PVD and related sputter tools spend a large share of their consumable budget on targets. Operations often track kWh, amp-hours, or a simple lot count to the next change. Those counters are useful. They are incomplete. The target surface develops an erosion groove—the familiar racetrack—and redeposited material builds elsewhere. Film thickness maps and particle adders can move while the counter still says the target has “life left.”
Photoresist strip and ECD plating are different modules. This note is about metallic (and related) sputter targets as a consumable with geometry, not only energy.
What changes as the target erodes
Magnetron fields concentrate erosion in a ring or track. As the groove deepens, local deposition rate and angular distribution shift. Edge dies and within-wafer range often move first. Some processes also see particle bursts when flakes from redeposited shelves release during a power ramp.
A target that is changed too late ships uneven films. A target changed too early wastes metal and PM time. Neither error shows up clearly if the only SPC is mean thickness on a center die.
A more useful changeout package
I would rather see:
- Thickness uniformity (range or sigma) tied to target hours, not only mean thickness.
- Particle / defect trends on monitor wafers after power cycles that stress redeposited shelves.
- Visual or measured racetrack depth at PM—OEM limits differ by target and cathode.
- MoC when power recipe, gas, or magnet pack changes; the old hour limit may no longer apply.
Public OEM notes and materials papers discuss erosion geometry in general terms; the plant window still belongs to that chamber family. Where data are thin, shorten the interval and measure, rather than stretching the counter because the last lot “looked fine.”
Disposition tip: if a sudden uniformity cliff appears with no recipe MoC, open the target history before rewriting the etch that follows. The next module can absorb a bias for a while. It should not be the primary target metrology.