Nanometric Particles Removal during Photoresist Stripping

Abstract

Microelectronics devices dimension have followed the Moore’s law in the last decades. In order to keep a high yield during integrated circuits production, a continuous improvement has been carried out in the particles cleaning area, switching from brushes and acoustic cleans to high velocity sprays. With the device nodes shrink, tinier and tinier particles are to be removed. Even though their adhesion and removal are well understood and documented for years [1], our industry still misses robust solutions to clean nanoparticles, keeping finest features integrity, and extremely low material consumption [2]. Well chosen resist stripping conditions enable an excellent particles removal without materials consumption nor damaging 14nm node polysilicon gates. The cleaning mechanism is investigated and key parameters involving the photoresist and wet stripping will be discussed. The aim of this work is to transfer this knowledge into an industrial optimized sequence. Previous studies deal with the fluids viscoelasticity role on the particles removal from a substrate [3]. It is thought non newtonian properties are involved during the resist removal process, with the fluid generated on the wafer through the well-known interface gel formation. This paper will present results obtained through the use of 300 mm wafers intentionally contaminated with 45nm silica particles, then coated by different polymers, stripped thanks to various methods. The particles counting before and after the complete resist removal step is assigned to a KLA Surfscan SP3 blue laser diffraction tool. The stripping conditions are a key parameter to enable an excellent defects removal. Indeed the PRE (Particles Removal Efficiency) varies from a few % to 70% when switching from SPM (Sulfuric acid / hydrogen Peroxide Mixture) conditions: too chemically active (radicals rich) with a stripping rate of few hundreds of nanometers per seconds on a single wafer platform to mild conditions, requiring few minutes to remove the resist on a batch spray tool, enabling the resist transition into a gel like phase (figure 1). Two samples have also been used as reference: first a resist removal by plasma ashing that doesn’t remove any particle, proving the particles removal doesn’t occur during the resist coating step and another one without resist, with the wet stripping to quantify the PRE brought by the SC1 (Standard Clean 1). Further tests are done with various solvents to substitute the SPM, but keeping the gel phase, required for the particles lift off. Not only results on blanket but also challenging patterned wafers will be presented, showing absence of fine feature damage and PRE data, with a fair comparison with high velocity spray methods. Eventually a basic mechanism will tentatively be proposed. Moreover, such results tend to prove an interest in continuing using resist stripping in wet instead of plasma ashing. Figure 1