Characterization of Crystal Qualily by Crystal Originated Particle Delineation and the Impact on the Silicon Wafer Surface D. Gräf, M. Suhren, U. Lambert, R. Schmolke, A. Ehiert, W. von Ammon, and P. Wagner Wacker Silt ronic AG, D-84479 Burghausen, Germany ABSTRACT Characterization of Si wafers by delineation of crystal originated particles (COP) provides insight into size and radi- al distribution of crystal related defects. A good correlation of the COP densities with gate oxide integrity and flow pat- tern defect densities is observed. The density and size distribution of COP in Czochralski Si ingots can be modified by the pulling rate and the cooling conditions of the crystal and is further influenced by high doping concentrations. The COP densities are comparable on wafers with (100) Si and (111) Si orientation as well as on p- and n-type wafers with mod- erate doping level. No COP are found on float zone (FZ) and on epitaxially grown wafers. Crystal defects are also delin- eated by chemomechanical polishing and can be detected on the wafer surface as light point defects (LPD). LPD densi- ties, however, do not necessarily correlate with the corresponding COP densities after SC1 treatment and do not reflect the quality of the crystals because polishing delineates only part of the larger crystal defects to a size which is above the detection limits of commercially available scanning surface inspection systems. High temperature annealing results in reduction of defect sizes and partial dissolution of COP. Investigations of FZ and oxygen doped float zone indicate that oxygen is participating in the formation of COP. Infroduction Characterization of crystal originated particles (COP) of polished silicon wafers by SC1 treatment is a powerful and valuable tool for monitoring Si ingots with respect to crystal defects.'3 The technique allows for fast mapping of the entire wafer surface with an automated scanning sur- face inspection system (SSTS), high lateral resolution, which provides, for example, precise localization of differ- ent defect regimes on a wafer, and information on the size distribution of crystal defects as derived from scattering cross sections.4 Several other techniques are available to characterize crystal related defects, e.g., gate oxide integrity (GOT), flow pattern defect (FPD), and infrared light scattering tomog- raphy (IR-LST). In most cases, a rather good qualitative agreement is obtained between all of these techniques. COP are detected in the vacancy rich regime of Czochralski (CZ) silicon crystals where also GOT, FPD, and IR-LST tmdefect densities are observed. These crystal related defects can be modified by appropriate pulling conditions6'° and by subsequent heat-treatments." Results are shown demonstrating the influence of crystal pulling and the impact of bulk parameters such as doping concentration and crystal orientation. The COP size distribution of the as-grown crystals significantly influences the efficiency of annealing treatments for defect dissolution. A discrepancy in correlating COP with FPD and GOT was reported for float zone (FZ) crystals. No COP were detected on FZ grown without nitrogen doping although similar defect regimes are involved as in CZ and high den- sities of FPD and GOT defects have been observed.'0 It is shown that COP can be delineated also on FZ wafers but only after prolonged treatment with SCT. A pronounced difference between FZ and CZ silicon crystals is the oxygen concentration, which is significant- ly higher in CZ crystals, where it is the prevalent grown- in impurity. Although it is generally accepted that COP are octahedral voids formed by agglomeration of vacancies incorporated into the silicon crystal during growth,'4 it is highly probable that oxygen also plays an important role in the formation of grown-in defect clusters.'5 New inves- tigations on COP in oxygen doped float zone (OFZ) wafers are presented which give experimental evidence for the impact of oxygen on the formation of grown-in defects and provide further insight into the origin of COP. Experimental A variety of material was used for the present investiga- tions with respect to doping element (B, P, Sb), doping concentration, crystal orientation (TOO and 111), oxygen concentration and wafer diameter (100 to 200 mm wafers). COP were delineated by immersion of the wafers in SC1 (NH4OH:HO2:H2O = 1:1:5) solution at 85°C for 1 to 24 h. Mapping of the wafers with a SSIS provides COP densities with a high sensitivity as well as size and radial distribu- tions of the COP. A Tencor Surfscan 6200 and a Censor ANS-100 were used for detection of COP. Other techniques for monitoring of crystal defects were applied for comparison. Electrical performance of the wafers was characterized by a GOT test (25 nm gate oxides on capacitors with 8 mm2 contact area; accelerated Qbd with yield criterion 5 1O' C/cm2). FPD were delineated by Secco etching and removal of 30 p.m of Si. Oxidation induced stacking fault (OSF) tests were performed by wet oxidation for 2 h at 1100°C and subsequent oxide stripping by HF followed by Secco etching. Results and Discussion Delineating COP by SC1 treatment—The Si removal during immersion of wafers in SC1 was almost indepen- dent of the wafer dopant concentration (Fig. 1). The label- ing p and p denote low and high boron concentration in the range of T0'5/cm3 and above 10'8/cm3, respectively, n doping with phosphorous (T0'5/cm') and n doping with antimony (5 . 10'8/cm2). An average Si removal of about 40 to 50 nm/h was obtained for the SC1 conditions given -c E C Co Co > 0 E U) Fig. 1. Si removal rate for immersion in SC1 (NH4OH:H202:H20 = 1:1:5) at 85°C. Removal rates on (111) Si are slightly lower as com- pared to (100) Si. No dependence on doping is observed. CZ p- CZ p+ CZ n- CZ n+ FZ n- FZ p- J. Electrochem. Soc., Vol. 145, No. 1, January 1998 The Electrochemical Society, Inc. 275