Sputtering targets: purity, erosion and the running cost nobody budgets
How to read a target certificate rather than the headline purity, why a bonded target fails when you raise the power, how much of a target you actually get to use, and what reactive sputtering control really depends on.
Consumables budgets for a deposition tool usually cover argon, wafers and pump oil. The line that gets left out is targets, and for a group working in platinum, gold or iridium the targets can cost more over five years than the tool did.
Worse, the target is the component that quietly sets film quality. A tool that is behaving badly is often a tool with a tired, contaminated or wrongly specified target in it, and no amount of adjusting the process recipe will fix that.
Purity, and what the extra nine buys you
Targets are sold by nominal purity: three nines, four nines, five nines. The number is a summary and it hides the thing you usually care about.
Read the certificate of analysis, not the headline. The certificate lists the individual impurities. What matters is whether the specific element that ruins your film is present, not the total. A magnetic contaminant at a level far below the headline purity will still ruin a magnetic measurement.
Ask whether oxygen, carbon and nitrogen are included in the stated purity. In many specifications the headline figure is metallic purity only, and the interstitial content is quoted separately or not at all. For titanium, tantalum and aluminium that omission is significant.
Match the grade to the application. A contact metal for a teaching device does not need the grade a semiconductor doping study needs. Buying five nines for everything is a way of running out of budget by March.
Density is part of the specification for ceramic and powder-pressed targets. A target at a low fraction of theoretical density outgasses, arcs and spits particles into the film. Ask for density as a percentage of theoretical and for the grain size, and treat a supplier who will not provide them as a supplier to avoid.
Bonded or monolithic
A target has to conduct heat out of its face into the cooled backing, or it cracks. How that thermal path is made is a specification point that gets skipped, and it decides how much power the tool can actually run.
Monolithic targets are a single piece clamped to the cooled cathode. Simple, no bond to fail, and the usual choice for common metals in a research tool.
Indium bonded targets are soldered to a backing plate with indium, which has excellent thermal conductivity. The limit is that indium melts at a low temperature, so exceeding the power the bond was designed for debonds the target. A debonded target overheats locally, arcs, cracks and can drop pieces into the chamber.
Elastomer bonded targets tolerate higher temperatures than indium in some cases and are used for brittle ceramics, at the cost of thermal conductivity.
The backing plate material matters. Copper is standard. Molybdenum is used where thermal expansion mismatch would crack a brittle target. Confirm which one you have, because a rebond service will ask.
Magnetic materials are their own problem. Nickel, cobalt and iron short the magnetron's field, and a standard target thickness in a standard gun will simply not strike a useful plasma. The fix is a thin target, a magnetron designed for magnetic material, or both. Suppliers quote a pass-through flux figure for magnetic targets, and it needs to be checked against the gun before the order goes in rather than after.
The racetrack, and how much target you actually get
A planar magnetron confines the plasma into a closed loop over the target face, and that loop is where erosion happens. The target does not wear down evenly. It develops a groove, the racetrack, while the centre and the rim stay close to full thickness.
The consequence is that a large fraction of an expensive disc is never sputtered. How large depends on the magnet design, on the target thickness, and on whether the magnet assembly moves. Rotating or scanned magnet designs spread the erosion and improve utilisation, and they are worth paying for when the target material is precious.
Change the target before you erode through. Sputtering into the backing plate contaminates every film in that run and can destroy the bond and the cathode. Measure the groove depth at each vent and record it.
Watch the rate at fixed power. As the racetrack deepens, the target surface moves away from the magnet field maximum and the deposition rate at a given power drifts. A rate log is a better wear indicator than a calendar.
Watch the arc rate. A rising arc count usually means either a poisoned surface, a nodule forming in the racetrack or a bond starting to fail. It is a warning, not a nuisance.
Keep the spent targets. Eroded precious metal targets retain most of their material and have a real reclaim value. Agree the reclaim route with the supplier at the time of the first order, because arranging it later, with an export of precious metal from Bangladesh, is much harder.
Flaking is the other consumable question nobody plans for. Material builds up on shields and on the chamber wall, and eventually it delaminates and lands on a substrate. Sputter-down geometries suffer more than sputter-up. Removable shields that can be cleaned or replaced on a schedule cost less than the wafers lost to a single flake event.
Reactive sputtering and the hysteresis problem
Reactive sputtering, running a metal target in a mixture of argon with oxygen or nitrogen, is how most useful nitride and oxide films are made. It is also the process most likely to be blamed on the tool when the real issue is control.
The target has two stable states. In metallic mode the surface is clean metal, the sputter yield is high and the rate is high, but there may not be enough reactive gas to fully form the compound. In poisoned mode a compound layer has formed on the target face, the yield collapses and the rate falls, though the film is fully reacted. Between them is an unstable region where a small change in reactive gas flow flips the target from one state to the other, and the transition does not happen at the same flow going up as going down. That loop is the hysteresis.
Pumping speed sets the width of the loop. A high pumping speed on the chamber narrows the unstable region and can, in some systems, remove it. This is why a reactive process that runs well on one tool will not transfer to another with a smaller pump.
Flow control alone is not control. Setting a reactive gas flow and hoping is how the process ends up in a different mode from run to run. Closed loop control on a signal that reports the target state, such as optical emission from the plasma or the cathode voltage, is what makes the process repeatable.
Pulsed DC or mid-frequency supplies reduce arcing when an insulating layer forms on the target, and for oxide work they are close to essential.
Pre-sputter every time. After a vent, and after any change of reactive gas, the target surface is not in the state your recipe assumes. A shuttered conditioning burn brings it back. Skipping it is the most common cause of a first run that does not match the rest of the batch.
A reactive recipe that is not under closed loop control is not a recipe, it is a coincidence that happened to work on the day it was written.: Common experience on shared deposition tools
Budgeting for targets in Bangladesh
The running cost of a sputter tool is dominated by three things: what the target material costs, how much of it you use rather than throw away, and how often you vent.
Every vent costs target life. The re-conditioning burn after a vent removes material without depositing anything useful. Batching runs so that the chamber opens less often extends target life measurably and is free to implement.
Lead times are the constraint, not price. A custom composition or a bonded target is made to order. For anything on a funded project's critical path, hold a second target from the start, because a target ordered when the first one wears out arrives after the deadline.
Precious metal shipments need their own paperwork. Value declaration, insurance and customs treatment for gold, platinum and iridium targets are not the same as for an aluminium disc. Build that into the procurement timeline.
Rebonding means the target travels. Sending a used target back for rebonding or reclaim is an export, and the documentation takes longer than the service does. Ask the supplier for the full round trip time, not the service time.
Write targets into the grant. A five year project with a deposition tool needs a consumables line that names targets explicitly. Equipment grants routinely fund the tool and leave the group unable to run it in year two.