Copper Interconnects For CPU's-How bout Silver?

[Howard]

I certainly didn't expect posts like Sixtyfour's in this forum.

 

[mordantmonkey]

Originally posted by: Howard
I certainly didn't expect posts like Sixtyfour's in this forum.

i take it you didn't look at his postcount/join date?
he thought he knew what he was talking about. but didn't. now the thread has gotten horribly off track.

 

[GimpyOne]

A couple of thoughts on the gold comments. First is that gold is extremely difficult to corrode.

When you are looking at old jewelry that appears corroded, it is usually for 1 of two reasons. Either:
1. It really isn't "gold" but rather golden colored brass, or
2. It is very impure gold and the corrosion is actually due to the other junk corroding
(as a correlary to this, pure gold from shipwrecks looks like the day it sank)

And yes I have degrees in both Chemistry and Materials Science, so I do know this stuff.

Now back on topic, gold is actually being studied to be used in interconnects, but there are problems associated with it. Namely, gold is relatively weak and the fibers drawn from it can be difficult to work with. Additionally, heat cycling can cause exaggerated grain growth of gold allowing the wires to fall apart. Not something you want to happen within a package.

 

[bobsmith1492]

Neither gold nor copper corrode from what I hear....

 

[GimpyOne]

Originally posted by: bobsmith1492
Neither gold nor copper corrode from what I hear....

Actually copper does, but it forms a protective surface oxide layer which prevents further corrosion.(thus the color of the statue of liberty!) So after it's initial reaction it's ok. As a note, this usually requires oxygen to be readily available, one of the reasons copper works well for plubming, as long as you don't continually empty the lines.

Now, this layer, as much as anything can be a problem. Even if only the surface reacts you will lose conductivity overall. On a normal wire that isn't a problem, but if you are looking at micron sized wires it suddenly becomes a large fraction of the overall part.
Part of the reason this isn't a problem with most chips is becuase the packaging effectively seals them off from the ambient atmosphere.

 

[imported_Kensai]

Originally posted by: GimpyOne
A couple of thoughts on the gold comments. First is that gold is extremely difficult to corrode.

When you are looking at old jewelry that appears corroded, it is usually for 1 of two reasons. Either:
1. It really isn't "gold" but rather golden colored brass, or
2. It is very impure gold and the corrosion is actually due to the other junk corroding
(as a correlary to this, pure gold from shipwrecks looks like the day it sank)

And yes I have degrees in both Chemistry and Materials Science, so I do know this stuff.

Now back on topic, gold is actually being studied to be used in interconnects, but there are problems associated with it. Namely, gold is relatively weak and the fibers drawn from it can be difficult to work with. Additionally, heat cycling can cause exaggerated grain growth of gold allowing the wires to fall apart. Not something you want to happen within a package.

:thumbsup: Now that's what we needed. 🙂

 

[f95toli]

A few more comment about silver.
I discussed this with a colleague of mine today and he pointed out that one problem with silver is that it will react with silcon oxide (forming silveroxide) wheras copper won't.
I also learned that another of my colleagues actually use silver to form contacts on ferroelectric films (he is working with oxides so I guess it is SRO), everyone else is using gold so I don't know why he choose to use silver instead (he was not around so I could not ask him).

A general comment: It is in general difficult to draw conclusions about how a material will behave when deposited as a thin film just from bulk properties because they depend on grain structure etc.
The properties of a thin film also strongly depends on how it was deposited, e.g. the best way to deposit gold contacs is usually sputtering, e-beam- and thermal evaporation results in gold with different propertie. It is mainly the interface properties (such as contact resistance) that change but even the color can be different.

The properties also depend on what you deposit on. I have actally worked on a material (PCCO) where we could not use gold to make contacts , the contact resistance was huge even though the mechanical sticking was ok (at low temperatures the contact turned into a Schottky diod).
We ended up having to use indium contacts (indium is a VERY messy material to deposit).

 

[onix]

Sixtyfour's posts up to this point have been mostly correct. I do IC design and gold is often used. In fact a lot of gold wire-bonding is done too, and that conceivably would also be "expensive". You guys have to look at the overall cost.

Now aluminum is the traditional metal used. Gold is a better conductor and it lasts longer, BUT it is highly incompatible with most CMOS processes, and hence its use is limited.

I am still interested in the question originally asked. Copper at one time was also difficult to fabricate IC's with, so I wonder now what ails Silver.

 

[f95toli]

What diameter are the bond wires used in commerical ICs?

At work I use a semi-automatic bonder from Kulicke&Soffa, if I remember correctly it uses 1 mil gold wires. I usually bond ro 44 pads which means that I use about around 10 cm of wire to bond a single chip, that is probably enough cold to deposit interconnects on many chips.

Does anyone know if the interconnects are deposited on oxidezed Si? In that case it would explain why Ag is not used since the Ag would react with the SiO as I have already mentioned.

 

[BitByBit]

Mmm, perhaps AMD could introduce a new FX 'Gold edition' line, for stratospheric overclocks. 🙂

 

[Eskimo]

Does anyone know if the interconnects are deposited on oxidezed Si? In that case it would explain why Ag is not used since the Ag would react with the SiO as I have already mentioned.

No, in both aluminum and copper processes a barrier metal(s) is usually deposited onto the ILD first. Ti and TiN are the most common. ILDs are usually some form of oxidiezed silicon (BPSG, TEOS, CDO, etc). Most low k materials in use today don't play well with copper. If you are directly contacting silicon you'd likely be using W since you have the obvious contamination issues with copper and aluminum can spike through silicon.

I am still interested in the question originally asked. Copper at one time was also difficult to fabricate IC's with, so I wonder now what ails Silver.

See my post above. If you've ever done any design on high power, high frequency devices you'll realize that the EM concerns alone would pose a major problem. And high power, high frequency applications are exactly where silver would be used, i.e. MPUs. Plus there are the process integration issues which still require more research.