5  Concentration Correction and Material Conservation

As a result of the numerical error in the FDM, the sum of the concentrations drift from unity and some of their values may be negative. There accumulated effect of the dose-stepping method can also result in a significant loss or gain of material. IMPETUS II employs a correction method and a conservation method. The main philosophy of these methods is that their alteration of the material distribution arrived at by the finite difference method should be just sufficient to bring them within the bounds of the natural laws, transgressions in the natural laws are resolved as locally as possible and the alteration must not create oscillations in the solution.

5.1  Correction Method

The numerical approximations inherent in the finite difference method have the consequence that solution concentrations after each dose step will generally sum to something other than unity at all grid points and, at some grid points, the concentration will take a negative value. The most straightforward method of correcting the solution is as follows:
for i=1,2,...,n do
        [^(q)]ij ¬   max([^(q)]ij,0)

for i=1,2,...,n do
        [^(q)]ij ¬ [([^(q)]ij)/(si=1n [^(q)]ij)]
where the original qij for i=1,2,...,n, j=1,2,...,N are the approximations to the concentrations resulting from the FDM. This is the method employed in the original IMPETUS. The main criticism of this method is that material is effectively added and removed arbitrarily, any tendency to conserve material inherent in the finite difference approximation (that is when summing all positive and negative conventrations) to the mathematical model is lost.
If it is assumed that there is a tendency of the FDM to conserve material in [0,xR] then the correction method must be such that the the total i-material is almost unaltered. The following method was developed to correct the [^(q)]ij and without altering the total material. In the method, the correction starts at j=1 and progresses to j=N. At any stage j, gi is used to represent the 'gain' in the i-material as a result of the correction method.
for i=1,2,..,n
        gi ¬ 0
for j=1,2,..,N
        for i=1,2,..,n
                [^(q)]*ij ¬ [^(q)]ij
        for i=1,2,..,n
                [^(q)]ij ¬   max([^(qij)]-gi,0)
        for i=1,2,..,n
                [^(q)]ij ¬ [([^(qij)]-gi)/(Si=1n [ [^(qij)]-gi])]
        for i=1,2,..,n
                gi ¬ gi + [^(q)]ij - [^(q)]*ij
On completion of the method the [^(qij)] values have clearly been corrected. The gi will generally be non-zero on completion of the method so that the correction method does not exactly conserve material. However, the gi is simply an estimate of the material gain(loss) and the estimated gain(loss) of any material in this method will generally be minute in comparison to the that in the earlier correction method. This latter method is employed in the IMPETUS II code.

5.2  Conservation Method

At any dose the conservation law requires that the material in the structure, and the sputtered material are equal in quantity to the material of the original structure and the total bombarding ions. More precisely, the law requires that the material in any region of the material at any dose f2 is equal to the material at an earlier dose f1, plus the material that has flowed into the region over that dose interval minus the material that has flowed out.
The determination of the quantity of material requires integrations of the yield-dose history and the concentration depth then numerical integration methods need to be employed and hence we must be mindful that only an approximation to material quantities can be made. Hence an exact conservation of material cannot be carried out. Generally, there is a marginal gain or loss in materials as a result of the computational solution.
The flow fi(x,f) of i-particles towards the surface at depth x and dose f is given by
fi(x,f) = (q(x,f)+ u(f)) qi(x,f)   (x,f) Î [0,¥) ×[0,¥) .
(14)
The flow of particles through the surface is precisely the yield, so that yi(f)=fi(0,f). Conservation of material requires that in any subregion [x1, x2] Í [0,¥) and over any dose interval [fA,fB] Í [0,¥) the i-material present in [x1, x2] at dose fB must equal the i-material present at dose fB with any net flow of i-material that has flowed into the region. That is the following equation must hold:
ó
õ
x2

x1 
qi(x,fB) dx = ó
õ
x2

x1 
qi(x,fA) dx + ó
õ
fB

fA 
(fi(xA,f) -fi(xB,f)) df
(15)
As discussed earlier, we cannot necessarily expect the above equation to hold after the application of the FDM
In IMPETUS II it is assumed that the linear diffusion methods of the previous section effects a negligible gain or loss of material and the materials are accounted for only in the domain of the FDM, after the completion of every 2 d dose step. Let [x1, x2] Í [0,xR] then the gain in i-material gi over the dose step f* ® f*+ 2 d is approximated as follows:
gi » \sf simp[ ó
õ
x2

x1 
qi(x,phi*+d) dx ] -\sf simp[ ó
õ
x2

x1 
qi(x,f*) dx ] - 2 d( fi(xA,f) -fi(xB,f))
where simp represents a Simpson rule approximation. Conservation of i-material can now be effected by a proportional decrease (increase) to correct the gain(loss) in the discrete qi(x,f*+d) values.
The technique for conserving material could theoretically be applied across any subrange of [0, xR]. However, if it is applied over every region [(j-1)D, jD] then local numerical errors can be magnified. Equally, if it is applied over a wide range then the redistributive effects of the technique tends to shift material from part of the region to another in practice. The method finally adopted in IMPETUS II was that of evenly dividing the domain of the FDM into its subranges [x1,x2] and, if there are interfaces within the region, a division of the region is placed there. This latter point ensures that the redistributive effects of the method are minimised. The conservation technique is independently applied in each sub-range.

5.3  Implementation

In IMPETUS II the conservation/correction method is applied after every d dose step. Note that it is difficult to satisfy all the physical constraints simultaneously. Note also that in the conservation method, the total i-material in the region (for each i) is not precisely obtained and hence it is infeasible to attempt to conserve material exactly.
In the IMPETUS II code the FDM solution is processed by first applying the correction, then the conservation method. Unfortunatley, the conservation method may have the effect of returning concentrations that sum to a value greater than one, so the correction method is applied a second time. After this process it is clear that the concentrations are non-negative and sum to unity. However, although the one application of the conservation method was applied, the second correction method will generally imply some gain (or loss) of material (although it is generally much smaller than the original gain (loss)). In order to apply the conservation principle rigourously, any measured gain(or loss) in in each material after each d dose step is recorded and an effort is made to remove (replace) the material at the following dose level.

5.4  Results from test problems

A good indication that materials are conserved in IMPETUS II is given in yield-depth curves of figure 6 resulting from the original material structure of figure 5. Each of the three layers contains equal volume concentrations of two materials and this is reflected in the results.
Further tests on IMPETUS II each consisting of 100 angstrom layer with sharp interfaces, the first interface at a depth of 200 angstroms were generally found to conserve material to within 1% in both material systems tested. For thin layers and sharp interfaces close to the surface, as considered in sections 2 and 3, the conservation rule is still more difficult to enforce and a greater proportional loss of material may be observed in these cases.