How to Calculate Metal Widths to Avoid Electromigration

[nikogian]

how to avoid electromigration

Hi all,
I need some help with calculating the metal widths in my design in order to avoid electromigration.
My IC process provides me with the maximum allowed DC current density for each metal layer, VIA etc at 110 degrees Celsius. I have the following questions concerning the width calculation:

1) What about AC currents? the process gives me a rule about the allowed peak AC current density but what about steady sinusoidal currents? Should I calculate the width according to the effective current Iac/(2*pi) ? (Iac is the amplitude of the sinusoidal current)

2) When a wire transfers both AC and DC current, Should I just add the DC current and the effective AC current and then calculate the width?

3) What about the temperature. I don't think my circuit will ever reach 110 degrees while working. Is there a chance the the given DC current values are too pessimistic? If so, is there a way to calculate the maximum current density at a lower temperature?

4) I have read in Tom Lee's book that electromigration is not a very big problem at high frequencies. If I calculate the width for sinusoidal currents using the effective current Iac/(2*pi), it seems like frequency is ignored. How can I account for the AC current frequency while calculating the metal width?

Thanks in advance,
John

 

[mdcui]

current density

as my understanding, electromigration will only present when a signal trace always has a one-way current flow, so only DC current counts, all AC current (either on top of the DC compnent or absolute AC) shall be ignored, that is what we did in our chip design.
think of the principle of electromigration, the collision of electron on the atom core moved the atom, so if electron are bumping from opposite direction, averagely the atom will not move, so that is why AC current shall never be counted, it is not like metal melt-down machanism which is related with average square of I.

 

[mswug]

calaulate metal width vs current

mdcui is right~~
And for different temperature, there is a coeffient in DRC file. So you can get other temperature electromigration by the coeffient multiple 110 degree electromigration

 

[nikogian]

electromigration metal 1 width

Hi,
I think (at least as I read in a Chenming Hu's paper) that there are 2 mechanisms of electromigration. One is the electron momentum transfer, which can be ignored in high frequencies, and the other is thermal electromigration, which must be accounted for AC currents.
The problem is the effective thermal equivalent must be calculated for AC currents

John