Some electronics fail because metal grew inside them. Seventy years on, nobody fully knows why

Most electronic failures are recognisable: something overheated, something cracked, something corroded, something was made badly. There is one category that is stranger than all of them, and it is still not fully explained.

Sometimes a circuit fails because metal grew inside it. Not corroded, not migrated in the loose sense — grew. A single-crystal filament of tin, a couple of micrometres thick and long enough to see with a lens, pushes its way out of a plated surface over months or years, reaches a neighbouring conductor, and shorts it.

Tin whiskers have caused satellite failures, grounded aircraft equipment, triggered medical device recalls and shut down a nuclear plant. They were first documented in the 1940s. There is still no complete theory of why they form, and no test that reliably predicts whether a given finish will produce them in service.

That deserves attention on its own. It also introduces a bigger idea, which is that solid metal at room temperature is not static at all — and once you accept that, three otherwise unrelated failure modes turn out to be the same story.

A whisker grows from its base, not its tip

The first thing to understand is that a whisker is not a crystal growing outwards from its tip the way frost or a stalactite does. It is being extruded — pushed out from underneath, like toothpaste, while the tip stays whatever shape it started as.

This was settled experimentally in an elegant way: mark a growing whisker partway along its length and watch where the mark ends up. It stays a fixed distance from the tip and the new material appears at the base. The tin is being supplied from the film below and forced out through a weak point in the surface oxide.

What forces it out is compressive stress in the tin layer. A thin tin coating under compression has very few ways to relieve that stress — it cannot easily flow, it is constrained by the substrate underneath and by its own tough surface oxide — and extruding a filament through a break in that oxide turns out to be one of the few routes available.

So the real question is where the stress comes from, and here there are several contributors that vary by situation, which is part of why the problem is so slippery.

The most important is usually intermetallic growth. Tin plated over copper does not sit there inertly; copper diffuses into the tin and forms an intermetallic compound at the interface, and it does so irregularly — preferentially along tin grain boundaries, wedging in and occupying more volume than the material it replaced. The film is being pushed from within, at room temperature, by an ongoing solid-state reaction. Added to that are residual stresses from the plating process itself, stresses from thermal expansion mismatch as the assembly heats and cools, and mechanical stress from bending or clamping a plated part.

It is worth pausing on that: a circuit board sitting in a drawer, unpowered, at room temperature, is chemically active at its interfaces. Atoms are moving. This is the same solid-state diffusion that drives sintering, operating much more slowly and with nobody wanting it.

Why lead mattered, and why removing it brought this back

Here is the part that makes tin whiskers a materials-policy story as well as a physics one.

Tin-lead solder and tin-lead plating, the industry standard for decades, essentially do not grow whiskers. A few per cent of lead suppresses them extremely effectively. The mechanism is still debated — explanations centre on how lead changes the grain structure and behaviour at grain boundaries, altering how stress is accommodated — but the empirical fact was never in doubt.

When electronics moved to lead-free finishes to comply with restrictions on hazardous substances, pure tin and high-tin alloys became widespread, and the whisker problem returned at industrial scale for the first time since the 1960s. This was not a surprise to metallurgists; it was a known and openly discussed consequence.

The response has been mitigation rather than solution, and exemption where mitigation is not enough. High-reliability sectors — space, aviation, defence, some medical equipment — continue to specify tin-lead solder, under explicit regulatory exemptions, precisely because whisker risk over a twenty-year mission is not acceptable and no lead-free finish has been shown to remove it entirely.

Readers who followed our article on piezoelectricity will recognise the shape of this exactly: lead zirconate titanate also holds an exemption, also because no lead-free substitute matches it across the full range of requirements. Two entirely different industries, the same honest position — the substitution is desirable, the science is genuinely trying, and in specific applications the replacement is not yet good enough to pretend otherwise.

The mitigations that do help are worth knowing. A nickel underlayer between copper and tin acts as a diffusion barrier and substantially reduces intermetallic-driven stress. Matte tin — deposited with larger grains and lower internal stress — is markedly better than bright tin, whose fine grain structure and organic additives make it among the worst finishes. Annealing after plating relieves some stress. Thicker deposits help. And conformal coating delays whiskers and can contain short ones, but a sufficiently long whisker can penetrate or bridge over a thin coating, so it is a risk reduction rather than a fix.

Electromigration: the failure that is understood

The second way metal moves is far better characterised, and it sets a hard limit on how long a chip can run.

Inside an integrated circuit, conductors carry current densities that would vaporise ordinary wiring — routinely a million amperes per square centimetre or more, tolerable only because the conductors are tiny and well heat-sunk. At those densities, the flowing electrons do not merely pass through the metal. They transfer momentum to the metal atoms themselves, nudging them along in the direction of electron flow.

Over time, this electromigration thins the conductor where atoms are leaving and piles material up where they arrive. The thinned region eventually opens as a void, breaking the connection — or the accumulated material forms a hillock that bridges to a neighbouring line and shorts it. Either way the chip fails, and it fails after a period that is statistically predictable rather than random.

That predictability is why electromigration is a design constraint rather than a mystery. Black's equation relates median time to failure to current density and temperature, and chip designers use it to derate interconnects so that the expected lifetime exceeds the product's intended life. It is one of the reasons the industry moved from aluminium to copper interconnects, and why every shrink in feature size forces another look at current density — because thinner wires carrying similar currents means higher density, which means shorter life unless something else changes.

