JOURNAL OF MICROELECTROMECHANICAL SYSTEMS, VOL. 15, NO. 3, JUNE 2006 541 Layout Controlled One-Step Dry Etch and Release of MEMS Using Deep RIE on SOI Wafer Liu Haobing and Franck Chollet Abstract—Deep reactive ion etching (DRIE) of silicon on insu- lator (SOI) wafer has become a popular method to build microelec- tromechanical systems (MEMS) because it is versatile and simple. However when the devices using this technology become large in size or have compliant beams, the stiction occurring during the HF wet release is a serious problem. We have observed that some structure patterns could be wet released more easily than others. In this paper, we discuss the relationship between structure patterns and their stiction property, and describe the notching effect, which is found to be the mechanism behind this dependence. We finally provide simple mask layout design rules to utilize this effect to our advantage. These rules allow etching the structure and releasing it with the same DRIE step, without any wet process. Alternatively, this method will completely remove the stiction appearing during wet release or other further wet processes. We show the application of these rules on the fabrication of a large moving stage. [1636] Index Terms—Deep reactive ion etching (DRIE), microelec- tromechanical systems (MEMS), notching, release, silicon on insulator (SOI), stiction. I. INTRODUCTION D EEP REACTIVE ion etching (DRIE) of silicon on insu- lator (SOI) technology is usually considered one of the easiest way to build high aspect ratio MEMS devices. Its use- fulness has been clearly established by commercial products and MEMS researchers which have used it [1]–[4]. With only one mask, high aspect ratio structure can be formed by DRIE. After wet etching of the oxide layer in an HF solution, the device structures are released and can move freely. This process works well when the structure is simple, small, and very stiff in the ver- tical direction. However, MEMS devices built on SOI wafer are becoming more and more complicated and large (e.g., several millimeters). For such devices, stiction problem during release is still a concern. Stiction happens when the wafer dries and microstructures are pulled to the substrate by surface tension or capillary forces at the receding rinse liquid/air interface [5]. Then a combination of forces appearing between the structure and substrate, e.g., van der Waals forces and hydrogen bonding, keeps them firmly bonded to each other. The adhesion force is so strong that a force large enough to detach them usually de- stroys the microstructure [6]. Many efforts have been used to reduce the stiction and in- crease the release yield [7]. Mechanical approaches include cre- ating bumps on the underside of the structure layer [8], tem- porarily stiffening the microstructure with polysilicon links [9], Manuscript received June 25, 2005. Subject Editor N. de Rooij. The authors are with the MicroMachines Centre, School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore 639798, Singapore. Digital Object Identifier 10.1109/JMEMS.2006.876660 Fig. 1. Patterns for release test on 50 thick SOI wafer. or building polymer supporting columns that can be dry etched after the wet release [10]. Other physical approaches include avoiding wet process by using HF vapor [11], reducing surface tension by replacing water with methanol before drying, using hydrophobic coating layer [12], avoiding liquid drying process by freeze-drying or supercritical drying [13], or using charge controlled overetching [14]. Although some of the existing approaches have reported high microstructure release yield, they often need complicated process or facility (e.g., supercritical drying), and the easier methods do not achieve good results [7]. In this paper, we will introduce a method that is simple but effective for releasing microstructures built on SOI wafers. II. MESH PATTERNS AND THEIR RELEASE PROPERTIES During the fabrication of actuators with the DRIE process, we observed that some mesh patterns seldom experience stic- tion while others almost always did. We investigated this phe- nomenon thoroughly by using mesh with different line width, pattern size and pattern shape at the end of an array of cantilever of different length. Fig. 1 shows the nine mesh patterns used for the test. Pattern 1 has a square unit with 12 line width and 50 gap. Pattern 2 changes the beam width to 6 , and pat- tern 3 changes the gap to 25 . Patterns 4 to pattern 7 adopt different shapes. Pattern 8 adds some so called anti-stiction tips on one side, and pattern 9 reduces the area to half of pattern 1. We then noted the maximum beam length that get released after a standard etch and wet release process (Fig. 2). The results are shown in Table I. The result shows a clear contrast. Patterns 1, 2, 8, and 9 easily stick to the substrate, as beams as short as 200 can not surely 1057-7157/$20.00 © 2006 IEEE