Extremely high current effects in semiconductors

  • Thread starter ChrisP
  • Start date
  • #1
ChrisP
4
0
TL;DR Summary
Can high current alone damage a device (assuming the device is cooled adequately)?
So, let's assume in a theoretical case we have a (semiconductor but not necessarily) device, e.g. a MOSFET, that is perfectly cooled and has zero heat resistance (infinite thermal conductivity), and that it's connected to a perfect heat sink (theoretical abstractions for now, I will come back to the real world later). Let's say that the MOSFET is switching so it has transient behaviour, it operates in normal region (not sub- or near- threshold) and its gate is charged and discharged at a high frequency by a perfect driver (let's say an ideal inverter). You also have an ideal power supply that can deliver as much current as the circuit asks (with perfect line and load regulation), and the voltage never exceeds the dielectric breakdown threshold.

Now, my question is, can the device be damaged from very high current flowing through the gate driver? Again, not because of overheating, but because of physical limitations in the device? I know that high voltage causes dielectric breakdown, due to a high electric field exceeding the dielectric withstand strength of the material, but is there such a thing for very high currents besides heat? Because the only physical effects I know related to high current besides heating is electromigration which is a relatively slow process and affects mostly long term reliability.

The question is motivated by two thoughts: 1) when you have a real IC and you cool it down with liquid nitrogen, cooling is not really an issue, so the only limiting factor to achieving maximum frequency should only be breakdown voltage right? (I know that there are other limiting factors to speed relating to setup and hold requirements, etc. but I mean from a physics perspective. There is also the case where heat is just not removed fast enough due to the thermal resistance of the junction to package, package to heat sink, etc. and it still ends up blowing your device). 2) in a superconducting system where you don't have any heat being generated, what's limiting you from having billions of amps going through the system (assuming that there is a load that can draw so much current, i.e. it has something to do with all this current, and a source that can provide that)?
 

Answers and Replies

  • #2
Baluncore
Science Advisor
12,339
6,410
2) in a superconducting system where you don't have any heat being generated, what's limiting you from having billions of amps going through the system (assuming that there is a load that can draw so much current, i.e. it has something to do with all this current, and a source that can provide that)?
Semiconductors cease to function as semiconductors at superconducting temperatures, because they become short circuits. You will need to select your semiconductor material for low temperature. Consider using GaAs transistors with Si conductors and SiO2 insulators.

Semiconductor PN junctions have a voltage determined by bandgap voltage. 1000 amps through a 1 volt junction will generate 1 kW, even with zero series resistance.

High currents produce high magnetic fields. Hall effects will come into play. Specs will change.

High currents when switched rapidly, with some conductor inductance, will produce voltage spikes that will punch through junction insulation.
 
  • #3
Rive
Science Advisor
2,490
1,928
Again, not because of overheating, but because of physical limitations in the device?
Overheating (in various forms) is one of the the most basic physical limitation of devices.
You can have insanely (!) high pulse currents, as long as the mass involved is enough to keep the junction temperature in the working range.
 
  • #4
Tom.G
Science Advisor
Gold Member
4,743
3,499
You may or may not consider this a physical limitation, but it is an inherent limitation that manifests itself thru localized melting of the semiconductor chip itself.

SCR's (Silicon Controlled Rectifiers) have a maximum di/dt rating, beyond which they fail shorted. The turn-on of these devices starts in a few very localized areas of the junction and then spreads across the full junction area. If the current rises faster than the conductive area can spread, there is intense localized heating at these filamentatious areas... enough to melt the semiconductor chip.

Personally, I would consider that an inherent limitation because the chip itself can spread neither the current nor the heat fast enough to function.

Cheers,
Tom
 
  • #5
ZapperZ
Staff Emeritus
Science Advisor
Education Advisor
Insights Author
35,994
4,716
Wait. Don't semiconductors, by definition, become insulators at low temperatures? So what current? Tunnel current?