Dendrites: metal that grows in water you cannot see

The third mechanism needs three things present at once: moisture, a voltage difference and ionic contamination.

Under those conditions, metal dissolves at the positive conductor, travels as ions through the thin film of adsorbed water on the board surface, and plates out at the negative one — growing a branching, tree-like conductive structure back across the gap. When it arrives, the circuit shorts. This is electrochemical migration, and unlike whiskers it needs no compressive stress; it needs an electrolyte, and a humid board with flux residue on it is an electrolyte.

This is why cleanliness after soldering is a reliability requirement rather than cosmetic, why ionic contamination testing exists, and why humidity in service is part of a design assessment. It is also increasingly relevant as spacing between conductors shrinks, since the distance the dendrite must travel shrinks with it.

The same physics has an important cousin elsewhere: lithium plating in a battery, described in our article on why batteries age, deposits metal unevenly for related reasons and can grow structures that reach the other electrode. Different chemistry, same underlying picture of a metal being dissolved in one place and deposited somewhere it was not wanted.

Three mechanisms, three driving forces — mechanical stress, electron momentum, electrochemical potential — and one conclusion. A conductor is not a fixed object. It is a population of atoms held in place by nothing more than the absence of a reason to move, and engineering reliability means accounting for every reason you have given them.

Why this is hard to test for

Tin whiskers are unusual in that the industry has a well-developed test standard and limited confidence in what it predicts.

The difficulty is that whiskers grow slowly, erratically, after long incubation periods, and their formation depends on stress states that are difficult to reproduce. Accelerated testing — elevated temperature and humidity, thermal cycling — does produce whiskers, but the acceleration factor between test conditions and field conditions is not reliably known, and a finish that performs well in a chamber can still produce whiskers in service years later. Some conditions that accelerate other failure modes actually relieve the compressive stress that drives whiskers, so a harsher test can look better.

The practical result is that high-reliability programmes manage whisker risk through material selection and design rules — avoid pure bright tin, use a nickel barrier, maintain conductor spacing, use conformal coating, and where the stakes justify it, use tin-lead — rather than by testing their way to confidence. That is an unusual and honest position for a mature industry to be in, and it is worth knowing about precisely because it is not how most reliability problems are handled.

Why it matters for students and researchers

Reliability physics is taught far less than design, and it is where products actually die. A graduate who knows that a solder finish has a failure mechanism, that interconnects have a current-density budget, and that flux residue is an electrolyte will avoid a category of expensive mistakes that do not appear until years after a product ships.

Tin whiskers are also a valuable example of something students are rarely shown: an unsolved problem inside a completely mature technology. Soldering is thousands of years old, tin plating is a commodity process, and the mechanism by which a tin film extrudes a single crystal filament is still an active research question with competing explanations. Not everything old is settled.

The open problems are real and consequential. A predictive model that connects plating chemistry, grain structure and stress state to whisker propensity would remove the need for conservative blanket rules. A lead-free finish with demonstrated long-term whisker immunity would let high-reliability sectors finally complete the transition. Accelerated tests with a trustworthy relationship to field behaviour would change how the industry qualifies materials. And as conductor spacing continues to shrink, the margin against all three of these mechanisms narrows, which makes the questions more pressing rather than less.

Frequently asked questions

Is lead-free solder worse than leaded solder?

For most consumer electronics it is entirely adequate, and the environmental and health case for removing lead is strong. It does carry real trade-offs: higher processing temperatures, different joint microstructure, and the return of tin whisker risk that leaded finishes had effectively eliminated. That combination is why space, aviation, defence and some medical applications still hold exemptions and continue to use tin-lead for long-life hardware.

Can I prevent whiskers on a board I am building?

You can substantially reduce the risk. Avoid bright tin finishes, prefer a nickel underlayer beneath tin, consider finishes other than pure tin where the application allows, keep conductor spacing generous where you can, and apply conformal coating on anything that must survive many years. None of this is a guarantee, which is the uncomfortable part, but the difference between a well-chosen and a poorly-chosen finish is large.

Does electromigration mean my phone has an expiry date?

In a sense, yes, though it is rarely the thing that ends a phone's life. Chip designers size interconnects so that electromigration failures fall well beyond the expected service life at normal operating conditions, so batteries, screens, software support and physical damage almost always end a device first. Sustained high temperature and sustained heavy load do shorten the margin, which is one more reason heat is worth avoiding.

Why does a conformal coating not simply solve the whisker problem?

Because it addresses the consequence rather than the cause. The stress in the tin film is still there and the film still wants to relieve it, so whiskers can grow beneath the coating, penetrate a thin or unevenly applied layer, or in some cases bridge over the surface. Coating meaningfully reduces risk and is widely specified for exactly that reason; it does not eliminate the mechanism.

Why does flux residue matter if the board works fine when tested?

Because electrochemical migration needs time, humidity and bias — and a factory test provides none of those. A board that passes today can grow dendrites across a fine gap over months in a humid environment, with ionic residue acting as the electrolyte. This is why cleanliness is verified by ionic contamination testing rather than by whether the unit powers up, and why "it worked at final test" is not evidence about year three.