Conclusion:A surface-potential-based model 1,2 is developed for the cylindrical surrounding-gate MOSFET (Fig. 1), of which the (non-differential) equation for the surface potential had not been known because of the cylindrical structure. Unlike other surrounding-gate MOSFET models, this model includes both drift and diffusion currents, 3,4 and also there is no inherent distinction between saturation and non-saturation.3-5 Its accuracy is demonstrated by comparison with device simulation without arbitrary fitting. The model could potentially be useful as a basis for developing compact models, just as the planer single-gate counterpart 1 has formed the theoretical foundations for the development of compact planar MOSFET models.
6-8Details: We formulate the model for a uniformly doped nMOS device assuming nondegeneracy using the gradual channel approximation. The gate oxide capacitance per unit length is given by C ox = 2π ox / ln(1 + t ox /a). The semiconductor charge Q s per unit length iswhere V GS is the gate voltage (source grounded), V fb is the flat-band gate voltage, and ψ s (z) ≡ ψ(a, z) is the surface potential. The semiconductor charge consists of mobile charge (electrons) and immobile charge (acceptor ions):If we use the depletion approximation,A is the ionised acceptor density, and d is the surface depletion depth. Note that the expression of d for the planar MOSFET [Eq. (4) of Jang and Liu 5 ] cannot be used for the cylindrical MOSFET. 9 We hereafter assume that the channel is fully depleted of majority carriers (holes), leading to Q im −πa 2 qN − A . The drain current I D (z) is proportional to Q mo (z) and the gradient of the electron quasi-Fermi potential V n (z):, where µ n is the electron mobility, and ∂V n ∂r = 0 is assumed (charge-sheet approximation). This equation includes both drift and diffusion currents. In a steady state, the electron density is given by, where n i is the intrinsic carrier density, β −1 = k B T /q is the thermal voltage, and V n is the electron quasi-Fermi potential. After some algebraic manipulation, I D can be written aswhere µ eff is the effective mobility, 6 and ψ sL ≡ ψ s (L) and ψ s0 ≡ ψ s (0).The surface potentials ψ s0 and ψ sL are needed to use Eq. (2). We therefore derive an equation for the surface potential. The Poisson equation for the fully depleted channel is, within the gradual channel approximation,whereA /n i ) and V np = V n − V p is the difference between quasi-Fermi potentials (Fig. 2). The boundary conditions for Eq. (3) In deriving Eq. (4), we did not make the charge-sheet assumption, and therefore the electric flux should be continuous at the interface between the channel and the gate oxide (if there is no interface trapped charge): (r, z). Then, from the second boundary condition,In order to perform the integral in Eq. (4), we use a charge-sheet model. In the charge-sheet model, mobile charge Q mo (z) is localised to the surface, and the electric flux is discontinuous there. Note, however, that ψ(r, z) is still continuous at r = a. The Poisson e...