Wafer Handling End Effector Machining: Flatness, Edge Radius and Mass
A wafer handling end effector is a cantilever with a 300 mm reach, a mass budget under 400 g, and a flatness requirement tighter than the wafer it carries. It moves 3 to 5 times per second in some atmospheric handlers, and every micron of tip deflection changes the pick height at the edge of a wafer that is already only 775 µm thick.
The part looks trivial: a plate with a pocket, some vacuum grooves, and a mounting hub. What makes it difficult is that flatness, mass, stiffness and particle generation pull against each other, and the material that solves three of them usually makes the fourth worse.
What the blade must not do
Start with the failure list. The blade must not deflect enough to change the pick plane, must not touch the wafer backside outside its intended contact pads, must not shed particles onto the device side, and must not grow or shrink relative to the robot arm when the chamber heats to 80 °C.
Deflection is the one that gets measured. A 300 mm blade fixed at the hub and loaded with 5 N at the tip deflects with the cube of its length, so a blade that is 0.5 mm out of flat is already consuming most of a 1 mm pick clearance. Handlers typically set pick height with 0.3 to 0.5 mm of margin, which means flatness and loaded deflection have to fit inside that number together.
Particles are the failure nobody sees until yield drops, and the specification that controls them is edge radius, not surface finish.
Flatness: 0.05 mm is a starting point
For a 300 mm blade, 0.05 mm flatness over the full length is the common callout, and it is adequate for atmospheric handlers with a 1 mm clearance. For a blade that enters a load lock through a 2 mm slit valve, or one that picks from a cassette with 4 mm pitch, 0.02 mm is the working number. Below 0.02 mm, you are fighting the material's own stability rather than the machining.
The problem with thin blades is that flatness measured on the machine is not flatness in service. A 1.0 mm alumina blade clamped on a fixture at 2 kN will read flat and spring back 0.03 mm when released. Machine it on a vacuum chuck with the back face as the reference, check it free-standing on three points, and specify that the measurement be taken with the part supported at its three mounting points, not on a surface plate.
Ceramic needs a different sequence. Fired alumina blanks arrive 0.1 to 0.2 mm out of flat, and diamond grinding removes that in 0.02 mm passes with a resin-bond wheel, taking equal stock from both faces. Grind one face only and the blade bows within a day as surface stress balances.
See also: Chamfer vs Fillet: Which Edge Treatment Belongs on the Drawing and Deep Pocket Machining: Depth-to-Diameter Limits, Corners and Deflection.
Edge radius, particles and the top face
The edge of a wafer blade should be a full radius between 0.2 and 0.5 mm, blended into both faces with no visible transition. A 0.2 mm radius is enough to stop chipping on ceramic; 0.5 mm is what you want on the leading edge that passes closest to the wafer. Hand polishing is acceptable only with a sequence ending at 1 µm diamond, with the radius checked on a profile projector at three points along the edge.
Surface finish on the top face should be Ra 0.4 µm or better, but it must be a controlled finish rather than a mirror. A lapped face at Ra 0.1 µm contacts the wafer over more area, so any trapped particle gets pressed into the backside. Specify Ra 0.4 to 0.8 µm with 0.5 mm raised pads covering 15 to 25 percent of the blade.
Vacuum grooves need the same care: 0.8 mm wide, 0.5 mm deep, a 0.2 mm bottom radius, and deburred from both ends where a hole breaks through. A burr inside a groove cuts hold force and shows up as a wafer slide under acceleration.
| Blade thickness | Mass at 200 × 300 mm | Tip deflection at 5 N | Typical use |
|---|---|---|---|
| 1.0 mm alumina (99.5%) | 235 g | 9.0 mm | 200 mm wafers, low acceleration |
| 1.5 mm alumina | 350 g | 2.7 mm | 300 mm wafers, atmospheric handler |
| 2.0 mm alumina | 470 g | 1.1 mm | Load lock, higher acceleration |
| 2.5 mm aluminum 6061-T6 | 405 g | 2.5 mm | Cost-sensitive tools, ambient only |
The table assumes a 300 mm cantilever fixed at the hub, 200 mm width, alumina at 300 GPa and aluminum at 69 GPa. Doubling thickness cuts deflection by eight, which is why 2.0 mm ceramic beats 2.5 mm aluminum at similar mass.
Ceramic or aluminum
Alumina at 99.5% purity is the default for semiconductor blades: 300 GPa modulus, 3.9 g/cm³ density, no outgassing, and it does not cold-weld to a wafer. Silicon carbide gives 410 GPa and better thermal conductivity, at roughly double the material cost and much higher grinding cost because of its hardness. Use SiC when the blade sees high-temperature processes or needs to shed heat.
Aluminum is the right answer when the tool runs at ambient, the wafer is 200 mm or smaller, and the budget matters. Use 6061-T6, not 7075: the zinc content complicates cleaning and the extra strength buys nothing at 2.5 mm. Skip anodize on the top face, since 20 µm changes flatness and cracks at the edge radius.
Mixing materials in one assembly is where CTE bites. Aluminum at 23.6 µm/m·K against alumina at 8.1 µm/m·K would grow 4.7 mm more over 300 mm at a 100 °C rise, but the blade is clamped, so the difference becomes stress at the hub and a bowed blade. Design the hub joint with slotted holes or a compliant insert, and keep metals and ceramics at the same temperature.
Handling and packing
A ceramic blade that survives a year in a tool can be destroyed by a shipping box. Pack vertically in a foam-lined slot with the blade on edge, never flat on foam, because a 1.0 mm blade supported only at the center will take a permanent set from its own weight in a warm truck.
Bag in cleanroom-compatible film, double bag, and add a desiccant pack to the outer bag. Mark the box with the flatness value measured at final inspection, not the drawing tolerance, so the customer can verify it on receipt.
If you are developing a new blade, send the arm interface drawing and the cassette pitch with the request. A shop that machines parts for semiconductor equipment will tell you whether the flatness callout is achievable at your chosen thickness. You can request a quote with the assembly model and get DFM notes in 24 hours.
Frequently asked questions
What flatness can be held on a 1 mm ceramic blade?
0.02 mm over 300 mm is achievable on a fired alumina blank with diamond grinding, provided stock is removed equally from both faces. Below 0.01 mm the blade becomes sensitive to clamping and to humidity, and the measurement means less than the number suggests.
Why does the blade bow after grinding?
Unbalanced surface stress. Grinding one face removes material and leaves a stressed layer, and the blade relieves toward the ground side over hours or days. Grind both faces in alternating passes and stress-relieve between rough and finish.
Should the top face be polished or left as machined?
Leave it at Ra 0.4 to 0.8 µm with defined contact pads. A mirror finish increases contact area, which presses trapped particles into the wafer backside and does nothing for grip.
How does CTE mismatch show up in service?
As a bowed blade and a loose hub joint. Aluminum grows about 4.7 mm more than alumina over 300 mm at a 100 °C rise, so the joint has to absorb that movement through slots or a compliant layer rather than through clamp force.
Related reading
Send a drawing for a quote and DFM review.
STEP, IGES, X_T, PDF or DWG up to 50 MB. Tell us quantity and finish, and the quote comes back with tolerance and process notes, not just a price.