How Are Flatness and Roughness of Optical Device Heat Sinks Controlled? Post-Sinter Flattening and Lapping Strategies
Published:
How Are Flatness and Roughness of Optical Device Heat Sinks Controlled? Post-Sinter Flattening and Lapping Strategies
Flatness and roughness on an optical heat sink are controlled in two stages. Sintering sets the bulk geometry and density. A post-sinter step then finishes only the functional faces: coining/sizing, surface grinding, lapping, or CNC, chosen according to the drawing callout. Specify flatness and roughness per ISO 1101 and ISO 21920 on the die-attach and mounting surfaces only. Leave non-functional features at as-sintered tolerance, and sequence finishing before plating. For EML/DFB and laser diode packages, finishing only those faces limits secondary operations while meeting the thermal-interface and optical-alignment requirements.
Writing the flatness and roughness requirement to standard
Most disputes over heat-sink flatness start with an ambiguous drawing. Call out flatness as a geometric tolerance per ISO 1101 or ASME Y14.5, applied to a named surface. A dimensional ± tolerance on thickness does not control flatness. A part can sit inside its thickness tolerance and still be bowed across the die pad.
State roughness per ISO 21920-2, which replaced ISO 4287 for profile parameters. Ra on its own says little about solder wetting or thermal contact. Where the die-attach process is sensitive, add Rz or a peak-related parameter, plus the evaluation length and filter. For lapped faces with directional or random lay, areal parameters per ISO 25178 describe the surface better.
The qualification side also shapes the spec. Telcordia GR-468-CORE governs reliability assurance for many telecom optoelectronic devices. It does not prescribe heat-sink flatness. Thermal cycling and damp heat can reveal die-attach voids and alignment drift associated with the thermal interface. The sintered material grade can be referenced against ISO 22068 or MPIF Standard 35 for MIM materials.
What sintering delivers before any finishing
Newlife - MIM produces laser diode heat sinks and EML/DFB thermal components in tungsten-copper (W–Cu) and high-purity copper. Optical transceiver heat spreaders use copper, W–Cu composites, or self-developed powder-metallurgy alloys such as W70Cu30 and W80Cu20. Laser diode heat sinks and optical transceiver heat spreaders are near-net shaped by MIM/PM to a sintered relative density of ≥98%, with mechanical properties close to wrought levels. As-sintered dimensional tolerance is typically ±0.3–0.5%. For laser diode heat sinks, the published route is PM + copper infiltration + CNC, which forms micro-cavities, steps and mounting features.
The linear tolerance converts directly into a feature-level band:
T = ±L × (0.003 to 0.005)
where L is the nominal dimension. On a 10 mm carrier, this gives ±0.03 to ±0.05 mm. On a 4 mm thickness, the thickness band is ±0.012 to ±0.020 mm.
This calculation drives the purchasing decision. Non-critical holes, steps, cavities and outline dimensions often fit inside this band, so they need no secondary operation and stay at near-net cost. Die-attach and bonding/reference faces usually carry flatness and roughness callouts beyond that as-sintered band. Budget a finishing operation for those faces only, and leave machining stock on them in the sintered geometry. Newlife - MIM quotes tighter-than-sintered precision part by part, through sizing or machining.
Choosing the post-sinter route
| Route | Typical role | Effect on flatness | Effect on roughness | Cost and throughput | Watch-outs |
|---|---|---|---|---|---|
| As-sintered | Non-functional faces, outlines | Limited by shrinkage uniformity | Reflects powder and mould surface | Lowest | Finish if the die-attach callout is tighter than as-sintered |
| Coining / sizing | Restoring flatness and key dimensions on ductile parts | Corrects bow on ductile sections | Can smooth slightly | Low per part; needs tooling | Less effective as tungsten content rises and stiffness increases |
| Surface grinding | Establishing a flat datum, removing larger deviations | Good | Moderate; leaves directional lay | Medium | Wheel loading on copper; burrs at edges |
| Single- or double-side lapping | Die-attach and thermal-interface faces | Suited to thin, flat parts; double-side also controls parallelism | Fine, random lay | Batch process; efficient for small parts | Abrasive embedment in soft copper; needs thorough cleaning |
| Precision CNC (fly-cut / milling) | Faces next to steps, cavities or pedestals that a lap cannot reach | Good locally | Good; tool-mark lay | Higher; per-feature | Fixturing distortion on thin parts |
For flat chip carriers and spreaders, a common sequence is coining or grinding to establish flatness, followed by lapping for roughness. Parts with raised pedestals or recessed die pockets cannot be lapped across the whole face. Those features need CNC finishing on the functional area, which is consistent with the PM + infiltration + CNC route used for laser diode heat sinks.
