Recovery of Over-Target RuO2-Based Thick-Film Resistors by Laser Surface Modification and Dual-Step Trimming: An 8P7R Feasibility Study Toward Multilayer Hybrid Integrated Circuits

Conventional laser trimming can only increase resistance; consequently, an as-fired RuO2 thick-film resistor already above its target cannot normally be recovered. This work evaluates a dual-step route in which a laser surface-modification step first lowers resistance and conventional closed-loop up-trimming on a shared laser trimmer then raises it to target. Commercial 8P7R SIP-8 arrays were used as a controlled vehicle before transfer to high-value multilayer hybrid integrated circuits (HICs). Two complete modification recipes were compared: a picosecond (Pico) route (10–12 ps, 1064 nm) and a nanosecond master-oscillator power-amplifier (MOPA) fiber-laser route (≈6 ns, 1064 nm); because pulse duration, average power, and focusing lens differ simultaneously, the routes are compared as complete recipes rather than as isolated laser-source effects. Across 15 panels (11 Pico and 4 MOPA), 6540 unique physical resistors had valid paired pre- and post-process measurements after 75 records with printing defects were excluded before analysis. In modified Rows 1 and 3, the ratio of modified to as-fired resistance (Rf/Ri) was 0.607±0.015 (n=3192) for the Pico route and 0.831±0.052 (n=1175) for the MOPA route; all Pico resistors, but only 77.6% (912/1175) of MOPA resistors, reached Rf/Ri ≤0.870, the level needed to bring a resistor 15% above target below target before up-trimming. Across seven up-trim kerf geometries (n=34 resistors per cell, two panels per route), the observed Pico–MOPA differences in trimmed ratio across the 15 matched kerf × row cells in which both route means lay within ±1% of target fell within an absolute engineering equivalence margin of ±0.005 in trim ratio (two one-sided tests (TOST)); the scan-cut kept every resistor within ±1%, the I-, D-, and M-cuts had at most 4 of 34 resistors per row outside it, and the U-cut did not hold it; because the cells come from a limited number of panels, this is presented as descriptive engineering evidence rather than population-level proof of route equivalence. After five thermal-shock cycles between −30 and +80∘C, 59/60 modified resistors and all 30 untreated controls remained within ±1%; the modified-minus-control increment was +0.15 percentage points (pp) for Pico (two panels) and +0.09 pp for MOPA (one panel). The MOPA thermal-shock result is presented as preliminary feasibility evidence only. Stage-mean two-point (25/125 ∘C) temperature coefficient of resistance (TCR) values remained within ±100 ppm ∘C−1 through five processing stages; after 155 ∘C storage, 2 of 10 MOPA Row 3 resistors exceeded +100 ppm ∘C−1. Top-view scanning electron microscopy (SEM) showed re-solidified surface morphologies without macroscopic delamination, and energy-dispersive X-ray spectroscopy (EDS) detected the principal O, Si, Pb, and Ru signatures; these observations do not establish the conduction mechanism. The contribution is a production-oriented feasibility validation of resistance lowering followed by precision up-trimming for otherwise untrimmable over-target resistors.