HiSIM version 1.0 Model Released in UTMOST III

 

Introduction

HiSIM is a MOSFET model for SPICE circuit simulation that has been developed by Hiroshima University and STARC Company.

This model present several advantage on the extraction point of view, with a reasonable number of parameters, a physical reliability of the equations for a wide range of geometries (down to 0.1um) and a unified description of devices characteristics for all bias conditions.

It has been implemented in Silvaco Spice Simulator, SmartSpice and in our current extraction software UTMOST III.

This article presents an efficient extraction procedure for HiSIMv1.0 parameters using UTMOST III. This procedure will lead to a model card accurate for a wide range of geometries. It is widely inspired from the one proposed by STARC in its “HiSIM1.0 User’s Manual”, but it has been adapted and optimized for UTMOST III users.

 

Data Collection and Initial Parameter

The extraction procedure presented in this article requires DC measurement that can be obtained using the BSIM3_MG routine on a Large and Wide device and an L-array of devices (See articles for BSIM3_MG extraction routine in the previous issues of the Simulation Standard.). In this case “BSIM3_MG” should be used as a measurement routine only.

Following STARC recommendations, IDS vs. VGS measurements should be performed for VDS=50mV and 1V. IDS vs. VDS measurements will be performed for VBS=0V and -1V. For those last characteristics, the Voffset value which is used to calculate the first VGstart value (Vgstart = VTextracted + Voffset) defaults to 0.2V.

The recommended number of points per sweep in the BSIM3_MG routine is 51 and the number of VGS steps and VBS steps are 5.

In addition, to be able to extract Quantum Mechanical Effect parameters and Poly Depletion effect parameters, Intrinsic Gate capacitance vs. Vgs should be measured on Large and Wide device.

The typical number of geometries used for model parameter extraction is 10. There should be a large device with wide W and long L (to avoid short channel or narrow width effects) to extract the root parameters (as mobility parameters, substrate impurity concentration and flat band volt-age). An array of short L devices will be used to extract short channel parameters. It has to be noticed that HiSIM includes two kind of short channel effect parameters: One for the standard short channel effect, and the other for the reverse short channel effect. We will see later in this section how this two effects lead to separate the L-array of devices in two parts depending on which parameters need to be optimized.

Restrictions This article gives a procedure to extract most of HiSIMv1.0 parameters. However, some effects are not taken into account, their corresponding parameters extraction will require another article in a future Simulation Standard. Those parameters are: Narrow channel, Channel-length modulation, Substrate Gate and GIDL current, 1/f Noise parameters.

Model Parameters
To get significant results for an HiSIMv1.0 extraction, the initial set of values for HiSIM parameters is important. UTMOST III provides you a default model card that should be used as a starting point. Detail of this model card is shown in Table 1.

