Surface Damage by Physical Cleans during Semiconductors Manufacturing

Keywords: physical damage, surface, cracks, defect revelation Introduction More than one third of process operations in integrated circuits manufacturing rely on Wet cleans. They’re used to remove contaminants such as particles which are a dies’ yield killer. This is achieved either by a chemical action lifting the particles at the cost of a significant materials loss and device performances. This can be also done with some physical cleans (acoustic wave [1], cryogeny [2], high velocity spray [3]) or a combination of both. Whereas features damage by physical cleans has been widely studied [4], surface damage is poorly documented [5] due to lack of evidence. Experimental and characterizations In this work, FDSOI wafers (12 nm Si on 20nm buried oxide) are used. Physical cleans consist in either bi fluidic (aqueous liquid and nitrogen) high velocity sprays, or nitrogen cryogeny methods. Defects inspection on bare wafers is performed on a SP3 Surfscan tool from KLA-Tencor at detection such as 53 nm. Results and discussion Several physical cleans are considered in this paper. First and foremost, high velocity sprays are considered. Their activation and deactivation are the most critical steps due to high risk of erratic behavior (figure 1). Therefore, only well stabilized sprays are hereby discussed. Surface damage on FDSOI surfaces has already been reported [6]. Nano cracks on hard surface can’t be directly detected. A revelation method is used to enlarge the defects, by etching the box (buried oxide) underneath the 12 nm Si layer, thanks to concentrated HF. The wet chemistry can’t penetrate the nano cracks (trapped air [7] and steric hindering [8]) until the path is slightly enlarged. Hence an oxidation step is performed either by ozonated water or plasma to oxidize the silicon along the nano crack and generate after HF step, a 5 nm hole for HF to etch µm wide cavities underneath the damaged silicon (figure 2). Other solid films have been characterized the same way, by changing the enlarging and revealing chemistry. Secondly, a direct characterization method has been developed to quicker compare physical cleaning methods. Indeed, the first method has one drawback, only cracks deep enough are revealed. Soft films are hence used to directly reveal surface damage. A relevant choice of the polymer has enabled to evidence various kind of defects [9] with the high velocity spray (figure 3) and also some alumina particles damaging the resist surface during a nitrogen cryogenic clean (figure 4). These latter weren’t detected on hard films. The various defects maps and tests are summarized in table 1. Eventually, nano indentation and nano scratch tests have been performed on the photo resist films to reproduce the behavior of high energy droplets. And consequences from surface damage will be discussed. Conclusions Physical cleans are widely used in semiconductors manufacturing and are mandatory to achieve a high yield. Whereas damage to features is well documented and easily detected, surface damage is not. Two methods are given to evidence these phenomena: either directly on soft matter or by indirect revelation on hard surfaces. Some new particles cleaning solutions like polymer coat and peel methods are emerging [10] [11]. They combine both an excellent cleaning performance without any damage. References [1] S. Brems, ECS J. Solid State Sci. Technol, Vol. 3 (1) N3010 (2014) [2] J. Lauerhaas, IEEE/SEMI conference and workshop on advanced semiconductor manufacturing , p 11-16 (2005) [3] M. Sato, ECS, 41 (5), p 75-82 (2011) [4] C. De Marco, ECS Trans., Vol. 11 (2) (2007) [5] K. Suzuki, Japanese Journal of Applied Physics, 50 (2011) [6] US2020/0083065 A1 patent [7] C. Virgilio, International Journal of Mechanical and Mechatronics Engineering, 10 (3) (2016) [8] P. Garnier, Solid State Phenomena, Vol. 282, p 141-146 (2018) [9] Y. cho, Microelectronic Engineering, 234 (2020) [10] A. Lallart, Surface Preparation and Cleaning Conference (2018) [11] T. Mercadier, Solid State Phenomena, 346, p 268-274 (2023) Figure 1