A New Surface-Potentials Based MOSFET Model : HiSIM

HiSIM stands for Hiroshima-university STARC IGFET Model. It has been developped at Hiroshima University starting in 1992. It has been released as version 1.1.0 in July 2002.

HiSIM and Conventionnal MOSFET Models

HiSIM is interesting because of the way it models channel current. Conventionnal MOSFET models simplify computation of channel current by splitting calculation between a linear dependent region (due to strong inversion) and a saturation region (due to velocity saturation). Discontinuities can appear in IDS, in the transition region. Therefore, to avoid these discontinuities, extra parameters are used to smooth the transition between the different set of equations. These parameters are not physical, they are just needed to correctly fit the device’s characteristics.

Another drawback of common models is the bad modeling of short-channel effects. For deep sub-micron MOSFETs, this effect dominates the IDS-VDS characteristic. Conventional models do not use equations based on physical concepts, but add fitting parameters to each modeled effect to account for short-channel effect. This results in many unphysical fitting parameters, and makes parameter extraction difficult.

The conclusion is that dividing the IDS current into different regions and equations is not correct anymore for short-channel transistors.

HiSIM is based on a charge-sheet model. IDS current is described using only one equation, and there-fore is continuous over the whole range of operating regions. This improves MOSFETs modeling regarding at least two points :

• Equations are continuous over all operation regions, as well as their derivatives. This is a key point for today’s analog circuits, where performance is very much dependent on high order derivatives.

• Parameter number is dramatically reduced (by a factor 5) for the same level of accuracy. Parameters are not interdependent anymore, making extraction easier. Furthermore, a set of parameters is valid for all channel lengths.

 

Surface Potentials

HiSIM is based on charge control and charge flow through the channel. The inversion layer charge and the depletion layer charge depend on the surface potential along the channel.

To compute these charges, the surface potentials at source side _S0 and _SL at drain side are needed. These two values are directly dependent on technological parameters. They are calculated by solving the Poisson equations :

The surface potentials side _S0 and _SL are distributed in the channel according to the schematic shown in Figure 1

Figure 1.

 

Both Poisson equations are solved iteratively, because they are implicit. Using approximations to get explicit equations with regard to terminal voltages would not be an improvement: it would reduce accuracy, and convergence is quickly obtained when solving these two equations.The internal New-ton’s algorithm converges within one to ten iterations, depending on the circuit. This is acceptable for a circuit simulator, since simulation times are comparable to those observed using other models.

The screenshot in Figure 2 shows the surface potentials evolution when VGS increases.

Figure 1.

 

New in Version 1.1.0

The last improvements are: shallow-trench-isolation (STI) is accounted for in leakage current model, a lateral-field-induced capacitance has been added, and the resistance model has been improved, requiring two more model parameters. These new modeling equations make HiSIM even more accurate.

 

Modeled Effects

HiSIM conputes charge control using dedicated parameters to account for the following physical effects:

• Short-channel

• Reverse short-channel

• Pocket implantation

• Quantum

• Poly-depletion

• Universal mobility

• Channel-length modulation

• Velocity overshoot

• Symmetry at VDS=0

• Shallow trench isolation (version 1.1.0 only)

 

HiSIM and SmartSpice

HiSIM is available within SmartSpice when LEVEL=111 is specified. This model has been implemented using reference versions 1.0.0 and 1.1.1. The user can select one of these version using a selector, VERSION. Beyond this material, SmartSpice provides all the services commonly proposed for MOSFET models. Among them are:

• Advanced geometry scaling (ACM)

• Simulation performance using VZERO and BYPASS options

• Friendly diagnostics to help with convergence issues

• Extrinsic elements such as junction diodes/capacitances

The model card for HiSIM includes the following parameters:

 

Technological parameter

TOX	Oxide thickness
XLD	Gate-overlap length
XWD	Gate-overlap width
XPOLYD	Difference between gate-poly and design length
TPOLY	Height of the gate poly-Si
RS	Source contact resistance
RD	Gate contact resistance
NSUBC	Substrate-impurity concentration
NSUBP	Maximum pocket concentration
VFBC	Flat-band voltage
LP	Pocket penetration length
XJ	Junction depth
XQY*	Distance from D junction to maximum electric field

 

Temperature dependence

BGTMP1	Bandgap narrowing
BGTMP2	Bandgap narrowing

 

Quantum effect

QME1	Coefficient 1 for quantum mecanical effect
QME2	Coefficient 2 for quantum mecanical effect
QME3	Coefficient 3 for quantum mecanical effect

 

Poly depletion

PGD1	Strength of poly depletion
PGD2	Threshold voltage of poly depletion
PGD3	VDS dependence of poly depletion

 

Short channel

PARL1	Strength of later-electric-field gradient
PARL2	Depletion width of channel/contact junction
SC1	Short-channel coefficient 1
SC2	Short-channel coefficient 2
SC3	Short-channel coefficient 3
SCP1	Short-channel coefficient 1 for pocket
SCP2	Short-channel coefficient 2 for pocket
SCP3	Short-channel coefficient 3 for pocket

 

Narrow channel

WFC	Threshold voltage reduction
MUEPH2	Mobility reduction
W0	Minimum gate width
WVTHSC*	SHort-channel effect at the STI edge
NSTI*	Substrate impurity concentration at the STI edge
WSTI*	Width of the high-field region at STI

 

Mobility

VDS0	Drain voltage for extracting the low-field mobility
MUECB0	Coulomb scattering
MUECB1	Coulomb scattering
MUEPH0	Phonon scattering
MUEPH1	Phonon scattering
MUETMP	Temperature dependence of phonon scattering
MUESR0	Surface-roughness scattering
MUESR1	Surface-roughness scattering
NDEP	Coefficient 1 of effective-electric
NINV	Coefficient 2 of effective-electric
NINVD	Modification of NINV
BB	High-field-mobility degradation
VMAX	Maximum saturation velocity
VOVER	Velocity overshoot effect
VOVERP	LGATE dependence of velocity overshoot
RPOCK1	Resistance coefficient 1 caused by the potential barrier
RPOCK2	Resistance coefficient 2 caused by the potential barrier
RPOCP1*	Resistance coefficient 3 caused by the potential barrier
RPOCP2*	Resistance coefficient 4 caused by the potential barrier

Channel-length modulation

CLM1	Hardness coefficient of channel/contact junction
CLM2	Coefficient for QB contribution
CLM3	Coefficient for QI contribution

 

Substrate current

SUB1	Substrate current coefficient 1
SUB2	Substrate current coefficient 2
SUB3	Substrate current coefficient 3

 

Gate current

GLEAK1	Gate current coefficient 1
GLEAK2	Gate current coefficient 2
GLEAK3	Gate current coefficient 3

 

GIDL current

GIDL1	GIDL current coefficient 1
GIDL2	GIDL current coefficient 2
GIDL3	GIDL current coefficient 3

 

Flicker noise

NFALP	Contribution of the mobility fluctuation
NFTRP	Ratio of trap density to attenuation coefficient
CIT	Capacitance caused by the interface trapped carriers

 

Symmetry at VDS=0

VZADD0	Symmetry conservation coefficient 1
PZADD0	Symmetry conservation coefficient 2

* available only in version 1.1.0