Parameter Description Units Default VERSION HiSIM Model version number (SmartSpice specific)   1.1 Technological parameters TOX oxide thickness m 5E-9 XLD gate-overlap length m 0 XWD gate-overlap width m 0 XPOLYD difference between gate-poly and design lengths m 0 TPOLY height of the gate poly-Si m 0 RS source-contact resistance Vm/A 80E-6 RD Drain contact resistance Vm/A 80E-6 NSUBC substrate impurity concentration cm-3 1E17 NSUBP maximum pocket concentration cm-3 1E17 VFBC flat-band voltage V -1 LP Pocket penetration length m 15E-9 XJ Junction depth m 0 Temperature dependence parameters BGTMP1 bandgap narrowing eV/K 90.25E-6 BGTMP2 bandgap narrowing eV/K^2 100E-9 Quantum effect parameters QME1 Coefficient for quantum mechanical effect Vm 10E-3 QME2 Coefficient for quantum mechanical effect V 300E-12 QME3 Coefficient for quantum mechanical effect m 0 Poly depletion effect parameters PGD1 strength of poly depletion V 10E-3 PGD2 threshold voltage of poly depletion V 1 PGD3 Vds dependence of poly depletion   0.8 Short Channel effect parameters PARL1 strength of lateral electric field gradient   1 PARL2 depletion width of channel/contact junction m 0 SC1 short channel 1 V-1 0 SC2 short channel coefficient 2 V-2 0 SC3 short channel coefficient 3 V-2m 0 SCP1 Short-channel coefficient 1 for pocket V-1 0 SCP2 Short-channel coefficient 2 for pocket V-2 0 SCP3 Short-channel coefficient 3 for pocket V -2m 0 Narrow channel WFC Threshold voltage reduction F F cm-2 m 0 MUEPH2 mobility reduction   0 W0 minimum gate width log(cm) 0 Mobility VDS0 drain voltage for extractions the low field mobility V 50E-3 MUECB0 Coulomb scattering cm2 V-1 s-1 300 MUECB1 Coulomb scattering cm2 V-1 s-1 30 MUEPH0 phonon scattering cm2 (Vs)-1 (Vcm-1)MUESR1 300E-3 MUEPH1 phonon scattering   25E3 MUETMP temperature dependence of phonon scattering   1.5 MUESR0 surface roughness scattering cm2 (Vs)-1 (Vcm-1)MUESR1 2 MUESR1 surface roughness scattering   2E15 NDEP coefficient of effective electric field 1 NINV coefficient of effective electric field   0.5 NINVD modification of NINV V-1 1E-9 BB high field mobility degradation   2 VMAX maximum saturation velocity cm/s 4E6 VOVER velocity overshoot effect cmVOVERP 10E-3 VOVERP Lgate dependence of velocity overshoot   100-3 RPOCK1 resistance coefficient caused by the potential barrier V2 A-RPOC1 umRPOC2-1 10E-3 RPOCK2 resistance coefficient caused by the potential barrier V 100-3 Channel length modulation CLM1 hardness coefficient of channel / contact junction   700E-3 CLM2 coefficient for Qb contribution   2 CLM3 substrate current coefficient   1 Substrate current SUB1 substrate current coefficient V-1 10 SUB2 substrate current coefficient V 20 SUB3 substrate current coefficient   0.8 Gate current GLEAK1 gate current coefficient 1 AV-3/2 C-1 10E3 GLEAK2 gate current coefficient 2 Vcm-1 V-1.5 20E6 GLEAK3 gate current coefficient 3    300E-3 GIDL current GIDL1 GIDL current coefficient 1 AV-3/2 C-1 5E-6 GIDL2 GIDL current coefficient 2 Vcm-1 V-1.5 1E6 GIDL3  GIDL current coefficient 3     300E-3 1/f Noise NFALP contribution of the mobility fluctuation Vs 1E-16 NFTRP ratio trap density to attenuation coefficient V-1 cm-2 10E6 CIT capacitance caused by the interface trapped carriers  Fcm-2 0 Conservation of the symmetry at Vds=0 for Short-Channel MOSFETs VZADD0 symmetry conservation coefficient V 10E-3 PZADD0 symmetry conservation coefficient V 5E-3

Table 1: HiSIMv1.1 default model card.

 

Geometry Selection
As mentioned above, we have three “group” of device geometry that will be used separately depending on the kind of parameters that will be extracted.

The Long and Wide device used for mobility, flat band voltage and substrate impurity concentration parameters extraction.

The split of the L-array should be done according to the Vth dependency on the channel length, which can be observed through UTMOST validate routine. The L-array of devices will be split in two parts depending on what kind of short channel parameters will be optimized.

Standard short channel effect makes VTH grows with L gate, while Reverse short channel effect has the opposite influence.

The bigger devices of the L-array will be used for reverse short channel effect parameters, while the smaller devices, will be used for standard short channel effect.

Figure 1. VTH dependence on L - UTMOST routine “Validate“.

 

On the example shown in Figure 2, the middle group includes devices from 10x5 to 10x0.2 while short group includes devices from 10x0.13 to 10x0.11. This split of the L-array is valid for low VDS.

