Ion contamination in IC's

[KR8K]

Hello,
I'm new to this forum. I'm not an electrical engineer, but I work in an investigative capacity in the automotive industry. Sorry if this sounds a little disjointed, but some of the information is proprietary and since I'm responsible for the "fix" when it happens, I need some advice. We have this IC located in a module on a car. The IC contains a logic and power board and amongst other things contains about 8 high side drivers, which is to say it controls B+ (12 volts) to lamps in the car. Due to an unfortunate series of events, one of the high side drivers that run the largest load is too small. After a period of inrush current events from the lights turning on and off, the driver gets damaged by the current, damages the control board and the chip can no longer protect itself from hi temp or over current. The resulting fire is quite spectacular. There is a current limiting bit, that can be set to limit peak current, this setting was set to limit which also causes a large amount of heat that must be dissipated due to the inrush current when the bulbs turn on. Between the bit setting and the sizing of the high side driver, we have a problem.

There is an unused high side driver that can handle the load, but due to another incorrect setting, there has been open load detection current (about 100 micro amps) that has been running through this unused driver for at least 25-35K miles. I can't equate that to inrush events because it depends on how often the doors are opened. The chip supplier came forth last week and told me that I should be checking the leakage and voltage drop of that FET before swapping the circuits. They're calling it ion contamination, caused by the leak detection current and know that theoretically it can happen. We've never seen one that has an issue, but then the only way it can be detected is on the manufacturer's end of line tester.

What I need to know is, has anyone ever heard of ion contamination in this type of situation? Is what the supplier is saying make sense? How likely is it that I may see this in this part? They're claiming that a contaminated part could cause voltage drop across the FET, localized heating and a possible thermal event. Read fire.
If I have to check this FET for ion contamination it is going to be a major deal. Is it real or fiction?

Thanks
Eric

 

[IRstuff]

There may be such a thing, but it wouldn't caused by a piddly 100 uA, no matter what else was going on. Sounds like a load of horse manure, but of the finest kind, of course.

While they full of you know what, the devices have been powered up and thermally excited, so they may have some reliability fails, due to aging.

TTFN

 

[KR8K]

Thanks for the reply, here's a little more information...The manufacturer of the chip admits that the problem is not the 100 micro amp current flow, but the fact that this pulls the gate up to near battery potential, which is 12 volts for discussion purposes (it could go as high as 14.5 with the engine running. The fact that the gate is near battery potential and is in a reverse biased condition is what causes the ion contamination. Does this make any more sense or do you stand by your first opinion?

Thanks
KR8K

 

[IRstuff]

No, a biased-gate can't CAUSE an ion-contamination, since that's all sealed under passivation on the chip. It can AGGRAVATE existing ion contamination.

A typical example would be some sort of contamination occuring during the gate oxidation process in the IC manufacturing. A bias on the gate would force all the ions to move toward or away from the gate, thereby affecting its external threshold voltage.

The bottomline line is they are attempting to blame you for a processing problem in their fab.

TTFN

 

[KR8K]

IR stuff,
Thanks for the reply. The battle rages on and I have another conference call with them on Tuesday. They admit that they did have some ion contamination in their fab process but they never saw it in this particular application of their chip. This is the same chip just in a different component. They instituted a "barrier" layer that is supposed to take care of that, in addition to cleaning up their process that caused the contamination in the first place. What you're saying makes sense to me, in that the ions are already there, possibly due to their process and now are trying to make sure that severely contaminated IC's don't end up in a thermal event. They claim to be doing some testing, the details of which I haven't seen yet. They are recommending a leakage test and voltage drop across the gate at 2X recommended current flow. The box that enables this test hasn't been built yet and as far as I know is still in the design stages.

Eric
KR8K

 

[kenvlach]

The mechanism is called electromigration.
Google gave me 79,900 hits for electromigration circuit defect:

http://www.google.com/search?source...-41,GGLG:en&q=electromigration+circuit+defect

 

[IRstuff]

Electromigration is due to high currents in the aluminum conductors transferring their momentum to the aluminum atoms, creating voids along crystal boundaries.

Electromigration can manifest as open circuits in the aluminum as well as shorted junctions, where the aluminum as migrated to the point of migrating into the junction boundary area itself, approximately 0.5 microns under the surface of the silicon.

If the transistor in question was under low current state, it shouldn't be electromigration. Most high current processes use an aluminum/copper alloy or copper alone to prevent electromigration.

That said, however, the OP's last posting of using barrier materials and thermal events certainly seems to support the possibility of electromigration. Barrier materials such as Titanium Nitride are often used between the metal conductors and the PN junctions to prevent aluminum from migrating into the silicon.

Unless, of course, the barrier material is something like a silicon nitride layer within the gate structure. The problem is that silicon nitride is a discarded non-volatile memory technology. There's some possibility that if your supplier is using a nitride/oxide gate insulator, that your long-term gate bias voltage might might have mimicked the programming conditions for a non-volatile memory, therebyu causing a negative threshold shift that might be similar to ion contamination. The test for voltage drop across the gate would seem to implicate the gate dielectric itself, as opposed to electromigration in the aluminum.

My data is pretty old, since I worked in the metal nitride oxide semiconductor (MNOS) memory devices more that 20 years ago. MNOS memories faded away about 15 years ago and it might be possible that your supplier has no one who is aware of the danger of using a nitride/oxide composite gate dielectric.

Since there isn't enough detail, it's hard to tell what the real story is. The OP is going to have to spend some more calories to get the full story from the manufacturer. I would start by asking for more specific details about the "barrier" material and exactly where it's located.

TTFN

 

[kenvlach]

FYI: ASM Int. has a clearance-priced book which may be helpful:
Materials and Processing Failures in the Electronics and Computer Industries: Analysis and Prevention
Author: A.S. Brar and P.B. Narayan
ISBN-10: 0-87170-468-4
Published: 1993
Pages: 330
#25. + s/h

http://www.asminternational.org/Tem...=Ecommerce/ProductDisplay.cfm&ProductID=11030

Look at other books (can save on shipping). Maybe:
Microelectronic Failure Analysis Desk Reference 2002 Supplement (Book and CD ROM) $48 (non-members):
You can browse the TOC & preview a chapter:

http://www.asminternational.org/Tem...=Ecommerce/ProductDisplay.cfm&ProductID=12021