Friday, December 8, 2006

Resist Trimming

The need for resist trimming arises from the gap between the line widths advanced lithography can print today and the desired gate length of the transistor. For the 90 nm technology node, the printed line width is typically somewhere between 90 and 100 nm and the final gate length 50 to 60 nm. Hence, a line width reduction has to be achieved at some point in the process flow, typically during the first step of the gate patterning etch (slide 1). The requirements for this so called resist trim step are: no deterioration of the critical dimension (CD) uniformity across wafer, adjustable dense / iso CD bias, line integrity (no resist bending or line clipping), adjustable aspect ratio of post trim resist shape, linear dependence of the CD reduction on trim time (i.e. linear trim curve).

Oxygen is the main etch gas in many resist trim processes because oxygen based plasmas etch polymers isotropically. Halogens such are HBr, HCl, Cl2 or fluorocarbon gaeses are added to provide control over the ratio of lateral vs. vertical etch rate. These gas additives inluence the trim rates for the dense and isolated lines. It iwas found that the trim rate is faster in HBr/O2 chemistry than in Cl2/O2 chemistry. Larger difference in trim rate between dense and isolated lines are found for the HBr/O2 chemistry (slide 2).

With HBr and HCl, the trim rate is faster in isolated than in dense resist patterns. Cl2/O2 chemistries result in a trim process which is faster in dense resist lines than in isolated resist lines (slide 3).

The effect of bromine and chlorine addition to an oxygen based trim process wass studies with in situ XPS (E. Pargon, O. Joubert, T. Chevolleau, G. Cunge, Songlin Xu, Thorsten Lill; JVST B, 23 (2005) 103). In HBr/O2 plasmas, very little Br concentrations are found on all resist surfaces exposed to the plasma. O only is involved in the resist transformation. Br is not reactive with respect to carbon. Thin reactive layers are typical of a very chemical type of etch. In Cl2/O2 plasmas, large chlorine concentrations are found on top and sidewall of the resist patterns indicating a competitive absorption between O and Cl. The increase in reactive layer thickness on the resist sidewalls with Cl2/O2 indicates that it is formed from resist etch products redeposition (slide 4).

For O2/Cl2 trim processes, the composition of the reactive layers changes when the O2 concentration in the gas phase is increased. The Cl concentration and thickness of the the top reactive layer decrease. The Cl concentration and thickness of the reactive layer on the resist sidewall decrease and the O/Cl ratio increases. A direct correlation between trim rate, O/Cl ratio and thickness of the reactive layer on the resist sidewalls exists (slide 5).

For O2/HBr trim processes, very little difference in reactive layer composition and thickness as a function of O2 concentrations was found. A slow down in trim rate at low O2 concentration (10% in HBr/O2) may be attributed to the presence of Br in the sidewall layer. At higher O2 concentration (60%), no bromine is present on the resist sidewalls, the trim rate increases by more than 50% (slide 6).

The pressure dependence of the Cl2/O2 trim process is shown in slides 7 and 8. An increase in pressure in Cl2/O2 generates a slow down in trim rate in dense as well as isolated resist patterns. The composition of top resist patterns is not affected by the change in pressure. For the sidewalls of the resist patterns, an increase in chlorine concentration, decrease in oxygen concentration and increase in reactive layer thickness is observed. This is consistent with the decrease in trim rate as a function of pressure.

Similarly to the Cl2/O2 chemistry, an increase in pressure of the HBr/O2 gas mixture generates a decrease in trim rate in dense as well as isolated resist patterns. XPS shows no difference in reactive layer formation in dense resist patterns as a function of pressure of the HBr/O2 gas mixture (slide 9).

In Cl2/O2 plasmas, increasing bias power has the same impact than increasing pressure: the trim rate decreases in dense as well as isolated lines (slide 10).

XPS studies of the active layer showed the following trends for increasing bias power in Cl2/O2 plasmas: 1. The top resist patterns not affected by the increase in bias power, their composition is only driven by O2/Cl2 ratio, 2. The idewalls of the resist patterns exhibit an increase in chlorine concentration, a decrease in O concentration and an increase in reactive layer thickness consistent with the decrease in trim rate as a function of bias power, and 3. the deposition of carbon etch products increases with bias power (slide 11).

Similarly to Cl2/O2, increasing bias power in HBr/O2 leads to a decrease in trim rate for both isolated and dense lines. An increasing bias power generates a decrease in trim rate difference between isolated and dense lines whereas there is almost no effect in Cl2/O2 (slide 12).

In summary, resist trimming is a very powerful and widely used method to achieve gate lengths which are beyond the resolution limits of photolithography. The trim process is chemical in nature and very dependend on the choice of gases. Slide 13 shows some of the limits of the resist trim process. Firstly, resist consumption and internal stress leads to profile deformation (resist bending). Secondly, initial resist roughness (line edge roughness) becomes a problem after ultimate trim.

More Plasma Etch Applications ...

2 comments:

Anonymous said...

Do you have any data for resist trimming with fluorinated chemistries ?

Clarycon said...

No XPS data, only empirical observations: http://www.clarycon.com/cd_control_pic.html
(slides 6 and 7).