F.1 Local Potential Perturbation



next up previous contents
Next: F.2 Gate-Bias Shift Up: Appendix F: Theoretical Consideration of Previous: Appendix F: Theoretical Consideration of

F.1 Local Potential Perturbation

We will analyze the first problem in detail. A connection with the second problem will be established later. When the charge is absent, the potential in the oxide and bulk, denoted as , is determined by the two-dimensional Poisson equation. A localized interface charge induces a perturbation in the system. Considering the perturbation to be small, the Poisson equation enables that a relationship

 

may be written for the bulk. We assume -type bulk, uniformly doped in concentration , with a neutral hole concentration ; . As clarified above, we further adopt the total depletion approximation. Equation F.1 simplifies to the Laplace equation which also holds in the oxide. Having the interface charge, the problem consists of solving the equation

 

in the interval , , with the permittivity for and when . is the surface charge density in the plane . The problem with the coordinate system is shown in Figure F.1. The interface charge is distributed in the interval in a space charge density of . The boundary condition is at the gate contact. For the sake of simplicity we adopt that the interface charge does not modulate the local depletion region width (the is connected with the distribution of ). It follows that holds. To solve the problem one may apply the superposition theorem [375]. Let denotes the charge at the position : . This charge induces a potential at a particular point . The total potential at the point , induced by the whole interface charge-sheet, is given by

 

Note that we have to solve a two-dimensional problem; the conditions shown in Figure F.1 extend in the -coordinate to infinity. The potential perturbation, connected with the line charge with strength constant in the -direction, is determined by the equation

 

where we abbreviated . Equation F.4 is solved in this work by a standard technique consisting of employing an integral transformation which converts the primary equation to a simpler form [311]. Other an approach is to use the method of images. In this problem we have to consider an infinite number of images in two image-planes, accounting for the presence of two dielectrics with different permittivity.
As integral transformation we choose the Fourier transformation, as done in [331][40] by solving the three-dimensional point-charge problem. Here we employ

 

The local coordinate system used henceforward has its origin in , Figure F.1 (right). Since the problem does not depend on the -coordinate, the left-hand-side of F.5 yields

 

Remember that, consistent with F.5,

 

holds for the Dirac "function" in one dimension. Replacing F.5 into F.6, with benefit of F.7, we obtain

 

In solving equation F.8 we distinguish three intervals

  1. ; semiconductor region. The general solution is given with

     

  2. an infinitely small interval , surrounding the interface at . After one integration of F.8, with benefit of the continuity of and applying , we obtain

     

  3. ; oxide region. The general solution reads

     

From the boundary conditions for it follows: and . The unknown coefficients follow after replacing the latter conditions and the relation of the continuity of at the interface in expressions F.9 - F.11. The solution in the semiconductor becomes

 

The potential is given with the inverse Fourier transformation

 

Since holds for F.12, it follows

 

in the half-plane . An equivalent expression may be written for the oxide. Although solution F.14 represents only an intermediate step in our derivation of the total potential perturbation, it is worthwhile to discuss the qualitative behaviour of this expression. Let us assume a finite and . Solution F.14 exists for all , as well at and , while it does not exist in the origin , , where the source of the field is placed, as can be easily concluded by analyzing the behaviour of the subintegral function for and gif. Note that the solution to the three-dimensional point-charge problem, where the point charge is located at the oxide/bulk interface, has equivalent properties. If , but is finite, solution F.14 exists in all points except the origin . Particularly, on the axis connected with , it is given in explicit form by

For the three-dimensional point-charge problem we found in this case

while the solution does not exist at . If both, and the integral in F.14 diverges for all and , while for the three-dimensional point-charge problem the expression we derived reduces to the Coulombic potential form: . For these special cases the solutions are known from elementary courses.

Replacing from F.14 in F.3, after an integration with respect to , one obtains the total potential at an arbitrary point in the bulk induced by the whole charge-sheet at the interface

 

For simplicity, but without much loss of generality, a uniform charge density is assumed for F.15. In order to prove the integrability of F.15 let us consider the subintegral function :

, which is a finite value.

.

Since is convergent for all when and is a continuous function, the integral in F.15 is convergent. Moreover, it has a finite value for and (in the origin), where we are interested to calculate the surface-potential perturbation. As a conclusion, F.15 exists everywhere.

We focus on the solution at the oxide/bulk interface , at the middle of the charge distribution . To further simplify the result let us suppose that . This condition is fulfilled in depletion and inversion in common cases. Setting , one obtains

 

with and the reciprocal effective permittivity for the oxide/semiconductor system .

When the charge-sheet extends to infinity, formulae F.15 and F.16 should provide the known relationships. To check this, we apply on F.15. Taking advantage of the relationship

and the evenness of , expression F.15 reduces to

 

At the interface, ; the charge is stored in two capacitors, oxide and bulk, connected in parallel. For we get , as expected. This result holds when applying on F.16, as well.

It is not a trivial task to calculate in F.16. An approach, by which many aberrations can be avoided, is to employ the expansion

 

Because of , this series converges for . Taking advantage of F.18 and using an intermediate result from [152] (3.945-3), one arrives finally at

 

Note that the case has been analyzed separately. The series representation in F.19 converges efficiently and together with F.16 represents the model for the surface band-bending in the middle of the localized uniform charge, when the depletion region width is large.

The potential of the theoretical formulation is confirmed by a comparison with the exact two-dimensional numerical solution, Figure F.2. The numerical results are computed by MINIMOS, assuming an -channel MOSFET with uniformly doped bulk and long channel (). The localized fixed interface charge is placed in the middle of the channel, thus avoiding an accidental impact of the source and drain junctions.

In two calculations, we fixed the gate voltage so that the bulk was in depletion and then we calculated changes in the surface potential at the middle of the trap-region; the results are points shown in Figure F.2. For the data sets, denoted as , the gate bias has been adapted to establish a surface electron concentration of at the middle of the trap region. Therefore is kept constant for all points in the latter casesgif. While the former data sets (, ) correspond to the problem 1, the latter (, ) represent the problem 2, mentioned at the beginning. The calculations are carried out for and , with no significant effect of on the results. Since a positive charge of induces a large band-bending of several hundreds , care has been taken to ensure that the bulk was in depletion for all points. For and data the bulk dopant concentration was , whereas for and data we assumed a low concentration of to remove the depletion region edge away from the interface. The corresponding depletion width

 

(where being with respect to potential deep in the bulk) is calculated to be , , and for , , and data sets, respectively. Solid line represents the analytical model for an infinitely large , given by F.16 and F.19. The dashed line is the theoretical result for expressed by F.21, as will be explained. All results in Figure F.2 are absolute values normalized with . Since the ratio is large for numerical data, the agreement between numerical and theoretical approaches is excellent. Small differences come out from a finite in the numerical calculation, because F.16 is strictly valid if . For a finite the analytical model is given by F.15, but no simple analytical solution of this integral has been developed up to now. The numerical modeling ( points) shows that a finite tends to reduce the surface band-bending, but only moderately. In fact, the extreme value of is not , but which is less. The total effect of a finite is to increase the surface band bending with respect to its extreme value, as is clarified below.



next up previous contents
Next: F.2 Gate-Bias Shift Up: Appendix F: Theoretical Consideration of Previous: Appendix F: Theoretical Consideration of



Martin Stiftinger
Sat Oct 15 22:05:10 MET 1994