Gold plating over Nickel on PCBs for wire bonding

"Gold Plating Technology"
by Reid & Goldie
(hard to find & expensive; if you
see a copy cheap, act fast)


on eBay or

AbeBooks

or Amazon

(affil links)

A. The term used a couple of years ago when I worked in a board shop was "deep gold". The meaning should be obvious: normally you only plate the tips (at the edge of the board, achieved by careful trimming).

With deep gold the whole board is immersed into the plating solution and plated just as you would plate the copper, tin, or tin-lead. The big difference is that the photoresist has to be much more resistant because of harsher chemistry.

I don't remember the specified thicknesses, but I think they were the same as on tips, e.g., 50-100 nickel with minimum 20 gold?

Hope this helps.

A. Don, it really depends a lot on the reason for the gold finish on the board. If it is just as a solderability preservative, then a relatively thin pure gold finish will normally work. Most in the industry prefer other finishes since gold will tend to reduce the reliability of solder joints...immersion tin, Benzotriazine, or solder plate/fuse are logical alternatives if solderability preservation is the objective.

If on the other hand the application calls for wire bonding, then relatively thick pure gold is required...normally at least 20 µin and more like 40 µin. Most of the time, due to the high cost of gold, masking is used for such an application to put the gold only in those regions that require wire bonding.

Definitely do not use hard gold if wire bonding or solderability are required ... you will have serious problems. Use only pure gold (i.e., 99.9% pure) for these applications.

Hope this helps.

Ted Adlam (from not quite so sunny but very humid Singapore)

Don,

Be careful about selecting electroless gold on you PCB's. Unless your plater & chemist are VERY familiar with the process and have A LOT of experience prepping the board, the E'less Au will end up growing laterally along the surface of the board instead of up from the deposit and your entire board will end up covered with gold. Although all E'less Gold suppliers will guarantee their Gold will build up 150+ microinches quickly, none of them will guarantee that the lateral growth will not occur. And it is thermodynamically the more favorable reaction by several orders of magnitude unless you know how to shut it down. Maybe an E'less Au techie out there will chime in with the answer to this pbm.

Good Luck
Dave Kinghorn, in neither hot nor humid VERY sunny Sunnyvale, CA

? What is the material that you are trying to bond with? Have you tried to burnish the copper surface? Can you attach a molytab with some kind of conductive epoxy then try to bond to the molytab?

A. Greg, This is a difficult question to answer. Much depends upon the surface roughness of the nickel as well as the substrate. It also depends upon the porosity of the gold. My experience comes from gold wire bonding to gold plating. In that case we found that 30 microinches was needed to eliminate any nickel oxide forming on the surface of the gold. We also observed that pulse plated gold was less porous than DC plated gold.

A. Pure gold (0.9999 which is what you need for wire bonding) can be plated at any thickness including 0.000 005 up thru 0.5 inch over ANY thickness of any kind of nickel (but the nickel must be fully "activated" for the gold to bond). Any statement to the contrary is not technically correct.

A. Hello Fellow Tank Jockeys:

This is a classic example of the difference between accuracy and precision. They are not the same thing. While it is factually accurate to state that "ANY thickness" of Gold can be plated over "any" thickness of Nickel, it is also accurate to state that you will plate out a "distribution" of thicknesses all over your part, and the grand average of all these thicknesses is your target mean...the aforementioned "ANY thickness" value.

A trained plater can target this mean value, say 75 microinches, quite "accurately." However when the very same trained plater measures this very same electroplate in, say, 100 different locations and plots a histogram of this data, a "distribution" of thicknesses will be observed. The "wideness" of this distribution is known as the "precision," and is characterized by a mathematical term known as the "standard deviation," or "variance."

Looking again at our 50-100 microinch callout example, the 75 microinch target mean, the grand average of all measurements, the "ANY thickness" value, can be hit exactly with little effort. Now look at the individual measurements. About 90-95 percent of them will be between 50 and 100 with most being between 60 - 90. 5-10 percent will be either high or low. The part fails, yet the "ANY thickness" value was hit perfectly. Why? Precision.

Most "typical" plating baths being comprised of a warmed bucket with anodes over the sides have a precision of ±50% of mean within ± 3 standard deviations. Now do the same experiment with the 100-400 callout. Plater targets 250 and hits it exactly. Plater takes 100 measurements and 100 are within the 100-400 callout. Same tank, the precision is the same, yet the part now passes. Why? The callout is now wide enough to fully envelope the plating processes' precision. And that's the difference between precision and accuracy, and also illustrates how just calling out a thickness is a dangerous way to specify and electroplate.

A. Bruno,
The reason PURE, Soft gold is better for wire bonding is because hard gold obviously has hardening agents known as alloys or brighteners. If the alloy contains nickel, cobalt, silver there will be additional intermetallic formations at the interface after bonding that will shorten the shelf life of the part, and many times fail. When I worked for a major precious metal supplier pure, soft gold was always recommended for wire bonding.
Good Luck.

Hi Mark,

Thank you very much for your answer. It's now clear to me. Meanwhile I bought the book "Wire Bonding in Microelectronics". If I understand well, the hardening agents work like impurities in the gold coating, leading to the "Horsting effect", namely intermetallics that precipitates in voids after some temperature change. Is that right?

Is there anywhere some case of studies, showing bad experience of hard gold?

Regards,

A. Hi Sangeetha,

Whilst wire bond failure may not necessary due to plating deposit but also wire bonding parameter setting, it seems your soft gold bath iron content is too high. Normally should be less than 25 ppm. Gold deposit purity should be at or above 99.95% by wt. Alternatively,you may reduce gold thickness by plate a thin layer of Pd over Ni follow by Au.

Regards,
David