Threshold voltage shift vs temperature

[ZeroFunGame]

TL;DR Summary

In a field effect transistor, does the threshold voltage decrease with increasing temperature? Imagine that there would be an increase in thermally generated carriers and thus the device would reach saturation faster? I realize the mobility would decrease, but the increase of thermal carriers outpaces phonon scattering effects such that conductivity rises. Is this the correct way to think about threshold voltage as a function of temperature?

In a field effect transistor, does the threshold voltage decrease with increasing temperature? Imagine that there would be an increase in thermally generated carriers and thus the device would reach saturation faster? I realize the mobility would decrease, but the increase of thermal carriers outpaces phonon scattering effects such that conductivity rises. Is this the correct way to think about threshold voltage as a function of temperature?

 

[trurle]

-2 mV/K for silicon

 

[ZeroFunGame]

Is it because the conductivity increases? If the threshold voltage does decrease, wouldn't this make a ring ossilator have an increasing fmax as T increased?

 

[trurle]

Is it because the conductivity increases? If the threshold voltage does decrease, wouldn't this make a ring ossilator have an increasing fmax as T increased?

No. Conductivity decrease with increased temperature is a separate effect from narrowing of bandgap which results in threshold voltage shift. And regarding current, mobility degradation typically dominates over threshold voltage shift unless transistor biased in very narrow range of gate voltages near threshold voltage. Therefore, the ring oscillator frequency falls with temperature.

 

[ZeroFunGame]

No. Conductivity decrease with increased temperature is a separate effect from narrowing of bandgap which results in threshold voltage shift. And regarding current, mobility degradation typically dominates over threshold voltage shift unless transistor biased in very narrow range of gate voltages near threshold voltage. Therefore, the ring oscillator frequency falls with temperature.

As I understand it, conductivity increases with temp in a semiconductor rather than decrease.
Fig 2.2.39 here: https://archive.cnx.org/contents/64...y-in-semiconductor-metal-and-its-conductivity

So the reason for the decrease in threshold voltage is due to a decreasing bandgap.

So you are saying mobility degradation due to phonon scattering is the main reason for reduced oscillator frequency with temperature? Even if conductivity increases?

 

[trurle]

As I understand it, conductivity increases with temp in a semiconductor rather than decrease.
Fig 2.2.39 here: https://archive.cnx.org/contents/64...y-in-semiconductor-metal-and-its-conductivity

You are mixing here the behavior of intrinsic conduction in high-purity semiconductor (actually intrinsic conduction was experimentally observed for germanium, but required materials purity is too good for modern silicon technology) and behavior of actual electronic devices which nearly never use intrinsic conduction, instead relying on fully ionized doping to provide charge carriers. Devices based on doped semiconductor have approximately fixed number of carriers and carriers mobility roughly proportional to T^(-2.5).

So you are saying mobility degradation due to phonon scattering is the main reason for reduced oscillator frequency with temperature? Even if conductivity increases?

It is a bit meaningless statement. Small-signal conductivity have a direct relation to carriers mobility.

 

[ZeroFunGame]

So as T increases, the thermally generated carriers do not help with conduction in a doped semiconductor device? But only in the intrinsic semiconductor case?

 

[ZeroFunGame]

Why are thermally generated carriers negligible in a doped semiconductor? Is it because the dopant density >> thermally generated intrinsic carrier density? Is this still true for wide bandgap semiconductors that have shown IC operation in >500C (for example SiC)? That is, the thermally generated carriers do not help much in a 500C environment?

 

[trurle]

Is it because the dopant density >> thermally generated intrinsic carrier density? Is this still true for wide bandgap semiconductors that have shown IC operation in >500C (for example SiC)?

True. Especially for wide bandgap semiconductors. Wider bandgap, less thermal carriers. The limiting factor for SiC at high temperature is actually failure due migration of dopants, not because of proliferation of thermal charge carriers.

Overall, i do not see where "thermally generated carriers" may be helpful. They may increase on current, but also increase off current by same amount. Pure waste of power except for may be non-existing exotics like "reversible anti-fuse".

 

[ZeroFunGame]

True. Especially for wide bandgap semiconductors. Wider bandgap, less thermal carriers. The limiting factor for SiC at high temperature is actually failure due migration of dopants, not because of proliferation of thermal charge carriers.

Overall, i do not see where "thermally generated carriers" may be helpful. They may increase on current, but also increase off current by same amount. Pure waste of power except for may be non-existing exotics like "reversible anti-fuse".