Zz.
 
  • #6
Baluncore
Science Advisor
12,339
6,410
Wait. Don't semiconductors, by definition, become insulators at low temperatures?
Some LEDs (used to set the bias voltages in RF preamps) become superconductors in liquid helium, just when you need them most.
 
  • #7
Rive
Science Advisor
2,490
1,928
Wait. Don't semiconductors, by definition, become insulators at low temperatures?
Semiconductor, as material indeed works that way (yeah: mostly :doh: ). Semiconductor, mentioned as component in electronics supposed to be in some kind of working conditions (unless specified otherwise).
 
  • #8
ZapperZ
Staff Emeritus
Science Advisor
Education Advisor
Insights Author
35,994
4,716
Some LEDs (used to set the bias voltages in RF preamps) become superconductors in liquid helium, just when you need them most.

They do? Or do you mean they have almost zero resistance at those temperatures? Having almost zero resistance does not automatically imply that it is a superconductor.

Still, LEDs are semiconductors pn junctions, aren't they? They aren't just simple semiconductors, and under normal forward bias operation, I'd like to know the mechanism that allows for the presence of electrons and holes in the n-type and p-type semiconductors at 0 K that make up such a junction.

If there are none, then these are possibly tunnel currents, which is a different beast.

Zz.
 
  • #9
Baluncore
Science Advisor
12,339
6,410
Still, LEDs are semiconductors pn junctions, aren't they? They aren't just simple semiconductors, and under normal forward bias operation,
LEDs operate forward biassed, with a PN voltage drop of a few volts.
 
  • #10
ZapperZ
Staff Emeritus
Science Advisor
Education Advisor
Insights Author
35,994
4,716
LEDs operate forward biassed, with a PN voltage drop of a few volts.

Yes, I know. But where are the majority charge carriers when the temperature drops to really, really low values?

Zz.
 
  • #11
eq1
206
71
In a mosfet, if power dissipation was taken out the equation, and the load could take the power, then probably it would hit saturation velocity (it’s no longer ohmic) and eventually the electric field would kill the device.

edit: I misread the question. With a safe Vds it could theoretically be ok in this hypothetical situation and it would just sit in saturation. The higher current density and large charge velocity does increase the probability of hot carrier injection damage accumulating and eventually the mosfet would not longer be a mosfet. It would probably just be always on regardless of driver state.

In reality things are different. Google “FET safe operating area”.

https://en.m.wikipedia.org/wiki/Saturation_velocity

https://en.m.wikipedia.org/wiki/Hot-carrier_injection
 
Last edited:
  • #12
essenmein
657
295
I'd asked a similar question, is there any limitation on current carrying capability of a MOSFET other than thermal. I wasn't going to super cool or anything, just getting paranoid about the A/mm2 I was planning to push. The only answers I got really was more about cyclic reliability than anything else.

Only thing I ever heard about limits of fet, other than breakdown voltage and temperature, is the parasitic NPN getting turned on with too large a dv/dt, although haven't seen a dv/dt limit talked about for a while now.

Since in the real world you cannot ignore thermal effects, even if the current doesn't bring Tj above its limit, resulting thermal strain from high current density in the ribbons (or wires) will cause them to let go prematurely (in a cyclic scenario), this source connection Achilles heel leads to things like copper clip and P2 pack type developments which massively increase the delta Tj you can do and the amount of current you can do before the device fails (again, cyclic conditions eg LV324) .

Speaking of thermal strain, I can't imagine pouring liquid nitrogen on a circuit does it any thermal shock favors!
 

Suggested for: Extremely high current effects in semiconductors

  • Last Post
Replies
1
Views
400
  • Last Post
Replies
2
Views
712
Replies
4
Views
361
  • Last Post
Replies
6
Views
260
  • Last Post
Replies
9
Views
570
Replies
10
Views
657
Replies
12
Views
133
Top