Roland A. - PVD for microelectronics (779636), страница 67
Текст из файла (страница 67)
ForPROCESS MODELING FOR MAGNETRON DEPOSITION~35510 -e<gi104191101 O01000Energy of Sputtered Particles (E)FIG. 10. ! Energy distributions of sputtered Ni atoms for Ar ~ b o m b a r d m e n t at 0 ~ (normal incidence)and 75 ~ (near grazing) 110.41.the case of sputtering and sputter emission, the TRIM-related models canpredict the average yield, the average kinetic energy of the sputtered particles, and whether the angular distribution is a cosine distribution or somevariant, such as an over-cosine distribution that has a larger relative fraction of the emitted particles emitted normal to the surface.The TRIM models are also useful in exploring other relevant ion-impactprocesses.
The most important to sputter deposition is the neutralizationand energetic reflection of sputtering inert gas species. This occurs primarily in cases where the substrate atomic mass is greater than the atomicmass of the incoming ion. The incident ions, which are neutralized as theyapproach close to the cathode surface, are nearly elastically reflected fromthe surface and leave the surface region with a significant fraction (2050%) of their initial energy. As these energetic reflected neutrals can impact the growing film, the energy and angular distributions of this flux canbe critical to film properties.356R. POWELL AND S. M.
ROSSNAGEL10.2 Transport ModelingSputtered atoms emitted from the cathode must pass through the background gas of the chamber, as well as potentially such geometrical obstacles as a collimator, on their way to the sample. Collisions with backgroundgas atoms will result in a general reduction in the kinetic energy of thesputtered atoms as well as a scattering or loss of directionality. Sputteredatoms that impact on physical surfaces such as a collimator or the chamberwalls are obviously no longer part of the depositing flux at the sample, andtheir loss will alter the net deposition process. Although there have beensome analytic approaches to modeling gas-phase transport [10.7-10.9],most groups have used a Monte Carlo approach [10.10-10.13].
In this scenario, single atoms of some known energy and direction are ejected fromthe cathode, and their collisions with various other atoms and surfaces arefollowed until the atoms are lost by condensing on a surface. After manytens of thousands of atoms are followed, it is possible to make some statistical assessments of the net transport trends.An example of this approach given in Chapter 6 showed the average kinetic energy of sputtered Cu atoms as a function of chamber gas pressure(Fig. 10.2).
The species of background gas is critically important due to itsrelative mass compared to the sputtered atom. Effectively, the higher themass of the gas compared to the sputtered atom, the more rapidly the sputtered atom is scattered. Somekh showed, in a classic early study, how increasing the mass of the working gas results in a more rapid quenching ofthe initial kinetic energy as a function of distance from the source [10.12].The results for the energetic fluxes are shown in Fig. 10.3. Similar studieshave focused on the reduction in the net number of energetic particlesalong with the increase in diffusing, thermalized atoms.
Early work byMotohiro and Taga shows both of these species as a function of the distance away from the source at constant pressure (Fig. 10.4) [10.13].In addition to determining the average kinetic energy of the sputteredatoms, two other features can be explored by modeling: geometrical filtering and species-dependent transport. An example of geometrical filteringis shown in Fig.
10.5, where the flux is modeled at it passes through thecell of a collimator. As atoms impact the surface of the collimator, theystick and are removed from the transmitted distribution. The angular divergence of the transmitted flux is significantly narrowed for the transmitted flux, compared to the case without a collimator [10.14]. The depositionrate is also strongly reduced (Fig. 10.5). It is then appropriate in this caseto explore the situation of a multicelled collimator.
Since the walls of eachPROCESS MODELING FOR MAGNETRON DEPOSITIONMC results, 350 KDh=510.6nmOh = 521.8 nmI0 . 2 A v e r a ~ ekincric energy 111'spultcrcd C u atoms ;IS a function of chanlher pressure. The results I)$' thc Monle Carlo calculation arc Ihc lillcd circlch. Expcrir~~cnttllnlcasurernents art. fromI 10. I01 and are shown a> opcn squilres and diunondh I 10.1 I 1.cell or hole in a collimator collect material. a sample position close to thecollimator should show the shadowing effect of these walls. This is shownin Fig. 10.6 for the case of a square collimator and a short sample distancebelow the collimator 110. IS].Since most transport codes are Monte Carlo in nature and track the dynamics of individual atoms, it is possible to compare the transport probability for atoms of different masses sputtered from the same cathode.
