Investigating Two Dimensional Implantation Effects Using The BCA Model

I. Introduction

The Monte Carlo Binary Collision Algorithm (BCA) implant module in ATHENA was introduced in an earlier article [Simulation Standard Vol 10, #5, May 1999] which discussed the accuracy of the simulation model compared to one dimensional measurements using Secondary Ion Mass Spectroscopy (SIMS). In this article we discuss the two dimensional effects that can be simulated using the BCA module that cannot accurately be simulated using analytic models.

In the BCA model the deflection of the moving particles is calculated in a strict binary way - between the moving ion and the closest atom in the lattice. The difference between the BCA model and earlier implemented Monte Carlo crystalline model is in the precise determination of the closest atom and more accurate calculation of impact dependent inelastic energy losses. The BCA model therefore excels in the prediction of well-channeled particle trajectories which results in close agreement with experiments, especially in those cases where channeling processes are dominant, such as for zero degree or 45 degree implants.


II. Device Geometry Effects

The new BCA Monte-Carlo Impant Model is a true 3D simulation which can account for all crystal directions in the crystal lattice. In Athena, the 3D simulation is then Integrated in one direction to achieve a 2D profile.

The BCA model also takes account of the vacuum above the structure, such that if an ion is transmitted through one part of the structure, it will continue to be simulated if it then impinges into the structure later on. An example of this is shown in Figure 1, where a high angle, 15keV boron implant at 45 degrees passes right through the top corner of a poly gate and is then implanted into the silicon below through a 100A gate oxide. Only a single "pencil" beam aimed at the top corner of the polysilicon was modelled in this instance to show the shaddowing effect more clearly. Notice how the BCA model has taken account of the beam scattering from the polysilicon. The lateral spread of boron in the silicon is caused by the single point ion beam scattering after it passes through the top corner of the polysilicon. The BCA module also has knowledge of the crystalline or amorphous nature of the material being implanted in each part of the structure, including unannealed damage caused by previous implants as will be demonstrated in section III.

The Shaddowing effects demonstrated in Figure 1 can play a significant role in the final doping profile of high angle implants which are commonly used in MOSFET processing. The BCA module should be used for the highest accuracy in these cases, especially when the high angle implant is crucial to the correct modelling of the electrically active region of the device.

Figure 1. Showing device geometry effects. The Ion Beam passes through the top corner of the poly gate and into the silicon substrate below. In order to account for these effects, BCA also models the vacuum above the device.


In Figure 2 a more realistic case of the same high angle implant is demonstrated. Here the whole structure is implanted at 45 degrees and zero degree rotation. Two effects are now seen in this example. The first effect is that the shaddowing of the polysilicon gate can be seen clearly with an obvious angle in the concentration profile at the edge of the gate.

Figure 2. The same implant conditions as Figure1 but a blanket implant instead of a pencil beam. Channelling and shaddowing effects are clearly demonstrated. Note the difference in implant depth between the polysilicon and the crystalline substrate.


The second effect is channelling at 45 degrees as this is an exact channelling angle for <100> crystalline silicon. The implant in this case penetrates much deeper in the areas where it is not scattered by the polysilicon as was the case in Figure 1. Notice how the depth of the implant in the polysilicon gate is significantly less than the depth of the implant into the silicon. This is due to the non-aligned grain structure of the polysilicon that prevents channelling from occurring.

Figure 3 shows the same implant but this time at zero degrees which is another channelling direction for <100> silicon. Once again, notice the difference in the implant depths between the scattered implant in the polysilicon and the channelling that occurs in the crystalline substrate.

Figure 3. A zero degree tilt implant into the same structure for comparison.


It is important to control the beamwidth for simulations where significant channelling occurs. It should be stressed here that experimental results for well channeled implants are quite sensitive to many factors including surface conditions (thickness and uniformity of oxide surface layer) and precision of the ion beam orientation and beamwidth. Even 0.5 degree deviation in these parameters could result in considerable changes of measured implanted profiles. For this simulation, the beamwidth parameter has been set to 0.1 degrees.

Analytic implant models fall short of Monte Carlo implant models in these cases for two reasons. The first shortfall of analytic models is that they do not account for channeling other than in the normal direction. Channeling profiles have been well modelled for normally incident ions in [100] and [110] directions. However, analytic models cannot account for large angles from the normal which would have increased channeling effects in other crystal directions. Hence, for a <100> silicon substrate case where for a large angle implant, no channeling in the [110] direction (or any other crystal direction) would be calculated.

A second shortfall of analytic models is the lateral straggle which is not well defined. This is particularly important for LDD and source/drain implants, because of the lateral straggle of the implant underneath the poly gate. This effect is not significant at large gate lengths, but as gate lengths reduce to a quarter micron and below, MOSFET threshold voltages could be inaccurate if the BCA model is not used.


III. Implantation Damage Effects

In silicon processing, the implant depth profile of a species can depend on previous processing steps. In this example, two implants are implanted at zero degrees tilt to emphasize the effect. In the first instance, 10keV Arsenic is implanted into undamaged silicon, followed by 10keV boron. Both implants also have the same dose. In the second instance, the implant conditions are identical, but the order of the implants is switched, such that the boron implant is carried out first.

The two implant species, boron and arsenic, are very different in mass and therefore cause different amounts of damage for a given dose and energy. Arsenic is heavy compared to boron which therefore creates significantly more damage and will amorphise the surface for typical MOSFET source/drain implant doses.

If the boron is implanted after the arsenic, the boron is now effectively being implanted into an amorphous substrate such that little channelling occurs even at zero degrees tilt. The peak concentration of boron will thus now occur closer to the surface of the substrate. Figure 4 shows the results of these two experiments.

This is a clear demonstration of the accurate damage model in the BCA module. Clearly the boron implant does not damage the silicon to the same extent as the the arsenic implant, since the arsenic profile is little affected by the previous boron implant.

In order to produce a similar effect using analytic models, one would need prior knowledge of the damaging effect of previous implants. Thus, in this case, an amorphous model would be used for the boron implant if it was after the arsenic implant but a crystalline model would be used if it was the first implant. However, one would have to choose between these two extremes since there is no analytical model for partially damaged silicon. In this case, it can be seen that the analytical approach would be inaccurate, since in Figure 4a it is clear that the boron is not completely de-channelled. The BCA model on the other hand continuously takes into account the gradual build up of damage as it occurs during the implant.

Any bipolar or CMOS fabrication process that incorporates a high atomic mass ion implant followed by a lower atomic mass ion implant without a thermal anneal in between will suffer from the effects demonstrated above. This problem is enhanced for very low energy implants used on today's aggressive deep sub-micron technologies. We strongly recommend the BCA module for accurate process simulation in these cases.


IV. Conclusion

Two dimensional ion implantation effects can be simulated accurately using the Monte Carlo Binary Collision Algorithm (BCA). The BCA model takes account of the vacuum above the device in order to model ions that fully penetrate one part of the device, travel through the vacuum region and impinge into another region. Un-annealed crystal damge due to previous implants is also correctly modelled. These effects are becoming increasingly relevant in the sub-quarter micron fabrication process where low energy implants are used routenely.