Electroless Copper on High-Frequency PCB: The 0.3 Micron Film Everything Else Depends On

Every other wet step in a PCB shop works on metal that is already there. Electroless copper is the one exception: it has to turn a fluoropolymer wall into a conductor by chemistry alone, and it has to do it with a film thinner than a wavelength of visible light. That film carries no current in the finished product — it is an electrode, not a conductor — but if it is missing anywhere along the barrel, the electrolytic step that follows cannot fill the gap, and no inspection after plating will find the void until the board is in a thermal chamber.

Drilling leaves a hole. Etching leaves copper where the pattern asked for it. Electroless copper is the step that has to create metal where there was none, on a surface that is by design an insulator — and on PTFE, an insulator with the lowest surface energy of any laminate in the shop.

Its purpose is narrow and easy to over-specify: make the wall conductive enough that the electrolytic tank has a continuous electrode to build on. That is a coverage requirement, not a thickness requirement, which is why the important measurement here is a backlight rating and a cross-section at the hole centre, not a thickness number.

This is the first of the two plating steps described in the plating guide, and it follows the drill and desmear covered in the drilling guide.

The seed is 0.25–0.5 µm thick, deposited at roughly 2–3 µm per hour. Those two numbers together explain the whole step: the process takes minutes, and it has almost no margin for a surface that does not accept it.

The sheet-resistance figure is the useful mental model: at 0.056 Ω per square, a 5:1 barrel is only a few squares, so the voltage drop during electroplating is negligible. The seed is not there to carry current efficiently. It is there to be continuous, and continuity is a binary property that either survives the whole wall or fails at one point.

This is also why the common claim that electroless copper is 1–2 µm thick does damage. At 2.5 µm/h, 1.5 µm is 36 minutes of bath time instead of 7, on a bath that is being consumed and dragged out the whole time — and the extra copper is the least ductile metal in the stack, sitting exactly where the barrel has to flex.

Alkaline degreasing removes oils, fingerprints and resist residue; a rinse follows; and on a high-frequency build the conditioning step also has to make the wall wettable. A fluoropolymer is not wetted by water-based chemistry to begin with, so a bath that beads on the wall cannot deposit anything on it.

The check is the same one used after inner-layer stripping: a water-break test. A properly conditioned wall holds a continuous film for about 30 seconds; a contaminated or non-wettable wall breaks into droplets immediately, and everything downstream inherits that failure.

Micro-etching does two things: it removes the oxide and handling film from the copper, and it roughens the copper so the seed can key into it. The chemistry is sodium persulfate or hydrogen peroxide with sulfuric acid — the peroxide or persulfate is the oxidiser that actually dissolves copper, and the acid keeps the dissolved metal in solution.

Dilute sulfuric acid on its own does not micro-etch copper. It cleans it. A line built on acid alone produces a bright, oxide-free, perfectly smooth surface that the seed adheres to poorly, and the failure appears later as a barrel that separates from the wall under thermal stress.

Depth matters in both directions. One to two micrometres of copper removal is normal. Beyond that the etch starts to eat the 17 µm inner-layer copper that the inner-layer step just finished controlling, and it changes the conductor thickness that the loss model assumed.

Activation deposits palladium catalyst on the dielectric. Without it, the electroless bath has nothing to reduce onto, and the wall comes out of the line exactly as insulating as it went in. The catalyst is applied as a colloidal suspension, and the colloid is the part that causes most of the confusion downstream.

On FR-4 the sequence is well behaved: condition, micro-etch, pre-dip, activate, accelerate. On PTFE it is not, because the surface has no polar chemistry for the colloid to attach to. The wall has to be modified first — sodium naphthalene etching or plasma — so that the palladium has somewhere to sit. Skip it and the activation step looks normal, the electroless step looks normal, and the cross-section shows a seed layer that is present on the surface and absent in the hole.

Catalyst concentration and immersion time are set by the surface area being processed, not by a fixed dwell. A load of small-diameter, high-aspect-ratio holes presents far more wall area than the same panel area of large holes, and a bath dosed for the latter will under-catalyse the former.

