Photoresist Adhesion Promotion
- Wirth, Justin C
Overview
The adhesion of spun photoresist on a silicon wafer, or other substrate, may not adequate without taking specific steps. Adhesion depends mainly of the surface state of the substrate when photoresist is spun, and ideal methods for promoting adhesion depend both on the substrate and on the spun film.
Surface States
Common substrates including Si, SiO2, and SiN naturally have a hydrophilic surface due to adsorbed water and exposed OH bonds in the native oxide. This can pose a problem for photoresist surface adhesion, which are typically hydrophobic compounds. HMDS (hexamethyldisilazane, [(CH3)3Si]2NH) can react with the surface oxide to form a hydrophobic surface, and is currently used at BNC as a spin-on adhesion promoter (as was common in the 1970s, per YES). However, for fine features, this poses problems for feature definition (HMDS - MicroChemicals). Spin coating results in a thick film that does not properly displace water or bond to the surface. Although spun on HMDS does indeed form a hydrophobic top layer, ammonia is released during the resist pre-exposure bake as a result of unreacted NH groups. The ammonia diffuses into the resist and crosslinks the resist near the substrate. Per Microchem "Substrate Cleaning Adhesion Promotion":
“In case of spin coating of HMDS, a too thick HMDS film forms on the surface. After resist coating during the softbake, this excess of HMDS releases ammonia which diffuses into the resist and crosslinks the resin near the substrate. As a consequence, through-development sometimes becomes impossible. For the same reason, NEVER apply HMDS in a spin coater. HMDS vapour will diffuse into all resist films subsequently coated, and partially crosslink the resist film during softbake, which lowers the development rate and can deteriorate the resist profile and attainable resolution.”
Therefore, proper application of undiluted HMDS requires the use of a bubbler (e.g. a YES Oven) to apply an ammonia-free monolayer of HMDS on the substrate surface. Such vapor deposition equipment is not currently available at BNC, motivating both the use of alternative adhesion promoters, and the phase out of undiluted HMDS use in photoresist spinners.
A better alternative than undiluted HMDS is a mixture of 20% HMDS and 80% PGMEA, which will give significantly better results for spin-on-deposition. This is available commercially from a number of vendors, examples include MCC Primer 80/20. Another option would be an ionic treatment, such as SurPass 3000/4000, depending on the substrate.
Surface Chemistry for Photoresist Adhesion
The non-polar, hydrophobic nature of photoresists leads to poor surface adhesion and wettability of photoresists spun on top of substrates with hydrophilic surfaces. For silicon substrates, the hydroxy (oxygen-hydrogen) group on the surface oxide layer (i.e. Si-OH), as well as the OH group of any adsorbed H2O, present such a hydrophilic surface. Adhesion promoters such as HMDS convert the hydrophilic Si-OH surface to a hydrophobic surface by substituting for the hydrogen on the OH group. The adhesion promoting molecules contain hydrophobic tails, and with proper application a uniform hydrophobic surface is formed. The treated surface then exhibits good surface adhesion with structures patterned in spun photoresist and good wettability when spinning.
For Si and SiN substrates (but not glass/quartz/SiO2), a potential alternative to adhesion promoters is the removal of the native oxide layer in an HF dip [5], followed by a dehydration bake hot enough to thermally crack any remaining OH bonds on the silicon surface [6]. This presents a hydrophobic hydrogen terminated (Si-H) silicon surface with excellent adhesion to spun resist [7]. In air, ambient airborne water (i.e. relative humidity) is responsible for the growth of the native oxide, so an immediate dehydration bake should be performed after HF dip and the subsequent H2O rinse. This should present an Si-H passivated surface for >3 hours [7], though spinning should be performed immediately after the surface has cooled after the dehydration bake.
As glass/quartz/SiO2 consists of bulk oxide, removal of “all” of the oxide is clearly not an option. Additionally, the surface of HF treated SiO2 (where the oxide has not been completely dissolved) will have very poor adhesion to resist [3] [8]. In these cases, HF should be avoided, and an adhesion promoter is essential for good adhesion. Additionally, if an adhesion promoter is to be used on a Si or SiN substrate, the OH surface bonds need to be preserved to get good bonding with the promoter. In those cases, HF and a thermal cracking step should likely both be avoided.
At the minimum, regardless of the substrate (possibly excluding flexible/polymer substrates) and whether a H or OH terminated substrate surface is present, a baking step to remove adsorbed water (i.e. a dehydration bake) should always be performed. Water desorbs at 100 °C, so a bake at 120 °C for ~3 minutes should be sufficient. A similar effect may result from an Acetone+IPA clean as long as a subsequent water rinse is omitted.