Complete Circuit Simulation Solution for SOI Technology: Four New Models

Manish Pagey, Vanderbilt University

Introduction

Several models for simulation of SOI MOSFETs have been proposed in the recent years. Some of the most significant models include the physics based models (fully depleted and non-fully depleted) from the University of Florida, BSIM3v3 bulk MOSFET model from the University of California, Berkeley, and SOI model from Honeywell. All of these models have been recently implemented into SmartSpice and UTMOST. The table below lists the different models and the level numbers in SmartSpice which are used to invoke these models.


Table 1. Model and level description of four new SOI models.

 

HSOI: Honeywell SOI Model

The SOI model proposed by Dr. James Lai at Honeywell has been used extensively by Honeywell and their customers for the past several years. The current version of this model is based on the MOS level 3 model and provides fast execution for large circuit simulation. It has parameters which allow simulation of effects related to partially depleted SOI MOSFETs such as the kink effect.


Figure 1. Typical IDS versus VGS simulation curves using Honeywell SOI model.


Figure 2. Typical IDS versus VDS simulation curves using Honeywell SOI model.


Figure 3. Ring oscillator simulation using Honeywell SOI model.

Default Model

.model defaulthsoi nmos (level=20 ws=0.0 wref=0.0
lref=0 vt0=0 phi=0.576 gamma=0 delta=0
+eta=0 theta=0 alpha=0 kappa=0.2 eg=1.16 gap1=7.02e
+gap2=1108 cta=0 ctp=0 pta=0 php=0.8 ptp=0 ptc=0
+tcv=0 wmlt=1 alphaz=0 wvcr=0 alpha=0 walpha=0
+vcr=0 lvcr=0 lmlt=1 nlev=0 rsc=0 rdc=0 bex=-1.5
+jws=0 trd=0 trs=0)


University of Florida FD/NFD SOI Models

The device modeling group at the University of Florida, Gainesvilles has developed physical charge-based models for both fully depleted (FD) and non fully depleted (NFD) SOI MOSFETs. These models include several SOI related phenomena such as parasitic BJTs, non-local impact ionization current, temperature dependence and noise modeling. Several other optional features are present to include short-channel effects. The current model implemented in SmartSpice corresponds to models implemented in SOISPICE-4.3 by University of Florida. Physics based models such as these two models are essential to understand SOI behavior due to phenomena such as self-heating, floating body effects, hysteresis effects and gate-substrate coupling. The current versions of the fully depleted and non fully depleted SOI models cover several of these phenomena. Specifically, the fully depleted SOI model includes:

Figure 4. Typical IDS versus VGS simulation curves using University of Florida fully depleted SOI model.
  1. Front-gate/back-gate charge coupling,
  2. Source/drain field fringing in the buried oxide,
  3. Coupled parasitic/bipolar current/charge modeling,
  4. Charge/capacitance modeling continuous across all boundaries,
  5. LDD/LDS options,
  6. Body tie options,
  7. Dynamic body charging in off/near off state (model becomes NFD),
  8. Temperature modeling with self-heating option,
  9. Physical or structural parameters which allow easier evaluation.


Figure 5. Typical IDS versus VDS simulation curves using University of Florida fully depleted SOI model.

Similarly, the non-fully depleted SOI model includes the following capabilities:

  1. Inhomogeneous channel doping,
  2. DC and transient floating body effects,
  3. Coupled parasitic bipolar current/charge modeling,
  4. Charge/capacitance modeling continuous across all boundaries,
  5. LDD/LDS options,
  6. Body tie options,
  7. Dynamic body charging in off/near off state (model becomes NFD),
  8. Temperature modeling with self-heating option,
  9. Physical or structural parameters which allow easier evaluation.

Both the FD and NFD SOI models implement temperature dependent physical model parameters. All of these temperature dependencies have been taken from published literature.

Model Equations:

  1. Parasitic BJT Model: The recombination current model, that dictates the significant part of the BJT current, accounts for the effects of an LDS region and also includes doping dependent carrier mobility and lifetime models. The detailed analysis and equations can be found in [1].
  2. Doping Dependent Mobility Model: An empirical mobility model dependent on doping density has been used. This model is given by:

    where N is the doping density. N_ref and alpha are empirical parameters and mu_max and mu_min are maximum and minimum mobility values. Typically, for n-type silicon N_ref=1.3e17/cm3, alpha=0.91, mu_max=1360cm2/V-sec, and mu_min=92cm2/V-sec. The corresponding values for p-type silicon are 6.3e16/cm3, alpha=0.76, mu_max=495cm2/V-sec, and 47.7cm2/V-sec respectively.
  1. Doping Dependent Lifetime Model: The carrier lifetime is a combination of terms due to SRH recombination and band-to-band Auger process:

    The models used for SRH and Auger terms are:
  2. Impact Ionization Current: The weak impact ionization current has been modeled as:

    The multiplication factor M is obtained by integrating the ionization coefficient over the high-field region close to the drain[2].


Figure 6. Typical IDS versus VGS simulation curves using University of Florida non-fully depleted SOI model.