Figure 2. VTH dependence on L for a High
VDS - UTMOST routine “Validate”.

 

VTH dependence on L for high VDS (VDS=1V) will show a middle group including devices from 10x5 to 10x0.5 while short group will include device from 10x0.15 to 10x0.11.

 

Local Optimization Strategies

After the data is collected, The ALL_DC routine can be used for local optimization. In this article’s example a single ALL_DC routine will be used. The different types of data will be displayed in the ALL_DC graphics screen for different optimization strategies. This may require more user interface but it is easier to follow each step of local optimizations this way. Later the user may automate the local optimization strategies by utilizing the different ALL_DC routines. The operation of the local optimization is explained in the UTMOST User Manual.

Strategy #30: idvg_large_HiSIM
This strategy is used for the wide W and long L device only. As it can be seen in the figure 3, it will optimize the “Current” of “ID/VG” characteristics. The Wide Wand long L device should be selected in the “Geometry Selection Screen” (figure 5) for each row in this strategy. The ID/VG characteristics of this device (wide W and long L) should be present in the graphics screen The first row of the optimization is used for the substrate impurity concentration and the flat band voltage parameters optimization. The “sweep/start” is set to 1and “sweep/stop” to 5. This will ensure that these parameters in row#1 will be optimized for VBS values. This Optimization is done only for low VDS. We have fixed the current from 1E-12A to 10% of maximum current because the region of interest is the subthreshold region.

Figures 3 and 4 show that the second row is exactly the same than the first one. It is used to iterate this optimization.

Figure 3. Local optimization strategy definition
screen for Strategy#30 (idvg_large_HiSIM)

 

Figure 4. Local optimization target selection
screen for Strategy#30 (idvg_large_HiSIM)

 

Figure 5. Local optimization geometry
selection screen for
Strategy#30 (idvg_large_HiSIM)

 

 

Strategy #31: idvg_middle_HiSIM
This strategy is used to optimized pocket penetration length and maximum pocket concentration (LP and NSUBP) in a first step, then reverse short-channel coefficient 1 and 3 for pocket (SCP1 and SCP3) in a second one. It will optimize the “Current” of “ID/VG” characteristics (Figure 6). The middle devices of the L-array should be selected for each row in this strategy. Pocket parameters are going to be optimized for all VBS value (“sweep/start” set to 1 and “sweep/ stop” to 5) while short channel parameters are optimized only for VBS=0V, as there is no VBS effect on standard short channel effect. All this strategy is done for low VDS. Those parameters are optimized in subthreshold and around threshold region. As for Strategy 30, step 1 and 2 are duplicated in order to iterate the optimization.

Figure 6. Local optimization strategy definition
screen for Strategy#31 (idvg_middle_HiSIM)

 

Figure 7. Local optimization target selection
screen for Strategy#31 (idvg_middle_HiSIM)

 

Strategy #32: idvg_short_HiSIM
As the previous ones, the strategy#32 will optimize the “Current” of “ID/VG” characteristics. The aim of this strategy is to optimize the short channel parameters PARL2, SC1 and SC3. As standard short channel effect is much more sensible on very small devices, the short devices part of the L-array will be used for optimization. Region of interest is subthreshold region for all VBS and low VDS. As with previous strategy, several iteration of the same step are required.

Figure 8. Local optimization strategy definition
screen for Strategy#32(idvg_short_HiSIM)

 

Figure 9. Local optimization target selection
screen for Strategy#32 (idvg_short_HiSIM)

 

Strategy #33: idvg_highVT_HiSIM
Strategy #33 aims to take into account VDS influence on the different effects we have studied previously (Flat band voltage, standard short channel and reverse short channel effects). This strategy, as the previous ones, will perform optimization on the “Current” of “ID/VG” characteristics. Region of interest is subthreshold for all VBS values, but this time optimizations are going to be performed at low and high VDS.

First step optimized flat band voltage (VFBC) on the Large and Wide device for low and high VDS.

