电感耦合等离子体中原子层蚀刻对晶圆级均匀性的影响

Atomic layer etching (ALE) typically divides the etching process into two self-limited  reactions. One reaction passivates a single layer of material while the second preferentially  removes the passivated layer. As such, under ideal conditions the wafer scale uniformity  of ALE should be independent of the uniformity of the reactant fluxes onto the wafers,  provided all surface reactions are saturated. The passivation and etch steps should individually  asymptotically saturate after a characteristic fluence of reactants has been delivered to each  site. In this paper, results from a computational investigation are discussed regarding the  uniformity of ALE of Si in Cl2 containing inductively coupled plasmas when the reactant  fluxes are both non-uniform and non-ideal. In the parameter space investigated for inductively  coupled plasmas, the local etch rate for continuous processing was proportional to the ion  flux. When operated with saturated conditions (that is, both ALE steps are allowed to selfterminate), the ALE process is less sensitive to non-uniformities in the incoming ion flux  than continuous etching. Operating ALE in a sub-saturation regime resulted in less uniform  etching. It was also found that ALE processing with saturated steps requires a larger total ion  fluence than continuous etching to achieve the same etch depth. This condition may result  in increased resist erosion and/or damage to stopping layers using ALE. While these results  demonstrate that ALE provides increased etch depth uniformity, they do not show an improved  critical dimension uniformity in all cases. These possible limitations to ALE processing, as  well as increased processing time, will be part of the process optimization that includes the  benefits of atomic resolution and improved uniformity.

Plasma based ALE is a self-limiting two-step process  capable of removing single atomic layers in each cycle.  The first step of the cycle passivates the surface in a manner  that naturally stops when the top surface layer is fully passivated. This step is ideally performed with an ion-free flux  of neutral radicals. In the second step, the passivated layer is  selectively removed by, ideally, a radical free flux of energycontrolled ions that chemically sputters the top passivated  layer but is not energetic enough to sputter the underlying  unpassivated atoms. When the fluence of reactants is large  enough to fully saturate both of the self-limited half-reactions,  the etch depth per cycle (EPC) should be constant. In this  saturated regime, the etch rate should also be independent of  small non-uniformities in the reactant fluxes. The adoption of  ALE techniques may then offer a way to obtain atomic scale  uniformity over large areas without the need for having correspondingly uniform reactant fluxes.

Self-limiting behavior is necessary to improve wafer scale  uniformity in the presence of non-uniform reactant fluxes,  but it is not clear what the relationship is between the uniformity enabled by ALE and process saturation. Most ALE  processes are not perfectly self-limited , and it is difficult  to estimate the effect of non-ideal etching reactions on process  uniformity. The following discussion focuses on the consequences of saturation during the ion bombardment phase of  ALE on wafer-scale uniformity, however the same trends hold  true for the passivation phase as well.

Fig1

ALE using an ion bombardment time of 3 s produced significantly more uniform etch rates than continuous etching,  with only a small positive correlation between etch rate and  ion flux, as shown in figure  2(b). For antenna A1, the etch  rate—measured here in monolayers of computational cells  (3 Å) per cycle—increased by 17% from the center of the wafer  to the edge, significantly less than the increase in Γi (133%).  This improvement in etch uniformity compared to Γi indicates  a fully saturated self-limited ALE reaction. Non-ideal continuous etching mechanisms result in some dependence of etch  rate on Γi, and produce the remaining non-uniformity in etch  rate.

While the ALE process significantly improved the etch rate  uniformity for antenna A1 and A2 when compared to continuous etching, the results for antenna A3 are less improved.  This result indicates that the ion fluence at the center of the  wafer, which is similar for each antenna, is only just large  enough to saturate the ion bombardment reaction. For the  A1 and A2 antennas, the ion flux and fluence increase with  radius. Given that both the passivation and ion phases are both  fully saturated, the increasing fluence at larger radius does not  produce a significantly higher etch rate. However, in the case  of antenna A3, the ion fluence drops below saturation as the  radius increases, causing the etch rate to depend more strongly  on ion flux which decreases with radius.

The ICP power used for antenna A3 was chosen to result in  similar ion flux at the center of the wafer as antennas A1 and  A2. This choice of ion flux resulted in sub-saturation behavior  at large radius where ion fluxes are the lowest when processed  using the same ion bombardment time as A1 and A2. If the  ion fluence was increased for case A3 so that the entire wafer  was within the saturation regime, either by increasing the ICP  power (resulting in a larger ion flux) or by increasing Ti, a  similar increase in uniformity could be obtained for A3.