Figure 7. Typical IDS versus VDS simulation curves using University of Florida non-fully depleted SOI model.

Default Model Parameters for FDSOI Model:

.model defaultFDSOI NMOS ( level=21
+nqff=0 nqfb=0 toxf=1.0e-8 toxb=0.5e-6 nsub=1.0e15
+ngate=1.0e19 nsf=0 nsb=0 tpg=1 tps=-1 tb=0.1e-6
+nbody=5.0e16 uo=700 theta=1.0e-6 bfact=0.3 vsat=1.0e7
+zeta=1 alpha=0 beta=0 gamma=0.3 tau0=1.0e-6
+llds=0 lldd=0 nlds=3.0e18 nds=5.0e19 jro=1.0e-15
+eta=1 lmod=1 cgfdo=0 cgfso=0 cgfbo=0
+rhosd=0 rhob=0 rd=0 rs=0 rb=0 dl=0 dw=0
+kappa=0.5 ldiff=1.0e-7 vjmin=0.7 fvbjt=0
+fnk=0 fna=1.0 )

Default Model Parameters for NFDSOI Model:

.model defaultNFDSOI NMOS ( level=22
+toxf=1.0e-8 toxb=0.5e-6 nqff=0 nqfb=0
+nsub=1.0e15 ngate=1.0e19 tf=0.2e-6 tb=0.1e-6
+thalo=0 nbl=5.0e16 nbh=5.0e17 nhalo=0
+uo=700 theta=1.0e-6 vsat=1.0e7 tpg=1 tps=-1
+nlds=5.0e19 nds=5.0e19 llds=0 lldd=0 zeta=1
+eta=1 lmod=1 alpha=0 beta=0 bfact=0.3
+cgfdo=0 cgfso=0 cgfbo=0 rd=0 rs=0 rb=0
+tau0=1.0e-6 dl=0 dw=0 rhosd=0 rhob=0
+jr0=1.0e-10 m=2 ldiff=1.0e-7 vjmin=0.7
+fvbjt=0 +fnk=0 fna=1.0 )

BSIM3SOI Model

University of California at Berkeley has released an SOI model based upon the latest bulk MOSFET model BSIM3v3. The basic MOSFET equations present in the bulk MOSFET model have been modified and augmented to included SOI specific phenomena. The current version includes the following features:

  1. The effects of localized self-heating
  2. The effects of floating body charging
  3. fully-depleted and partially depleted SOI model.
  4. computtionally efficient and physical


Figure 8. Typical IDS versus VGS simulation curves
using University of California, Berkeley SOI model.


Figure 9. Typical IDS versus VDS simulation curves
using University of California, Berkeley SOI model.

BSIM3SOI Model Improvements

  1. Self Heating: The self-heating effects are modeled by a thermal resistor assuming that the device temperature is linearly proportional to the dissipated power. Due to poor heat conductance of the material, the temperature in the device does not change instantaneously. The temperature's finite rise time is modeled as a transient behavior with a single time constant dependent on geometry of the device. The time constant is represented with the equivalent voltage Vtemp that is inserted in the BSIM3 temperature model.
  2. Floating Body Model: The absence of the body creates the kink effect present in partially depleted (PD) devices. The kink is a result of several effects:
    1. Leakage Diode: The behavior of the floating body is greatly influenced by PN junction diodes at both the source/body and drain/body junctions. The diode model has been enhanced around 0 bias where most of the time the diodes will operate. Model has been further improved to account for high reverse bias tunneling currents.
    2. Implant Ionization Current: The SOI model is improved over the bulk model since at thinner oxides the electric field distribution is different in SOI devices compared to bulk devices. Several model parameters are added to improve modeling of the kink effect and impact ionization transients.
    3. GIDL Current: This effect is modeled using the Fowler-Nordheim type equations since at very large bias conditions the bends near the drain/body junction are bent significantly resulting in thinner barriers so the charge begins to tunnel through.
    4. Parasitic Bipolar Effects: This is modeled by placing extra current sources in the junction leakage diode to give the Ebes-Moll type model.

Default Model Parameters

.model defaultbsim3soi NMOS (level=23
+ rth=0 cth=0 tnom=27 ute=-1.5 ua1=0 ub1=0
+ uc1=0 at-3.3e4 kt1=0 kt2=0 alpha0=0
+ beta0=0 linep=0 seta0=1000 sgamma=0 bbjt=0
+ bgidl=1 ngide=1 agide=0 is1=1e-14 is2=0
+ is3=0 n=1 is3n=1 tsi=1000 ktsi=0 kflag=0
+ rsi=0 vth00.7 k1=0.5 k2=0 rsh=0)

References:

[1] Jin Young Choi, "Modeling and Simulation of the Fully Depleted Silicon-On-Insulator MOSFET for Submicron CMOS IC Design," Ph. D. Diss., University of Florida, 1991.

[2] S. Veeraraghavan, and J. G. Fossum, "A Physical Short-Channel Model for the Thin-Film SOI MOSFET Applicable to Device and Circuit CAD," IEEE Tran. Electron Devices, vol. 35, pp. 1866-1875, Nov. 1988.