晶片清洗工艺评估

ABSTRACT

This paper presents an experimental study of the aging of contaminated testwafers prepared forparticle removal (cleaning) process evaluation insemiconductor wafer processing. Two wafer preparation techniques werecompared: a conventional wet technique in which bare silicon wafers weredipped in a particle-laden solution and then dried, and a dry technique inwhich particles were deposited from a dry particle-aerosol stream. The extentof wafer aging was quantified through cleaning tests. Wafers contaminatedwith silicon nitride and tungsten particles were cleaned on different daysfollowing particle deposition, and the variation in the cleaning efficiency withwafer storage time was monitored. The tests showed that the dry-depositedwafers aged very little compared to the wet-dipped wafers, particularly fortungsten particles. The low-aging characteristic of the dry-deposition processhas positive implications for the repeatability and consistency of testsinvolving new particle materials of interest such as copper and low-kdielectrics.

INTRODUCTION

Particle removal (cleaning) is a critical aspect of semiconductor wafer surface preparationCleaning processes involving various mechanisms have been developed to date (1-4]. Theremoval of particles as small as 64 and 41 nm has recently been demonstrated (3-67. Theefficiency of a cleaning process is determined through tests involving contaminated wafersprepared using a variety of techniques. The most common methods include immersion in asolution containing suspended particles (7] and exposure to a wet particle-laden aerosol (8. Amore recent method involves exposure of wafers to a dry particle-aerosol stream (97. Other noveltechniques have been developed, but are not yet well established. These include spin coating withparticle-laden solutions 10 , exposure to saturated vapor of the contaminant 11 , and exposureto reactive chemical solutions (12].

In the immersion technique (Figure la), the wafer is dipped in a deionized-water bath containinga dispersion ofthe test particles, and then spin-dried. In the wet-aerosol technique (Figure 1b), thewafer is exposed to a particle-laden deionized-water spray, then spin-dried. The two techniqueshave been found to yield fairly equivalent results in terms of particle-wafer adhesioncharacteristics, with the dipping method showing somewhat better wafer-to-wafer consistency inthe particle count and distribution than the aerosol technique (87. In both cases, wafer aging isobserved, i.e., particle-wafer adhesion increases over time. Overall, the techniques are simple andsatisfactory for most cleaning tests aimed at process characterization.

The third technique involving dry-aerosol-based deposition has some advantages over the wettechniques in terms of repeatability and control over the particle deposition, making it particularluseful for the fine-tuning and optimization of cleaning processes. The control is achieved througlthe use of speclal instrumentation for the deposition 9 . A schematic of the instrumentation isshown in Figure 2. A particle-laden solution is atomized and the droplet spray is dried to yield adry stream of particles. These particles are then electrostatically deposited onto the wafer surfaceThis method was developed originally for the preparation of reference wafers for the calibrationof wafer scanners. but has recently been successfully adopted for the preparation of test wafersfor cleaning studies (3-57. Although much has been learnt about the deposition process itself, littleis known about the particle-substrate adhesion characteristics of the test wafers prepared usingthis dry technique. Of particular interest is the aspect of wafer aging, since it has direct bearing onthe interpretation of cleaning test results.

The focus of this study, therefore, was the aging characteristics of dry-deposited wafers. While itis conjectured that wafer aging is not as severe as in the case of wet-deposited wafers, because othe relative dryness of the deposited particles, the effect has so far not been quantified. In thisstudy, the aging characteristics of wafers prepared using the dry-aerosol and wet-dippingtechniques are compared in a controlled experiment. The aging is measured indirectly throughcleaning experiments in which the variation in efficiency of the cleaning process is recorded as afunction of wafer storage time.

EXPERIMENTAL

2.1. Test Wafer Preparation

Since both non-metallic and metallic contaminations are of interest in wafer cleaning, siliconnitride (Si;N ) and tungsten (W) were chosen as the test particles. Test wafers were prepared bthe wet-dipping and dry-deposition techniques for both materials. Although the depositedparticles typically have a wide size range (from a few nm to several um), this study focuses onlon particles of size > 0.15 um, in keeping with a similar earlier study 187 comparing the wetdipping and wet-aerosol techniques. The conclusions drawn from this study are expected to bevalid for smaller (< 0.15 um) particles as well.