Current peak at minimum transistor length

[jpm]

Hello!

I am current running some simulations on Hpsice and Virtuoso. The simulation consists of nothing more than an on nmos transistor from a 65nm library. The transistor is powered in the threshold regime at 250mV. I am sweeping the length and width and measuring the current through the transistor. My issue is that the results don't fully match what I was expecting to see. I apologise in advance for the crudity of this diagram:



When the width is at the minimum possible (120n), the current rises from the minimum length (60n) due to the Reverse Short Channel Effect and then, after peaking, begins a steady decline due to the increase in resistance of the channel. This is exactly what I suspected.

However, when the width is increased, something I cannot explain begins to occur. At the minimum length, a current peak starts to rise. At around 500n, this surpasses the peak between the RSCE and the linear resistance region. The lower sketch shows the response at 1u.

Can anyone explain why this current peak appears? Is there some physical effect that I am unaware of causing this effect?

 

[beetwee]

What do you mean with: " The transistor is powered in the threshold regime at 250mV".?
The common terms to describe the operation region of a mosfet is either weak inversion (sometimes called sub-threshold) or strong inversion. The region between strong and weak inversion is called moderate inversion. Do you mean that vgs=250mV?

 

[jpm]

What do you mean with: " The transistor is powered in the threshold regime at 250mV".?
The common terms to describe the operation region of a mosfet is either weak inversion (sometimes called sub-threshold) or strong inversion. The region between strong and weak inversion is called moderate inversion. Do you mean that vgs=250mV?

Sorry for this confusion. This is being tested atm as a digital device. Henceforth I mean both Vgs and Vds are 250mV.

 

[beetwee]

Ok, so your nmos is working in weak inversion (subthreshold), as your Vth will probably be around 400-500mV.

I guess that you want to compare your results with the formula Id=W/L I0' *exp[(vgs-vth)/nUT]. Vth is strongly dependent on the length and in case of very small widths also width dependent.

Plot also Vth as function of W and L. you'll see that the Vth is dependent on L. For large L the Vth remains constant, resulting in a linear relation between the length and current. For small W and L, Vth can change a lot and I think that this is the source of the effect you simulated. To know for sure that the non-constant vth it really is the dominant source, you have to check it yourself with the Weak-inversion formula, as i'm not a device physicist.