Second step is used to optimized reverse short channel effect dependency on VDS (SCP2). This is done on middle devices of the L-array for low and high VDS. As mentioned above (in geometry selection section), the middle group is not the same for low VDS and high VDS.

Third step is used to optimized standard short channel effect dependency on VDS (SC2). Optimizations are performed on short devices for low and high VDS.

Figure 10. Local optimization strategy definition
screen for Strategy#33 (idvg_highVT_HiSIM).

 

Figure 11. Local optimization target selection
screen for Strategy#33 (idvg_highVT_HiSIM).

 

Strategy #34: idvg_highVT2_HiSIM
As preceding strategy, the aim of strategy #34 is to optimize short channel effect dependency to VDS. While strategy #33 was a rough extraction of SC2 and SCP2, this one has a refinement purpose. This is obtained by optimizing the “Current” of “ID/VG” characteristics in subthreshold region for high VDS only. The first step optimized SCP2 on middle devices while the second step optimized SC2 on short devices. As for previous strategies, those steps are iterated. Check that short and middle devices groups are the ones defined for high VDS.

Figure 12. Local optimization strategy definition
screen for Strategy#34 (idvg_highVT2_HiSIM).

 

 

Strategy #35: idvg_lowMue_HiSIM
This one is dedicated to mobility parameters optimization. Optimizations are going to be performed on the “Current” of “ID/VG” on the Large and Wide device. Many different iteration are necessary due to the strong correlation of those parameters effects in the different area of the characteristics.

First step will be used for a rough extraction of MUECB0, MUECB1, MUEPH1 and MUESR1 in subthreshold region for a low VDS and 0V VBS (sweep_start = sweep_stop = 1).

Second step will refine MUECB1 in subthreshold region for low VDS and all VBS.
Third step is used for MUEPH1 optimization; this will be done just after threshold region for high VDS and all VBS.

Fourth step is used for MUESR1 optimization in saturation region for high VDS and all VBS.

Fifth step refine MUECB0, MUECB1 in subthreshold region for low VDS and all VBS.

Sixth step refines MUEPH1 in just above threshold region for high VDS and all VBS.

Last step refines MUEPH1 and MUESR1 together in saturation region for high VDS and all VBS.

Figure 14. Local optimization strategy screen
for Strategy#35 (idvg_lowMue_HiSIM).

 

Figure 15. Local optimization target selection screen
for Strategy#35 (idvg_lowMue_HiSIM).

 

Strategy #36: idvg_highVD_HiSIM
The strategy#36 is used for the high VDS, ID/VG-VB characteristics of all the L-array devices. The aim is to optimize velocity parameters on saturation region of the curve for all VBS. Optimized parameters are VMAX, VOVER and VOVERP.

Figure 16. Local optimization strategy definition
screen for Strategy#36 (idvg_highVD_HiSIM).

 

Figure 17. Local optimization target selection
screen for Strategy#36 (idvg_highVD_HiSIM).

 

 

Strategy #37: idvd_saturate HiSIM
Strategy #37 is applied to “Current ID/VD-VG” characteristics for the smaller devices of the L-array for both low and high VBS. The aim is to optimize gate overlap length (XLD) in addition to velocity parameters (VMAX, VOVER, VOVERP). The optimization is performed in saturation region.

Figure 18. Local optimization strategy definition
screen for Strategy#37(idvd_saturate_HiSIM).

 

Figure 19. Local optimization target selection
screen for Strategy#37 (idvd_saturate_HiSIM).

 

 

Strategy #38: idvd_linear_HiSIM
Strategy #38 is used to optimize contact resistance and potential barrier resistance parameters (RS, RD and RPOCK1, RPOCK2) on ID/VD-VG curve at low VBS. A single step is used for those optimizations. They are performed in linear region of the characteristics (VDS start=0, VDSstop=0.5, sweep start=3, sweep stop=5) for Large and Wide device and middle devices.

Figure 20. Local optimization strategy definition
screen for Strategy#38 (idvd_linear_HiSIM).

