surface touching an oxide is heavily doped p-type and fixed to GND, the interface does not invert, which removes the reason for enclosed-layout transistors; the readout can be designed by reusing standard-cell libraries. Between pixels, the n layer is segmented, a grounded shield electrode is placed in the gap, and - new relative to Report 2 - a high-permittivity, negatively charged dielectric is analyzed: its permittivity reduces the potential rise per trapped charge by a factor of two to five and extends the trapped charge at which the inter-pixel barrier vanishes by a factor of about two to four, and its negative charge offsets the trapped charge one for one up to an upper bound. We further note that the literature on irradiation effects in a-Si:H is described on the axis of displacement damage (dpa), and show by estimate that the cumulative dose expected in these gamma-dominated environments is more than two orders of magnitude below the damage at which photoconductivity disappears. The structure needs neither bonded nor SOI substrates, and the prototype can be built by combining an existing CMOS line with back-end deposition. Moreover, the photodetector is a back-end process after the interconnect is completed and the ring reset gate is realized by layout alone; introduction is therefore staged, with the isolation module untouched up to stage 2, where a single pixel p+-ring implant is added to a standard CMOS process. RHDS+ is placed beyond that.