Figure 5 in the link below shows the correlation of conductivity on T for *extrinsic* semiconductor:
https://nptel.ac.in/content/storage2/courses/113106062/Lec8.pdf

This is what I had in mind, where as T increases the conductivity also increases. However, you mentioned that due to dopants >> intrinsic carriers, the conductivity is not dominated by the number of carriers, but by mobility. If mobility continues to degrade due to phonon scattering with increasing T, why is it that conductivity increase at high T for an extrinsic semiconductor?

 

[trurle]

Figure 5 in the link below shows the correlation of conductivity on T for *extrinsic* semiconductor:
https://nptel.ac.in/content/storage2/courses/113106062/Lec8.pdf

This is what I had in mind, where as T increases the conductivity also increases. However, you mentioned that due to dopants >> intrinsic carriers, the conductivity is not dominated by the number of carriers, but by mobility. If mobility continues to degrade due to phonon scattering with increasing T, why is it that conductivity increase at high T for an extrinsic semiconductor?

Figure 2,3, and 5 in your reference have conductivity increase at high temperature due incomplete ionization of (arbitrary) dopant. Practical devices select dopant of "shallow" type, which ionize completely below 200K. Such dopant selection (boron, phosphorus, arsenic in silicon) allows better stability of device parameters over extended temperature range.

 

[ZeroFunGame]

Figure 2,3, and 5 in your reference have conductivity increase at high temperature due incomplete ionization of (arbitrary) dopant. Practical devices select dopant of "shallow" type, which ionize completely below 200K. Such dopant selection (boron, phosphorus, arsenic in silicon) allows better stability of device parameters over extended temperature range.

I see, so the plot represent deep donor levels. Basically, you are saying that at high enough T, if Figure 5 continued to extend to the left, then the mobility would peak (representing the complete ionization of all donors) and eventually begin to reduce due to phonon scattering?

 

[trurle]

I see, so the plot represent deep donor levels. Basically, you are saying that at high enough T, if Figure 5 continued to extend to the left, then the mobility would peak (representing the complete ionization of all donors) and eventually begin to reduce due to phonon scattering?

The correct statement would be
"at high enough T, if Figure 5 continued to extend to the left, then the conductivity would peak (representing the complete ionization of all donors) and eventually begin to reduce due to phonon scattering".
conductivity~number_of_carriers*mobility_of_carriers

 

[ZeroFunGame]

The correct statement would be
"at high enough T, if Figure 5 continued to extend to the left, then the conductivity would peak (representing the complete ionization of all donors) and eventually begin to reduce due to phonon scattering".
conductivity~number_of_carriers*mobility_of_carriers

In reference to Fig 5 here: https://www.semanticscholar.org/pap...-Hou/058f3e4087df56522a8b98d84a67f2effa598cda

It looks to me like f_osc increase with T from RT to 300C - 400C (depending on supply voltage), before decreasing. Similarly, the inverter delay is reduced at higher T up until 400C.

This is counter intuitive to what I would expect from transistor performance from RT to 400C, since wouldn't increase in lattice scattering add to the delay and thus reduce the RO frequency as T increased?

 

[trurle]

In reference to Fig 5 here: https://www.semanticscholar.org/paper/A-600-°C-TTL-Based-11-Stage-Ring-Oscillator-in-Shakir-Hou/058f3e4087df56522a8b98d84a67f2effa598cda

It looks to me like f_osc increase with T from RT to 300C - 400C (depending on supply voltage), before decreasing. Similarly, the inverter delay is reduced at higher T up until 400C.

This is counter intuitive to what I would expect from transistor performance from RT to 400C, since wouldn't increase in lattice scattering add to the delay and thus reduce the RO frequency as T increased?

From abstract, seems the device is highly anomalous. Unfortunately, cannot say why because the full text is behind paywall.

 

[ZeroFunGame]

From abstract, seems the device is highly anomalous. Unfortunately, cannot say why because the full text is behind paywall.

TTL implies BJT, so I was wondering if they behaved differently than majority carrier devices like JFETs or MOSFETs. My initial thought was that at higher T, the natural depletion region is reduced, so this could imply faster switching operation, but was just a guess and not entirely sure if the logic follows.

 

[trurle]

TTL implies BJT, so I was wondering if they behaved differently than majority carrier devices like JFETs or MOSFETs. My initial thought was that at higher T, the natural depletion region is reduced, so this could imply faster switching operation, but was just a guess and not entirely sure if the logic follows.

Base width modulation in other words may be. Without full-text it remain a unproven hypothesis though. I requested a full text at researchgate.net, let`s wait for the author response.

P.S. From figures seems the anomaly is due low-doped resistors at collector which seems to be going into intrinsic conduction mode at high temperature. The logic high is 5V below supply rail at room temperature - likely indicating the collector resistors value were higher than optimal for best speed.