After a humidity-driven failure, the board may be completely dry by the time anyone inspects it.
Diagnosis can therefore begin after part of the evidence has vanished. Condensation can create a conductive film, produce leakage current, trigger electrochemical migration and then disappear. Unless a metallic dendrite remains, the original cause may be much less obvious than the component that finally failed.1
At the Technical University of Denmark, CELCORR is trying to move part of that diagnosis before failure. Its researchers combine climate testing, electrochemical measurements, droplet imaging and multiphysics models.23
The colorful field plot is not the interesting part. The model's job is narrower: sweep combinations quickly and send useful cases back to physical experiments.
A microscopic puddle
A microscopic amount of water can change the electrical behavior of the surface.
Humidity and temperature changes deposit moisture on a PCB surface. Hygroscopic manufacturing residues can make that worse by attracting water. With an electrical bias present, the film becomes an electrolyte between conductors.23
Leakage current then signals that an ionic path actually exists between conductors.
Metal ions can leave one electrode, migrate and redeposit as dendrites. Once that growth connects two potentials, the result can range from intermittent malfunction to a short circuit.2
Droplets have to meet
A 2025 paper gives the model a useful experimental boundary.2
Anish Rao Lakkaraju's team used a custom test PCB, DC chronoamperometry, in-situ imaging and a COMSOL secondary-current model to follow condensed droplets as they grew and merged.2
For the PCB layout they studied, the result was clear: droplet coalescence was necessary for the observed leakage currents and dendrite formation.2
Relative humidity alone therefore does not describe the risk. Water still has to create a physical path between biased regions.
That turns several design choices into parameters worth sweeping: electrode spacing, conductor geometry, moisture-layer thickness, applied voltage and structures that interrupt or encourage continuity.1
Vary without fabricating
IEEE Spectrum labels the article as sponsored by COMSOL. It is therefore a usage demonstration, not an independent solver benchmark.1
The workflow itself can still be checked against DTU's published research.
CELCORR built simplified applications around a 3D board with two oppositely biased electrodes and a water layer. Industrial partners can change inputs and compare the resulting leakage-current calculations.1
At the bench, every new geometry may require a board, fixture, climate cycle, measurement run and sometimes destructive analysis.
Computation can sweep trace spacing or water thickness in batches. Only the useful cases need to return to the bench.
The limitation is published
The DTU paper also documents something software demonstrations often hide: what the model does not represent yet.2
Its secondary current-distribution model omits mass transport. The authors say it can still provide insight into how droplet coalescence affects leakage current, but it does not describe the full chemistry driving dendrite growth.2
The IEEE case says the team is now developing a tertiary current-distribution model that expands those basic inputs to include transport properties and reaction-rate constants.1
The team can start with a deliberately limited model, isolate dominant variables and add new phenomena only where disagreement with experiments demands them.
More physics buys nothing when the extra inputs cannot be measured reliably.
High voltage changes the scale
High-power electronics changes the problem again. Creepage distances, electric fields, packaging and corrosion mechanisms cannot be scaled mechanically from a small low-voltage PCB.
CELCORR's CRED project, established in 2024, specifically targets high-voltage and high-power electronics used in areas including renewable energy and electrification.5 Its program includes humidity and corrosive-gas testing, transport models through polymer packages, high-voltage test structures and work on safe creepage-distance guidelines.5
CRED describes a research program here. Nothing in the IEEE article shows that the small water-film model already predicts a complete power converter's lifetime.
An earlier power-PCB case study does show the simulation-plus-electrochemistry approach being applied to reactions under condensation and weak-organic-acid contamination.4
Find the safe boundary
For an engineer, the transferable lesson is less “buy a multiphysics solver” and more define the design boundaries that physical testing alone explores too slowly.
When a failure depends on humidity, voltage, material and trace spacing at the same time, a handful of prototypes gives isolated points. A sufficiently validated model connects those points and helps select the next experiments.
That produces better questions than “does this board survive 85 percent humidity?”
- when does water-film continuity make leakage current jump?
- what spacing interrupts the conductive path?
- which contamination moves that boundary?
- does the mechanism still hold as voltage rises?
Simulation does not reveal the root cause of a failed board by itself.
Its value is narrowing the plausible causes until an experiment can decide.