Leftover Gold in Ceramic CPU waste

[NobleMetalsRecovery]

I started recovering gold from electronic scrap 15 years ago. I have been saving the broken pieces from the CPUs. This consists mostly of 486s and early Pentiums. I have piles scattered about here and there. Today while cleaning out a warehouse that I rent, I came across about 100 pounds of the CPU broken fragments. A number of them have visible gold left where the silicon wafer detached from the ceramic.

A lot of these I purchased from another CPU refiner. He broke them up smaller than I did, and soaked them longer in AR than I did as well. I'm wondering how much gold could be left in this scrap? Anyone have a rough estimate?

 

[butcher]

Kind of depends on how finely broken means, the fine gold wires in small pieces may have a little left, I would regrind some to fine powder and test them, a batch can tell you.

 

[NobleMetalsRecovery]

Small meaning average size 1/4 to 3/8 inch.

How should I grind them? Make a ball mill?

I'd like to hear from someone that has processed ceramic broken up CPUs, then later gone back and ground them up and reprocessed them.

Thanks in advance. Steve

 

[butcher]

I do till fine powder, not as much yield smaller they get though, but I try till they squeal, then mix the powder up with my refractory for homemade furnaces or insulation. Plumbing pipe and caps makes a good cpu crusher, ball mill for very large batches, screening sizes (classifying) helps to separate sizes, making crushing easier if by hand. I'm liking Acid peroxide and then HCl bleach method, but sometimes still use nitric Aqua regia on them.

 

[goldsilverpro]

The Au/Si eutectic braze (if it was used) under the silicon chip, is a major source of the gold originally on these parts. When in AR, the braze underneath dissolves and the chips fall off or, at least, can be slid around on the pad. When I ran these parts, this was the indicator of when all the gold was dissolved. The gold under the chip is usually the last thing to dissolve, since the AR can only penetrate from the edges of the chip. Grinding the parts breaks up the chip and the underlying ceramic. This increases the chip edge length and speeds up the dissolving. If none of the chips remain stuck to the pad and if all of the base metals have been totally dissolved, I would think that very little, or none, of the gold would remain. If base metals are still present, there could, of course, be cemented gold powder present.

There are other non-gold brazes that can be used to attach the chip. Some can oxidize or tarnish and turn yellow, giving a false indication of gold.

I would first process 2 or 3 fairly large representative samples by boiling them in AR and then testing and recovering any gold that might be present.

Gold was originally present under the chip, under the lid, in the bonding wires, and in the plating. All can be dissolved, along with the base metals, with hot AR, even without grinding. I see no reason to further grind the parts.

I sometimes first ran the whole parts in hot 50/50 nitric, until all the base metals were dissolved. I then removed the nitric and dissolved the gold in AR.

These are just my opinions based on what I have experienced. I can't see the material. Can you post close up photos?

 

[NobleMetalsRecovery]

I will add photos as soon as possible.

I always thought there was gold all throughout the ceramic. Not in the form of wires, put a plated on "trace", like copper on a circuit board. The electrical path from the legs to the silicon wafer in the center needs to be some type of electrical conductor. I've always assumed it to be gold (I don't know why, maybe wishful thinking).

If that's the case, then most of the ceramic chip would have gold traces in it.

 

[goldsilverpro]

The traces, between the ceramic layers, going from the legs to the inside bonding fingers, are typically moly-manganese. In a paste form, which contains about a 5% glass frit. It is silk screened onto the ceramic and fired. The glass flows and bonds it to the ceramic, which also contains a glass frit (to hold the ceramic particles together).

Look at the inside bonding fingers, if they are still there. When the gold is removed, you should see a dark grey metal (Mo/Mn). If it's tarnished, remove the tarnish with a pencil eraser. If you have a new part, erase, scrape, or sand the gold away and look at what's underneath with an eye loupe.

 

[NobleMetalsRecovery]

Thank you for the excellent research and for dispelling a myth of my own making.

Does that mean that the similar looking traces in white ceramic chips are also moly-manganese?


One thing that lead me to believe they were gold is the high gold yield in some of the TCM CPU chips made by IBM.

How did you learn of the moly-manganese traces?

 

[goldsilverpro]

I owned a business that recycled side braze CPU packages for reuse. Our main customers were Intel and AMD. We processed about 100,000 of these each week. I had to know the construction of these packages inside and out.

That was a long time ago and changes could have come about. However, I was involved in refining about 500# of CPUs per day, a few years ago, and there was no change in what I saw. Take a new part, break it, and examine the edge of the trace with a loupe. If it's yellow, it's gold. If it's grey, it isn't gold. If gold was used for traces on certain parts, it's only purpose would be improved conductivity. In this case, it would have to be very pure gold, which could only be yellow.

The gold plating on these packages is 999.9 pure and is several times thicker than plating on fingers. The chip and lid are brazed on at high temperatures and lesser gold would discolor.

I am talking about the construction of the package itself, before mounting the chip and lid. The IBM parts are fully assembled. Most probably, 80/20 Au/Sn braze was used for the lid and about a 96/4 Au/Si braze for the chip. Probably 70 to 80% of the gold value on an assembled part is found in these brazes. A 40 lead side braze CPU, containing these Au brazes, is worth about $200/pound. The smaller ones can be worth $400 per pound, or more. The newer CPUs have much more ceramic and are, therefore, worth less per pound.

After writing the above, I looked up the IBM TCM chip. I have never processed these, so I can't really comment on them.

 

[NobleMetalsRecovery]

Excellent. That helps.

 

[FrugalRefiner]

The moderators would like to thank all of the members that contributed to the original thread upon which this thread was based, as well as those that asked questions showing what was missing. We encourage all members to read, comment, and ask questions in the original thread, Leftover Gold in Ceramic CPU waste.

The Library threads should not be considered to constitute a complete education. Instead, they're more like reading a single book on the subject of recovery and refining. There is so much more information on the forum, and it is impossible to include it all in these condensed threads. Members are strongly encouraged to read the rest of the forum to round out their education.

For those who prefer a printed copy, a pdf file of this thread is provided below.