Measurement of the Spatial Evolution of the Deprotection Reaction Front with Nanometer Resolution using Neutron Reflectometry

The use of chemically amplified photoresists for the fabrication of sub-100 nm features will require spatial control with nanometer level resolution. To reach this goal, a detailed understanding of the complex reaction-diffusion mechanisms at these length scales is needed and will require high spatial resolution measurements. In particular, few experimental methods can directly measure the spatial evolution of the deprotection reaction front and correlate it with the developed structure. In this work, we demonstrate the complementary use of neutron (NR) and x-ray (XR) reflectometry to measure the reaction front profile with nanometer resolution. Using a bilayer geometry with a lower deuteriumsubstituted poly(tert-butoxycarboxystyrene) (d-PBOCSt) layer and an upper poly(hydroxystyrene) (PHOSt) layer loaded with a photoacid generator (PAG), we directly measure the spatial evolution of the reaction front. We show that the reaction front profile is broader than the initial interface after a post-exposure bake and the compositional profile changes upon development in an aqueous base solution. We also directly correlate the final developed structure with the reaction front profile. The spatial detail enabled by this general methodology can be used to differentiate between and evaluate quantitatively reaction-diffusion models. INTRODUCTION The successful development of chemically amplified photoresists for the fabrication of sub-100 nm features will require nanometer scale spatial control of the deprotection reaction front. There are many challenges toward meeting this goal because of the difficulties in measuring and understanding correlations between processing conditions and critical dimension control at these length scales. Detailed information about the physics and chemistry controlling the local deprotection reaction will be required. In addition, physical quantities from larger length scale, bulk measurements may not be applicable for the fabrication of smaller structures. For example, ultrathin photoresist imaging layers will be required because of the increased optical absorption of the photoresist at smaller imaging wavelengths. There are reported changes in the material properties of ultrathin imaging layers such as the glass transition temperature, Tg, or interfacial effects such as interdiffusion in a bilayer photoresist that can affect the fabrication process [1-3]. Direct measurement of the reaction front in ultrathin photoresist systems would be needed. It is well established that the spatial evolution of the deprotection reaction front is controlled by many factors including the distribution of the amount of photogenerated acid proportional to the local DUV dose and acid diffusion during the post-exposure bake (PEB). There have been significant efforts toward quantifying acid diffusion rates and determining their effect on the final structure. Experimental methods have included the use of ion conductivity measurements [4,5], correlation with analysis of the final structure [6], Fourier transform infrared spectroscopy (FTIR) measurements in varying sample geometries [7, 8], ellipsometry [9], thickness changes after reaction and development [1], and scanning electron microscopy on cleaved, chemically decorated structures [10]. These experimental methods have been limited by either spatial resolutions that are significantly larger than the nanometer length scales required at future technology nodes or the lack of direct information of the spatial evolution of the deprotection reaction front. Fundamental understanding and quantification of the reaction-diffusion process at these reduced dimensions will be needed to control, develop, and model the lithographic process. Many models have been developed to describe the reaction-diffusion process in chemically amplified photoresists. Differences between the models largely lie in the description of the acid diffusion process and the detail in which the reaction kinetics are incorporated. Predictions from reaction-diffusion models have been compared with the total extent of reaction with FTIR data before development in several systems [7, 8, 12-14]. Top loss deprotection experiments and comparison with different acid diffusion descriptions showed that Fickian diffusion cannot explain the experimental results [11]. Further, it was suggested that the diffusion constant is a function of the deprotection fraction. Experiments with an acid feeder layer coupled to a deprotected polymer layer show fast reaction rates at short times (less than typical PEB times) then saturation at longer times [7, 10]. A single acid diffusion constant also cannot explain the data. As a result, it was proposed that the mechanism for reaction front propagation includes an enhancement of the local acid mobility at short times due to the evolution of deprotection reaction products. It was postulated that the volatile reaction products increase the local free volume to allow increased acid mobility. Additionally, the acid is subsequently trapped in the deprotected polymer matrix leading to a net reduction in the available acid for deprotection. Numerical calculations based upon this model show that the reaction front propagates as a sharply defined interface into the protected polymer layer [14]. Other successful modeling approaches include the use of a simulation incorporating in detail the deprotection reaction kinetics as well as diffusion mechanisms. These calculations show that FTIR data may be fit with a physical picture where Fickian acid diffusion coefficients are significantly different within the protected and deprotected polymer environments [8]. There is difficulty in clearly distinguishing between reaction-diffusion models because of the lack of a direct measurement of the reaction front before development or without any further sample preparation. Progress has been made by created contrast for SEM measurements by chemically decorating deprotected polymers with the silylation of samples cleaved normal to a diffusion-couple sample [10]. These experiments enable estimates of the spatial extent of the deprotection front and suggest that the reaction front propagates as a sharply defined band. However, the details of the silylation chemistry and subsequent etch step are not known. Further, although there are general guidelines for the composition of the solubility switch and simulation models for development effects on LER, there has been no direct correlation between the reaction front, its width, shape, and composition, and the final developed image. As development steps may also be sources of line-edge roughness, this connection is critical. In this work, we present the first direct measurements of the reaction front with nanometer resolution both after PEB and after development using neutron reflectometry. Compositional and density depth profiles are measured from a well-defined bilayer structure with a lower layer consisting of the protected polymer and an upper layer consisting of the deprotected polymer loaded with the photoacid generator (PAG). The protected polymer is synthesized with deuteriumlabeled protection groups providing strong neutron scattering contrast. The neutron scattering contrast is then proportional to the local extent of reaction. X-ray reflectometry is used to measure the density profile of the bilayer structure enabling a direct comparison of the developed film structure with the spatial extent of the reaction front. EXPERIMENTAL Bilayers for the reflectivity measurements were prepared on <111> silicon substrates (3 mm thick, 75 mm diameter) primed with hexamethyldisilazane (HMDS). The lower layer consisted of poly(tert-butoxycarboxystyrene) (d-PBOCSt) (Mr,n = 21000, Mr,w/Mr,n = 2.1), the protected polymer, with a deuterium-substituted protection group. The upper layer consisted of poly(hydroxystyrene) (PHOSt) (Mr,n = 5260, Mr,w/Mr,n = 1.12), the deprotected polymer, loaded with a 5 % mass fraction of the photoacid generator (PAG) di(tert-butylphenyl) iodonium perfluorooctanesulfonate (PFOS). Bilayers were also prepared with PHOSt without PFOS as control experiments. The deuterium-substituted ptert-butoxycarboxystyrene (d-PBOCSt) was synthesized using the following procedure. First, 2-methyl-2-propanol-d10 was reacted with 1 equivalent potassium metal in dry tetrahydrofuran (THF) to form potassium tert-butoxide-d9. The reaction was considered to be complete when all of the potassium appeared to dissolve in the THF; this generally took about 48 h at reflux temperature. This salt was then used as starting material for the published literature procedure for di-tert-butyl dicarbonate [15]. 1.1 Equivalents of di-tert-butyl dicarbonate-d18 were reacted with PHOSt with 0.1 equivalents 4-(dimethylamino)pyridine in dry THF. Precipitation into hexanes afforded a 95 % yield of a white polymer that was 92 % protected as measured by thermal gravimetric analysis. Control measurements comparing the deprotection kinetics by FTIR of exposed blanket films of d-PBOCSt and similarly prepared hydrogenous PBOCSt loaded with PAG show little difference between the two polymers.