Thisis most important in cases where the two components of an alloy targethave widely differing masses. as, for example, would TiW and AlCu. I t isalso relevant in these cases that the heavier species has a mass that significantly exceeds the working gas (Ar, 40 AMU). and the lighter species issignificantly less than the working gas. This means that the lighter specieswill be more readily scattered and stopped by the background gas.
whereasthe heavier species will tend to push its way through more efficiently. Itwould then be expected that the transport for the heavier species would behigher and the angular distribution less isotropic. The result will be asmaller angular divergence for the heavier species compared to the lighterone (Figure 10.7) [10.16].R. POWELL AND S. M. ROSSNAGEL3589~,,,-9,,,; ....,-..,,,.... k,9,,--.,,.~.,,..10 510 5AICrl,.. vl.U10 5AIW10 4KrytonCr10 30166200Xe n o n300400P r e s s u r e Distance (Pa-mm.)FIG.
10.3 Kinetic energy (in degrees K) for sputtered atoms as a function of the pressure-distanceproduct 110.121.The net result of this unequal transport and differing directionality foreach species is potentially a pressure-dependent change in the film composition and, perhaps more importantly, a topography-dependent filmcomposition. In this latter case, the film stoichiometry in deep featuresmay be different from that on planar structures because of the difference indirectionality of the two species. However, this is for transport alone.When other factors at the depositing film surface are taken into account,such as reflection or resputtering from the film surface, the compositionalprofile may change (see Section 9.7).PROCESS MODELING FOR MAGNETRON DEPOSITION359100\< 0.2eV,Ag\Ag2eV0FIG.
10.4110.131.10Distance (cm)20Percentage of ballistic and thermalized ( < 0.2 eV) particles as a function of distance10.3 The Wafer SurfaceThe physical processes that occur at the film and wafer surfaces aremostly well known and characterized.
The fundamental problem withcontemporary computer models of film deposition and growth is that, dueto the limited, practical size of available computers, it has not really beenpossible to track each atom in the depositing film; there are simply toomany. This has led to two general classes of models: one based on an approximation of a surface as a series of short line segments and anotherthat uses aggregates of large numbers of atoms in the form of disks thatthen are used to construct the film.
The line-segment approach is intrinsically two-dimensional and is thus limited to modeling surface featuressuch as trenches, which are essentially infinite in one direction. The diskapproach, also known as a molecular dynamics approach, can be extended to three-dimensions, although most published work seems to be intwo dimensions.The incoming flux to a surface can be approximated by an examinationof the experimental conditions. For example, if the deposition was occurring at low pressure where the mean free path of the sputtered atoms wasmuch longer than the cathode-to-sample distance, the arriving atoms360R.
POWELL AND S. M. ROSSNAGEL0.5Or}c-=0.4!el,!>" 0.3f-ii",~0.2rrNO'v90-60-300306090Angle ( d e g r e e s )-o~100r:\.......,.....,o----o Experiment~SIMSPUDN~ 80 ~rrC.o._600a40._>rr"0.00.51.01.52.0Collimator Aspect RatioFIG. !0.5 (Top) Modeled angular divergence of sputter-deposited atoms for: A; no collimator, B; acollimator of aspect ratio I:1, and C; a collimator of aspect ratio 2:1. (Bottom) The calculated net deposition rate compared to experimental values as a function of collimator aspect ratio 110.14].would have the same angular and energy distributions that they left thecathode with. Increased pressure would result in a wider angular distribution as well as a reduced average energy.Some models tie the transport part of the model to the film depositionpart [ 10.16].
The transport model predicts a certain angular and energy distribution of atoms that arrive at the sample location, and this informationis fed directly into the film deposition part of the model.PROCESS MODELING FOR MAGNETRON DEPOSITION361S,:,u.e;Tar.e,:....AAi AI/1i'13k~,1~Ik .... I1 ,,'1. /t ~.. K1.~.~,......--~....~~e,oht~,ch,ate_/Al..........i!~~,ii~........
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
