|
With immersion lithography at numerical aperture (NA) at or exceeding 1.2, the process window optimization of 42 nm
line/space (L/S) patterning is a difficult challenge as the k1 factor approaches 0.26, very close to the theoretical limit.
Advanced immersion resists used to print these patterns are extremely thin and do not enable use of a thick bottom antireflective
coating (BARC) due to etch selectivity limitations. Conventional BARC optimization based on reflectivity
simulation alone does not provide an accurate process window as the resist profile is not fully correlated with substrate
reflectivity. Reference experimental tests show that, by varying BARC thickness, we can obtain straighter profiles with
1.9% second-minimum reflectivity as compared to 0.3% first-minimum reflectivity. The Brewer Science, Inc.,
OptiStackTM simulation tool was used to simulate the optimal conditions based on a full diffraction model where the
design criterion is the optical phase shift of the reflection. Two metrics comprise the simulation output: the foot exposure
(FE) that characterizes the phase shift, and the effective reflectivity (ER) that is calculated from standing wave
amplitude. The objective is to obtain the minimum ER at the target FE. Two experiments were conducted in order to
validate this concept. In both set of tests, the films were characterized experimentally by analyzing the process window,
resist profile, and line width roughness, and by simulating the FE and ER. In the first experiment a reference BARC,
Brewer Science ARC®29A coating, and an advanced variable-k BARC, Brewer Science ARC®121 coating of the
ARC®100 coating series, selected from simulation are compared. Even though the reference materials did not show a
large variation of FE and ER in the wide thickness range studied, optical simulations explained the tapered profiles and
the smaller process windows. The variable-k BARC presented a larger FE range that included both the target FE value
and locally minimized ER. Process window analysis shows that the optimal process was not correlated to minimum
reflectivity but to the metric previously described, minimum ER at target FE. The second experiment, designed to better
de-correlate FE and ER through adapted k and thickness, using again an ARC®100 series BARC, confirmed the strong
effect of FE value at a given ER on resist profiles and process window.
|