 

Figure 21. Local optimization strategy target selection
screen for Strategy#38 (idvd_linear_HiSIM).

 

Local Optimization Sequence
The following sequence is presented as an advice to obtain a good model extraction procedure. It is NOT recommended to run all the sequence at once. The user should run each strategy one by one and observe the optimization results after each strategy is completed. By running sequentially the sequence, the user will be able to repeat some strategies or to come back in the procedure to obtain a better fit. The following sequence should be seen as a hint to achieve a good extraction:

- Rough optimization of technological then mobility parameters:

idvg_large_HiSIM
idvg_lowMue_HiSIM
idvg_large_HiSIM

- extraction of short channel effect parameters:

idvg_middle_HiSIM
idvg_large_HiSIM
idvg_middle_HiSIM
idvg_short_HiSIM
idvg_middle_HiSIM
idvg_short_HiSIM (x2)

- extraction of VDS dependence of short channel effect:

idvg_highVT_HiSIM
idvg_highVT2_HiSIM(x4)

- extraction of mobility and velocity parameters:

idvg_lowMue_HiSIM
idvg_highVD_HiSIM
idvd_saturate_HiSIM

- extraction of resistance parameters:

idvd_linear_HiSIM

 

Poly Depletion and Quantum-Mechanical Effect Direct Extraction

UTMOST III now includes a routine dedicated to Poly depletion and Quantum-mechanical effect direct extraction. This new algorithm can be accessed from INTCAP - CGG routine. INTCAP is an UTMOST III routine dedicated to MOSFETs Intrinsic capacitance measurement, extraction and optimization. Description of how to use this routine can be found in “UTMOSTIII - Extraction manual Vol 1”.

This algorithm should be applied to a Large and Wide device.

To access QMExx and PGDxx parameters extraction algorithm, in your “SPICE MODEL FILE” screen, HiSIM model should be selected (Figure 22).

In INTCAP “Fitting Variables” screen (Figure 23), set “qme/pgd” variable to - 0 for QME1,2,3 and PGD1,2 parameters extraction,

- 1 for QME1,2,3 parameters extraction only,

- 2 for PGD1,2 parameters extraction only.

Figure 22. Spice model selection screen	Figure 23. INT.CAP fitting variables screen

 

 

If variable “hisim_tox” is set to 1, TOX parameter will also be extracted.

To activate the extraction algorithm, when INTCAP-CGG characteristic is displayed, press “Options / Fit” from the “GRAPHICS” screen.

Warning: QME and PGD extraction algorithm requires the following parameters to be known precisely: LP, NSUBP, BGTMP1, BGTMP2. Be sure they have been extracted and that the corresponding values are in the “Optimized column” before using it.

 

Conclusion

A total of 9 local optimization strategies and one direct extraction routine for HiSIMv1.0 model have been presented in this article. The local optimizations should not be used without any user’s modification. Go into each strategy and change the selected geometries in the “Geometry Selection Screen” according to the available devices, and the observation of the “VTH dependence on L”. The user should also check if the region of interest specified in each strategy is properly selected by the criteria defined in the “Target selection screen”.

 

References

1. M. Miura-Mattausch, H.Ueno, H. J. Mattaush, H. Kawano, D. Kitamaru, K. Hisamitsu, T. Honda, S. Matsumoto, D. Miyawaki, H. Nagakura, S. Nara, D. Navarro, T. Okagaki, S. Ooshiro, Y. Shiraga, K. Suematsu, M. Suetake, M. Tanaka, Y. Tatsumi, T. Yamaoka, S. Kumashiro, T. Yamaguchi, K. Yamashita, N. Nakayama "HiSIMv1.0 User’s Manual" STARC
2. Silvaco International, "UTMOST III Extractions Manual Volume 1, MOSFET Modeling routines"
3. Silvaco International, "UTMOST III User’s Manual"

 

Contacts

Hiroshima University

Silvaco International, http://www.silvaco.com

STARC, http://www.starc.jp/