The palladium catalyst is delivered as a tin-palladium colloid, and each particle is wrapped in a tin hydroxide shell. That shell is what keeps the colloid stable in the bath, and it is also what would insulate the palladium from the electroless bath. The accelerator — usually a dilute acid or complexing solution — strips the tin shell and exposes the palladium core.

So the accelerator is not removing excess palladium, and treating it as a palladium stripper leads to the wrong control decision: an operator who thinks the step removes catalyst will extend the dwell to “clean up” the surface, strip the colloid entirely, and produce a wall with no active sites at all. The control parameters are concentration and time, held inside the supplier window, with the failure mode being under-acceleration (tin left in place, slow start) rather than over-exposure.

Electroless copper is a catalysed reduction, and like most such reactions its rate roughly doubles for every 10 °C. Production baths run at 45–50 °C for that reason.

The failure mode of a cold bath is not a thin seed — it is a seed with the wrong structure. A slow, low-temperature deposit is fine-grained, heavily stressed and prone to the same adhesion failure as an over-thick one, and the line takes five times as long to produce it. Running at 20–30 °C to “be gentle” on a high-frequency laminate is a misunderstanding of what the bath is doing.

Loading, copper concentration, formaldehyde or hypophosphite concentration, pH and stabiliser level all drift with use, and all of them affect the deposit rather than the thickness alone. Analysis by titration at a defined interval, with a Hull-cell-style deposit check, is what keeps a bath inside its window; a bath controlled by the calendar will pass dimensional checks and fail adhesion.

The measurable output of this step is coverage, and the standard measurement is backlight: a plated cross-section is examined against a light source and rated on how much light passes through the barrel. A rating of 8–9 out of 10 or better means the seed is continuous; anything less means there is a path with no copper on it.

Coverage degrades with aspect ratio for the same reason throwing power does — the chemistry inside the hole is depleted faster than it is replenished. At 6:1 on PTFE, the wall is at the edge of what a standard line covers reliably, which is why the drilling step caps aspect ratio rather than the plating bath being pushed harder.

Three mechanical aids buy coverage without changing chemistry: panel movement or vibration in the tank, which exchanges solution inside the holes; air knives or squeegee rollers between baths, which prevent drag-in and carry-over; and a pre-dip that matches the activator chemistry so the colloid is not shocked at the interface.

Most seed voids are not chemical failures. They are bubbles: air trapped in a blind hole, a gas pocket clinging to the upper wall of a horizontal barrel, or hydrogen evolved during deposition that stays attached because the wall is not wetting properly.

Countermeasures are all physical — orientation of the panel so that gas can escape upward, vibration, surfactant in the bath, and a wetting step before activation rather than relying on the alkaline cleaner. On a small via in a thick panel the bubble has nowhere to go, which is why aspect ratio appears in this article for the third time.

Because the defects are invisible, the discipline is statistical: backlight panels per shift, cross-sections at the centre and at both ends of the barrel, and a thermal shock sample from the same load. A batch that passes continuity testing proves only that the barrels that were tested were continuous.

Electroless copper is a ten-minute step that decides whether a twenty-five micrometre barrel exists. Its output is coverage, its enemy is a surface that will not accept a colloid, and its most common error is being asked to deposit four times too much copper at half the correct temperature.

The claims worth discarding are that the seed is 1–2 µm (it is 0.25–0.5), that the bath runs at room temperature (it runs at 45–50 °C, and every 10 °C halves or doubles the rate), that dilute sulfuric acid micro-etches copper (it needs an oxidiser), and that the accelerator removes palladium (it removes the tin shell around it).

XCE PCB runs electroless copper on Rogers, Taconic and F4B builds with sodium naphthalene or plasma modification, backlight verification per shift and thermal shock from the production load: 2–72 layers, impedance ±10%, panels to 600 × 1200 mm, 5,000 m² of production space and 80+ engineers, with samples in as fast as 24 hours.