Sie sind auf Seite 1von 3

Excerpt from the Proceedings of the COMSOL Multiphysics User's Conference 2005 Paris

Serge Gidon

Double Gate MOSFET Modeling

Abstract We use the MOS transistor model from COMSOL as a template to do our own model of double gate MOSFET. Double gate transistors are developed to resolve short channel effect problems in actual MOSFET structures [1,2]. So that such architectures are directly related to the constant reduction of the feature size in microelectronic technology. At the present time, it seems that double gate devices- going to non-planar transistor architectures- could be a solution for sub-32nm nodes. In addition, new design flexibility is allowed when gates are not interconnected. However, appropriate models must be developed In our investigation (yet 2D), we overcome the high aspect ratio of the transistor (thin channel compared to its length), by introducing an anisotropy scale factor in its geometry description. The model shows the involved phenomenon (appearance of channels) and gives the intended drain current curves versus gate voltages: Id (Vg1,Vg2) using a dedicated written Matlab file. Next step might be to study 3D designs Keywords: MOSFET, Double Gate, FinFET, Gate All Around, Comsol Multiphysics 3.1

Source contacts
G1 G2

Drain contacts

Oxide
Figure 1 : MOS thin FET geometry (Y coordinate being magnify by a factor 5)

In fact, the channel is so thin (compared to length) that it is interesting to magnify the Y direction by an anisotropy factor (G=5 in our case) to allow a clearer view of the data (current lines,) and to equilibrate the mesh. This coordinate transform is usually incorporated into the material parameters (diffusivity, permitivity,). Depending on a convention used to determine the results, the anisotropy matrix may be written with G2 in its second diagonal term. The diffusion of both carriers species can be taken into account in the simulation. However, considering that majority carriers (P for the modelised NMOS) do not contribute significantly to the current flow, it is efficient to reduce the system of equations to the ones of Poisson and N carriers drift-diffusion process. Though Comsol Multiphysics includes a script solver to build a solution step-by-step (used in the transistor model, for example, to initially estimate a distribution of the voltage in the structure), we use a Matlab mfile to determine the fundamental electrical characteristics of the transistor: Id (Vg1, Vg2), the drain current versus gate voltages considered as parameters (voltages being referenced to the one of the source). We choose to determine these curves in the linear region, Vd=5 mV and in the saturation region Vd=1V so that the transistor will not be in saturation. Three regimes can be noted in this case. When the two voltages are low, each interface is in weak inversion. When one voltage is low and the other is high, one interface is in weak inversion and the

1 Transistor geometry and related considerations


The MOS is described in a 2D cross view. Front and back grids are respectively shown on top and bottom of the channel (P doped). Linearly Implanted areas with smoothly varied doping concentration face to the double sources and Drain contacts.

S. Gidon CEA Grenoble Leti Tel: +33)4 38 78 99 51 Fax: +33)4 38 78 50 46 E-mail: serge.gidon@cea.fr

Excerpt from the Proceedings of the COMSOL Multiphysics User's Conference 2005 Paris

second is in strong inversion. In this two last case, an interface coupling exists. Each canal current depends on the both gate voltages. The last case is when both voltages are high. Both interfaces are in strong inversion, no interface coupling exists. Each channel is independent. The algorithms, first gives access to a panel of initial conditions (concentrations and potential) using Vg1=0, Vg2=[0:0.2:1]. Secondly, depending on the initial conditions, different Id curves versus Vg1 are determined using the Comsol Multiphysics parametric solver.

2 Results
Figure2 shows the current density (3D representation) and current lines in the previously described (distorted) geometry, using Vg1=Vg2=1V (strong inversion regime in saturation: Vd=1V). We see the double neck due to drain/gate potential that pinchs the current flow (see the carrier concentration on the bottom of the figure). We notice the current flow lines that climb in the channels.

Figure 3: Logarithmic scale carrier density cross section (at the middle of the MOS). We notice the appearance of both channels near Vg1=Vg2=0.5V (Vd=5mV)
9 8 7 6 5 4 3 2 1 0 -1 Id Vg2=1 x 10
-5

Id(Vg1)

Vg2=0.8 Vg2=0.6 Vg2=0.4 Vg2=0.2 Vg2=0

0.1

0.2

0.3

0.4

0.5 Vg1

0.6

0.7

0.8

0.9

Figure 2: Current density and carrier concentration in saturation regime. Figure 4 (over) shows linear regime (Vd=5mV) in linear representation, for different values of voltage Vg2. Insert shows carrier density (in logarithmic scale) for Vg2<>Vg1.