Process interactions that cause rejects
Plating replicates the substrate. These components support Au/Ni plating, vacuum brazing, soldering and bonding. A Ni/Au stack follows the underlying topography, so plating does not correct surface roughness. Measure roughness after finishing and again after plating, and state on the drawing which condition the spec applies to.
Finishing sets the join surface. Grind or lap the die-attach face, then clean it before plating, brazing, soldering or bonding. Include void inspection of the attached die (X-ray or acoustic) in first-article qualification.
Soft copper picks up abrasive. High-purity copper is prone to abrasive embedment during lapping and to burr formation during grinding. Specify the lapping media, the final cleaning step, and an edge-break requirement. Otherwise particles and burrs can end up under the die or in the optical path.
Removal can release or introduce stress. Removing stock from one face of a thin part can bow it again. Remove symmetric stock where possible, or use double-side lapping to control flatness and parallelism together. Verify flatness after any later thermal step, such as brazing.
When MIM is a poor fit. For large, plain plates with no 3D features, where every face needs full machining anyway, the advantage of near-net shaping shrinks. CNC from wrought W–Cu or copper plate may then compete on cost. MIM/PM pays off when micro-cavities, steps, fins or micro-channels would otherwise need extensive machining.
How Newlife - MIM supports flatness-critical heat sinks
The control chain starts at the powder. Newlife - MIM has more than 20 years of materials and powder engineering experience and produces its own ultrafine iron, copper and nickel powders (D50≤3μm). Controlling particle size distribution, morphology and purity gives more consistent sintering behaviour, and therefore more consistent stock allowance from lot to lot. This matters because finishing yield depends on the variation in the sintered part, not only its nominal size.
The same team runs the full chain from mixing and moulding through debinding, sintering and secondary operations. Machining allowances, coining tooling and lapping fixtures are therefore designed together with the sintered geometry. Copper and W–Cu thermal parts form micro-cavities, steps and thin walls near net shape, so only the functional faces need finishing. For new optical packages, OEM/ODM co-development and rapid prototyping let engineers confirm flatness, roughness and post-plating results on real parts before tooling is released.
Next steps
1. Mark the die-attach and mounting surfaces on the drawing. Apply ISO 1101 flatness and ISO 21920 (or ISO 25178) roughness callouts to those faces only. 2. State whether each roughness value applies before or after Ni/Au plating. 3. Use T = ±L × 0.3–0.5% to sort features into as-sintered and finished groups. Only the finished group should carry secondary-operation cost. 4. Choose coining, grinding, lapping or CNC by part geometry and tungsten content, using the table above. 5. Add die-attach void inspection and post-thermal-cycling flatness checks to the GR-468-style qualification plan. 6. Contact Newlife - MIM for custom engineering support. Send your drawing and flatness/roughness targets via http://www.newlife.cn to request a design consultation, a finishing-route proposal and a part-specific quotation for your laser diode, EML/DFB or transceiver heat sink.
References
1. ISO 1101:2017, Geometrical product specifications (GPS) — Geometrical tolerancing — Tolerances of form, orientation, location and run-out. 2. ISO 21920-2:2021, Geometrical product specifications (GPS) — Surface texture: Profile — Part 2: Terms, definitions and surface texture parameters. 3. ISO 25178-2, Geometrical product specifications (GPS) — Surface texture: Areal — Part 2: Terms, definitions and surface texture parameters. 4. ASME Y14.5-2018, Dimensioning and Tolerancing. 5. Telcordia GR-468-CORE, Generic Reliability Assurance Requirements for Optoelectronic Devices Used in Telecommunications Equipment. 6. ISO 22068:2012, Sintered-metal injection-moulded materials — Specifications. 7. MPIF Standard 35, Materials Standards for Metal Injection Molded Parts, Metal Powder Industries Federation. 8. German, R. M., & Bose, A. Injection Molding of Metals and Ceramics. Metal Powder Industries Federation. 9. Newlife - MIM product pages: Laser Diode Heat Sink; EML/DFB Laser Thermal Management Components; Optical Transceiver Heat Spreader. http://www.newlife.cn