Contact Us
  • Home
  • BLOG
  • Which Surface Finish Works for Wire Bonding?

Which Surface Finish Works for Wire Bonding?

The finish decides whether a wire bond holds, and the answer splits by wire. Gold ball bonding needs soft gold of 0.5 micron or more, or ENEPIG with a palladium layer; aluminium wedge bonding needs ENIG or ENEPIG with the gold held between 0.05 and 0.1 micron. Hard gold belongs on a connector finger, and HASL, OSP and immersion tin do not belong on a bond pad at all.

A bond is a solid state weld made at 150 to 240 C with ultrasonic energy, and it needs an oxide free, uniform, soft metal surface inside a narrow thickness window. The finish is the only part of that the board supplier controls, and it is usually specified last.

The four finishes and the window each one gives

ENIG. Nickel of 3 to 6 microns over copper, which is 100 to 200 microinches, with gold of 0.05 to 0.1 micron from displacement plating. Aluminium wedge bonding works well on it, because the aluminium meets nickel through a gold layer that is only a few atomic layers thick. For gold ball bonding ENIG is marginal: there is not enough soft gold for the ball to deform into, so the weld forms on nickel, and nickel oxide is what limits the pull strength.

ENEPIG. Nickel of 3 to 6 microns, palladium of 0.05 to 0.15 micron, gold of 0.03 to 0.1 micron. The palladium is a diffusion barrier. It keeps nickel away from the gold surface, which is the mechanism behind black pad, and it keeps the gold from diffusing into the joint during service. It accepts both gold and aluminium wire and survives several reflows, which is why it is the choice when one board carries bonded die and SMT parts.

Soft gold. 0.5 to 1.5 microns of pure gold over nickel. This is the gold ball bonding surface at fine pitch, because a ball needs softsh into. It costs more than ENEPIG and does not like repeat reflow.

Hard gold. Gold with a cobalt or nickel hardener, Knoop hardness above 130, 0.8 to 2.5 microns. It is made for edge fingers that slide into a socket. The hardener is exactly what makes a bond unreliable: the gold does not deform under the ball and the cobalt migrates to the interface during the weld.

Why thickness is the number that decides

Thin gold is not automatically better and thick gold is not automatically safer. At 0.05 to 0.1 micron the gold is consumed by the aluminium during the weld and the joint becomes an aluminium nickel joint, which is what a wedge bond wants. Push the gold past 0.2 micron without palladium underneath and the gold and aluminium keep reacting at 150 C: the gold is consumed, voids open on the gold side, and the joint fails in pull after a thousand hours with no visible change on the surface. This is the Kirkendall mechanism, a slow failure rather than an immediate one.

Below 0.03 micron the gold is porous enough that nickel oxide forms inside the pores before the bond is made, and black pad becomes possible: a corrosion layer on the nickel that looks black after a pull, is brittle, and fails between 1 and 2 grams on a wire that should hold 3 or more. Specify a range rather than a target, and ask for the thickness on the cross section report rather than the plating tank log.

Geometry follows the same logic: a 25 micron gold wire needs a pad opening of 2 to 3 wire diameters, so 75 to 100 microns, and the ball lands at 2.5 to 3 times the wire diameter. A mask defined pad has less copper than a copper defined one, and the smaller pad craters first.

Where a bonded board fails first

The second bond. Parameters found for the ball bond do not transfer to the stitch, and the pad under the stitch has less copper and less support. A board that passes the ball pull and fails the stitch usually has a finish problem at the pad edge, where the gold is thinnest.

Black pad on nickel. It starts in the plating bath and appears months later, which is why a supplier that reports nickel thickness and no bond test tells half the story.

Contamination applied after plating, and storage. Flux residue from an earlier step, silicone from a conformal coating, a fingerprint on a pad handled without gloves: all of them stop the weld and none are visible at 10 times magnification. Gold does not oxidise, but the nickel under a thin gold layer does, and six months in a humid room changes a pull result.

Three mistakes that show up in a first build

Treating gold as one material. Ordering 30 microinches of hard gold for a bond pad because the same spec worked on a connector finger. The hardener that makes the finger survive 500 insertion cycles is why the bond fails at 1.5 grams.

Thickening the gold to be safe. Moving from 0.1 to 0.4 micron over nickel, without palladium, to add margin. It adds a Kirkendall void instead, and the failure appears at the reliability test, not the line.

Leaving pad definition and cleanliness out of the purchase order. Which layers define the pad, the ionic cleanliness limit and the packing are free to specify and expensive to correct after the die is attached.

What SUPERB PCB supplies for a bonded board

ENIG at 100 to 200 microinches of nickel and 2 to 3 microinches of gold, selective ENIG, ENEPIG, immersion silver, immersion tin, lead-free HASL, and hard, soft or flash gold where the panel also carries an edge connector. Minimum line and space is 0.05 mm, laser drilling starts at 0.075 mm, BGA pads go to 8 mil on ENIG, the aspect ratio is 10 to 1 on the standard line and 25 to 1 on the high layer count line, and every panel ships with a cross section report.

The bond itself runs at the package house or at the customer assembly partner. What leaves here is the finish, the coupon data behind it, and a panel that has not been handled after plating. Where a board carries both bonded die and SMT parts, the finish is chosen once for both, and ENEPIG is usually the compromise that suits the aluminium wedge bond and the solder joint next to it. The substrate side is on IC substrate fabrication, the general window on PCB capability, and the high reliability route on high reliability PCB.

One honest limit: gold thickness varies across a production panel, and 0.05 to 0.1 micron is a real range, not a rounding of 0.1. If the bond process needs 0.1 micron everywhere, that is a tighter control and it is quoted as one.

What to send

Send the wire material and diameter, the bond pad opening and pitch, whether the pad is defined by mask or by copper, the gold thickness range and the palladium thickness if the finish is ENEPIG, the number of reflows the panel will see, and the storage interval before bonding. That is enough to fix the finish before the first panel is plated.

Need a quote or a review of your files against these criteria? pcba@superb-tech.com · WhatsApp +86 133 9608 1443