Why Laser-Welded Cold Plates Leak: Defects and Root-Cause Checks

Liquid cooling · Engineering guide

Investigate a leaking laser-welded cold plate by validating the test, separating fixture leakage from part leakage, locating the suspect area and correlating it with physical evidence. Do not increase laser power merely because a leak test failed. Incomplete fusion, connected defects, cracks and fit-up can be involved, but so can test background or fixture sealing.

Preserve the failed condition and its evidence

Quarantine the part and retain its identifier, material batch, drawing revision, weld program, fixture identifier and test program. Preserve original records before re-welding, grinding or cleaning removes evidence. Keep an accepted comparison part to distinguish a broad production shift from an isolated event.

Check that records actually correspond to the same part and time. Separate identifiers for welding, helium testing and rework can break the root-cause chain even when each station stores data. Define that association alongside the liquid cooling welding and helium testing process chain.

Eliminate test-system ambiguity first

Check the method, program revision, tracer conditions and fixture interfaces. Follow the equipment procedure for calibration and background checks. Examine seals, temporary plugs, connections and test fittings for damage, contamination or incorrect assembly. Use the agreed fixture-check method and reference parts without assuming that one reference part represents every leak behavior.

Environmental helium, exhaust and tracer mixing can affect results; INFICON's tracer-gas guidance highlights these controls. Do not remove failures by changing the threshold, bypassing calibration or shortening stabilization. A test-system problem and a real part leak can exist together and require separate records.

Build a location-and-evidence matrix

The matrix generates hypotheses; location alone does not prove a cause. An integral vacuum result does not directly identify the exact leak location. Add localization through the approved process when needed, using the distinction explained in vacuum chamber testing versus sniffer localization.

Interpret four evidence categories separately

Incomplete or locally insufficient fusion

Check beam-to-joint position, joint section, fit-up and energy distribution. A visible bead covering the joint does not demonstrate continuous interface fusion. Link sections to suspect and accepted locations rather than using one good section to represent a complete closed seam.

Pores and connected paths

Whether a void contributes to a through-leak depends on its relationship to geometry and other defects. Review cleaning, plating, material condition and shielding without assigning every pore the same cause. Select examination methods appropriate to the location and inspection task.

Cracking

Correlate material condition, restraint, thermal history, filler strategy where applicable and crack location. More power or a repeated cover pass is not a universal repair. Approve the repair route and requalification requirements before proceeding.

Channel-side damage or distortion

Check local breakthrough, internal particles, channel boundaries and released-part shape. A part can pass leakage acceptance while failing a channel or functional-surface requirement, so sealing and functional checks remain separate.

Confirm the cause through controlled changes

1. Freeze the test program and confirm measurement validity.
2. Select a small set of testable hypotheses supported by location and physical evidence.
3. Retain the baseline and change fixture, fit-up or process variables through a planned trial.
4. Recheck the complete seam, sections, leakage and relevant functions.
5. Validate the correction across agreed batches, part variation and continuous production.

If the source configuration changes, record the new profile and control revision and re-establish the process window. One non-leaking sample is not a completed corrective action. The machine-selection matrix helps organize fixture and equipment inputs.

Frequently asked questions

Why can an attractive seam still leak?

Appearance shows only part of the evidence. Internal fusion, connected paths, ports and test connections still require investigation.

Does passing leakage prove pressure and thermal performance?

No. Those tests answer different questions and require separately agreed acceptance conditions.

Will more power fix every incomplete-fusion problem?

No. Correct fit-up, path and optics problems directly; additional heat may create other issues.

Prepare a useful diagnostic package

Provide AOGEO with a failure-location map, material and surface details, fixture and program revisions, original test results and preserved accepted and failed samples. Where acceptance needs clarification, use the helium-test conditions checklist before requesting a part and process evaluation.