Plasma Technology for Advanced Devices
ARDE in SAC Etching
ARDE (Aspect Ratio Dependent Etching) is an effect where features with a high aspect ration (depth/width) have a higher etch rate then those with a small aspect ratio. Under certain conditions this effect can be reversed and is then called reverse ARDE. RIElag is a synonym for ARDE.
The following experiments were conducted in a magnetically enhanced reactive ion etcher (MERIE). Slide 1 describes a method for measuring ARDE where a series of contact holes with variable CD’s is printed and etched simultaneously. The etch depth for each hole size is measured by cross section SEM. This method allows to determine the average etch rate as a function of CD and etch time, final etch depth and final aspect ration. It requires that features with variable CD's are part of the photomask.
In case these special litho features are not available, an etch time scan can be used. Several wafers with the same pattern are etched for a series of etch time and the etch depth for each hole size is measured. Gives the average etch rate as a function of etch time, final etch depth and final aspect ratio (slide 2).
In order to truely measure the etch rate as a function of aspect ration, the change in etch depth has to be determined for an infinitely small etch time. In this method, several wafers with the same pattern are etched for a series of etch times. The etch depth of each hole size is measured. The same wafers are etched for a short time and the change in depth is measured again (slide 2). Using this method, the etch rate can be expressed as a function of the aspect ratio as shown in slide 3, where BPSG was etched in a C4F8, CO, Ar chemistry. For an aspect ratio of around 1, the etch rate is 350 nm/min and is drops to 100 nm/min for an aspect ratio of 5.
Slide 4 shows the effect of ARDE for several different types of oxides: silane oxide, BSG, BPSG, and PSG. The results show that ARDE varies for different types of oxides. It appears to be more pronounced for slower etching oxides, in particular PSG. This can be explained with the mechanism that leads to ARDE which involves the balance between etching and polymer deposition. Slide 5 shows a more detailed comparison between TEOS and BPSG where the faster etching BPSG exhibits less ARDE than the slower etching TEOS. For a 320 nm hole and the recipe used in this particular experiment, TEOS runs into etch stop after 200 s.
Since ARDE is a result in a change of the balance wetween etch and deposition as the the aspect ration increases, one would expect that any external influence on the polymer precurser density in the plasma, their transport down into the feature and their deposition rate will impact ARDE. Since the mask resist is being etched in the process and the etch by-products of this process contribute to the plasma chemistry, one would expect that ARDE is also a function of the resist type which is being used. Slide 6 shows that for the same type of BPSG oxide and similar feature sizes, i-line resist exhibits severe ARDE much sooner than DUV resist. The surprizing element of this finding is that DUV resist etches faster than i-line resist. The explanation must be in the difference in the chemical composition of the reaction by-products.
If the formation of polymers plays a key role in ARDE during oxide etching, the effect of adding a gas which suppresses polymer formation should have a significant influence. Slide 7 shows that clearly. Addition of oxygen indeed suppresses the formation of polymers which is good for ARDE reduction, however it would also decrease the nitride corner selectivity which is very critical in SAC etching. A balance has to be found in the process. The addition of CO provides a better process window between ARDE reduction and nitride corner selectivity. Larger CO flows do suppress the formation of polymers and hence reduce ARDE and at the same time provide enough carbon to the plasma to maintain the oxide to nitride selectivity. Slide 8 shows that with the addition of 170 sccm CO, ARDE was absent for aspect ratios o up to 6 for contact holes in BPSG.
Slide 9 summarizes the key findings:
The ARDE effect for a C4F8 / CO / Ar chemistry for oxide etching in a MERIE type reactor has been found to depend on:
ARDE/RIElag models given by Sekine et al. and Joubert et al. which consider the strong deposition component of the given process and possibly charging of the deposits can provide a phenomenological explanation of the above findings.