Excerpt from the Proceedings of the COMSOL Multiphysics User's Conference 2005 Paris

4 Annexe: Detail of the written solver script


% Solver script fem1.xmesh=fem.xmesh; fem1.sol=asseminit(fem1,'u',zeros(flngdof(fe m),1)); fem0.xmesh=fem.xmesh; fem0.sol=asseminit(fem0,'u',zeros(flngdof(fe m),1)); init = asseminit(fem,'u',fem0.sol); fem.sol=femnlin(fem, ... 'init',init, ... 'nullfun','flnullorth', ... 'solcomp',{'phi0'}, ... 'outcomp',{'lm3','phi','phi0','nc'}, ... 'pname','Vd', ... 'plist',[0.005], ... 'hnlin','on'); fem0=fem; init = asseminit(fem,'u',fem0.sol); fem.sol=femnlin(fem, ... 'init',init, ... 'nullfun','flnullorth', ... 'solcomp',{'lm3','phi','nc',}, ... 'outcomp',{'lm3','phi','phi0','nc'}, ... 'pname','Vd', ... 'plist',[0.005], ... 'hnlin','on'); fem1=fem; %Determination of initial solutions at Vg1=0 % Constants fem.const={'Vg1','0'} fem.sol=femnlin(fem, ... 'init',fem1.sol, ... 'nullfun','flnullorth', ... 'solcomp',{'lm3','phi','nc'}, ... 'outcomp',{'lm3','phi','phi0','nc'}, ... 'pname','Vg2', ... 'plist',[0:.2:1], ... 'hnlin','off'); fem2=fem; %Loop on Vg2 values and variation on Vg1 for nn=1:6 % Constants fem.const={'Vg2',0.2*(nn-1)}; init = asseminit(fem,'init',fem2.sol,'solnum',nn); fem.sol=femnlin(fem, ... 'init',init, ... 'nullfun','flnullorth', ... 'solcomp',{'lm3','phi','nc'}, ... 'outcomp',{'lm3','phi','phi0','nc'}, ... 'pname','Vg1', ... 'plist',[0:.05:1], ... 'hnlin','off'); %Table of the fem strutures femT(nn)=fem; end

Figure 5 shows the linear regime (Vd=5mV) in logarithm representation, for different values of volatge Vg2.

3 Conclusions
Comsol Multiphysics algorithmic allows modelling the drift diffusion process involved in semiconductors physics. We show that a thin FET 2D geometry or a double gate transistor can be investigated. The next step, might be to study 3D gate MOS architectures such as finFET or Gate All Around

Figure 6: Example of 3D MOS grid architecture Acknowledgements Thanks Erik Danielson (COMSOL) for his MOS model, Gilles Lecarval (Leti) for his tutorial and Olivier Rozeau (Leti) for his motivating suggestions. References

[1] Bonded Planar Double-Metal-Gate NMOS Transistors Down to 10 nm, M. Vinet, T. Poiroux, J. Widiez, J. Lolivier, B. Previtali, C. Vizioz, B. Guillaumot, Y. Le Tiec, P. Besson, B. Biasse, F. Allain, M. Cass, D. Lafond, J-M. Hartmann, Y. Morand, J. Chiaroni, and S. Deleonibus, IEEE Electron Device Letters, vol. 26, no. 5, may 2005 p. 317 [2] M. Vinet, T. Poiroux, J. Widiez, J. Lolivier, B. Previtali, C. Vizioz, B. Guillaumot, P. Besson, J. Simon, F. Martin, S. Maitrejean, P. Holliger, B. Biasse, M. Cass, F. Allain, A. Toffoli, D. Lafond, J. M. Hartmann, R. Truche, V. Carron, F. Laugier, A. Roman, Y. Morand, D. Renaud, M. Mouis, and S. Deleonibus, Planar double gate CMOS transistors with 40 nm metal gate for multipurpose applications, in Proc. Solid State Devices and Materials Conf., 2004, pp. 768769.

Das könnte Ihnen auch gefallen