Roland A. - PVD for microelectronics (779636), страница 68
Текст из файла (страница 68)
. ,,0.:,2]%oFIG. 10.6 Three-dimensional view of flux transmitted through a collimator at a short distance belowthe collimator [ 10.15 ].R. POWELL AND S. M. ROSSNAGEL3620.089"',",'i,','L' ,,-/~/\/10.06ot--(D9/,-'\\~".!,~-----.i'---~TitaniumTungstenI/Iu.. 0 . 0 4i'r"0.020.00-90FIG. 10. 7//-60', "-,.-300306090The relative divergence of sputtered Ti and W species at 7 mTorr [10.16].10.3.1 LINE-SEGMENTMODELSThis approach is intrinsically two-dimensional, which is appropriate forlong trenches or perhaps circular holes. The substrate surface is broken upinto many hundreds of short line segments. The length of the line segmentsis arbitrary, but typically models use a few hundred line segments to modela feature several microns wide, resulting in approximately 60 atoms perline segment. The deposition process occurs by randomly choosing a linesegment and allowing some amount of film material to land on the segment: The amount is determined by consideration of the incident fluxes(perhaps a result of the transport models described above) as well as theposition and angle of the line segment.
In addition to deposition, it is alsopossible to include diffuse or specular reemission from the entire segment,which might be due to either nonsticking of the incident flux or evaporation of substrate material. Specular reemission would be most probable forgrazing-angle deposition that might occur on the steep sidewalls of a feature. In many cases, it is also possible to consider resputtering of the filmon the line segment. This might be due to the effects of inert gas bombardment, energetic neutral bombardment, or energetic, depositing metalspecies such as might be present during ionized PVD.
The material resputtered from the film can be tracked and redeposited on nearby line segments. The line-segment approach uses bulk values for such things as thePROCESS MODELING FOR MAGNETRON DEPOSITION363sputter yield, reflection probability, sticking coefficient, etc., which is anobvious limitation from a physics point of view.The overall film topography is then determined by tracking theseprocesses at the randomly chosen deposition/etching sites over a period oftime. In general, it is also necessary to develop certain criteria for dealingwith discontinuities in the film profile, and the details of these effects aredescribed in many articles [ 10.17-10.20].As an example of the two-dimensional, line-segment approach to deposition modeling, Fig.
10.8 shows the sequence of conventional sputter deposition into three trenches varying in aspect ratio from 0.5 to 2.0 [ 10.21 ].The mostly isotropic nature of the deposition can be seen, particularly withthe higher aspect ratio feature, in the build-out of the upper sidewalls of thedeposit and the eventual closure and void formation.
This modeling approach has also been used to combine the effects of a neutral, isotropic deposition with an ionized, directional deposition. This is shown in Fig. 10.9,in which the relative ionization is varied from 33% to 67% of the total depositing flux [ 10.19].A final example of this approach explores the effect of resputtering ofthe depositing film and the subsequent deposition of the resputtered material (Fig. 10.10). This is particularly important in I-PVD cases where thedepositing metal ion energy has a measurable sputter yield ( > tens of eV).As can be seen from the figure, increased depositing ion energy results insignificant resputtering of the deposited film.
This forms bevels on theupper sidewalls of the trenches, and the resulting redeposition of thissputtered material on the opposite sidewall tends to close off the feature,forming a void. (Similar studies for resputtered liners were shown inSection 8.3.)The line-segment approach to modeling of the depositing film has several limitations, although it can be very useful for the prediction of topography in trenches. The model is intrinsically two-dimensional, whichmeans it is inappropriate for vias or complicated geometries (vias underlines, etc.) that are intrinsically three-dimensional.
Since it is not really anatom-based model, it is not possible to introduce physical effects such assurface diffusion, grain formation, or film structure. For those features, itis necessary to go to the molecular dynamics approach.However, even with its limitations, line-segment modeling has beenfound to be very useful in predicting the topography development duringcomplicated deposition conditions and can be used as a diagnostic tool either to understand and calibrate the degree of directionality and/or resputtering, or to imply other physical properties, such as effective sticking coefficients [10.22].
In the former case, the use of this type of model in364R. POWELL AND S. M. ROSSNAGELFIG. 10.8 Line-segment model showing conventional magnetron deposition (cosine flux) ontotrenches of aspect ratio 0.5 to 2.0 [ 10.211.PROCESS MODELING FOR MAGNETRON DEPOSITION365FIG. 10.9 Model of ionized PVD deposition for the case of (left-to-right) 33% ions and 67% neutrals, 50% ions and 50% neutrals, and 67% ions and 33% neutrals [10.19].conjunction with experimental samples has been used to "measure" suchplasma properties as the degree of ionization of the depositing atoms or thefunctional density and temperature of the neutral species.1 0 . 3 . 2 MOLECULAR DYNAMICS FILM GROWTH MODELSThis class of models approximates both the arriving flux and the alreadydeposited film as an array of small disks.
Pioneering work by Mueller atCSIRO mapped out a wide variety of physical deposition phenomena usingModeling of increased ion energy, resulting in a higher etch-to-deposition ratio forI-PVD deposition [ 10.19].FIG. 10.10366R. POWELL AND S. M. ROSSNAGELthis approach [10.23]. An example of the microscopic surface is shown inFig. 10.11 [10.16]. In this case, a new disk is incident on an existing filmstructure.
Potential sites for adsorption are shown in the figure adjacent togray disks, which represent nearest potential neighbors. The incident diskhits the surface and relaxes into some nearby site based on a calculation ofthe lowest surface energy.One of the intrinsic advantages of this approach is that it allows the development of a physical structure to the film. If, for example, the substrate/film surface is cold and no rearrangement due to surface diffusionoccurs, the film may show a very columnar structure. As the substratetemperature is increased, adatom diffusion occurs and the resulting grainsare much wider and less columnar.
(This was described in detail in Section 7.2.1 and Fig. 7.3.)This approach can also explore deposition at angles other than 90 ~.Fig. 10.12 [ 10.16] shows the tilting of the intrinsic columnar structure of acold deposition as the incident flux is inclined by 30 and 60 ~. This couldalso readily show the effect of an asymmetric deposition into trench features, similar to what might occur at a deep trench near the edge of a waferusing a long-throw, low-pressure deposition [ 10.16, 10.22]. (This was described in Section 6.2.) A related example shows the film coverage andstructure for deposition over a step when the depositing flux is 5 ~ fromnormal.
This would be similar to deposition in the middle to outer regionsof a 200-mm wafer during a long-throw deposition. The simulations showFIG. 10.11 Schematic of deposition using a surface composed of disks approximating groups ofatoms [ 10.16].PROCESS MODELING FOR MAGNETRON DEPOSITION367FIG. 10.12 Deposition at 0, 30, and 60 ~ angles of incidence (from the surface normal) with a fixeddiffusion length of 0.02 microns (similar to Fig. 7.3d) [ 10.16].a shadowing effect on the side of the feature away from the depositing flux(Fig. 10.13) [10.16].This same approach can be used to examine the topography of depositsinto deep features such as trenches. One of the first things that is seen withthese simulations is the low density of columns on the steep sidewallswithin a feature (Fig. 10.14a).
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
