REFERENCE: Wijaranakula , W. , Matlock, J.H., and Mol- lenkopf, H., "Nucleation and Growth Kinetics of Bulk Microdefects in Heavily Doped Epitaxial Silicon Wafers", Semiconductor Fabrication: Technology and Metrology, ASTM STP 9 9 0 . Dinesh C. Gupta, editor, American Society for Testing and Materials, 1989.
ABSTRACT: Nucleation and growth mechanisms of the bulk microdefects in oxygen controlled silicon sub- strate wafers heavily doped with boron and antimony were studied following pre- and postepitaxial enhanced precipitation annealing at various tempera- tures. It was observed that "grown-in" microdefects in both dopant species are nucleated by a similar mechanism during the crystal growth cooling period.
Microdefect growth behavior appears, however, to be different, probably due to differing effects of the dopant species. It was also observed that the pre- and postepitaxial annealing as well as the epitaxial deposition processes have a strong impact on the growth of the bulk microdefects. Based upon this study, generic growth characteristics of the bulk microdefects of both P* and N"*" materials are establi- shed and a growth kinetics model is proposed.
KEYWORDS: silicon epitaxy, oxygen precipitation, internal gettering
Thermally induced defects which include oxidation- induced defects (OSFs and swirls) [1-3] and those induced by oxygen precipitation [4-5] have been studied for many y e a r s . Defects associated with oxygen precipitation seem
to draw the most attention because well controlled D r . Wijaranakula is a senior engineer/R&D Materials Characterization, Dr. Mollenkopf is the department manager and D r . Matlock is the Vice President of Technology at SEH America, Inc., 4111 SE 112st Avenue, Vancouver, WA 98662
oxygen precipitation with respect to the internal gettering (IG) process could lead to high device performa- nce and yield [6-9]. In epitaxial technology where heavily-doped substrate wafers are used, the differen- ce between precipitation behavior in boron doped P"*" and antimony doped N"*" silicon is observed [10]. Extensive studies on this subject have been conducted but no clear picture of the nucleation and growth mechanism has yet been reported. The precipitation model [11] based on com- plex formation between an impurity and a donor-type self- interstitial seems to be plausible. However, a similar argument can also be postulated in the case where a com- plex between boron and donor-type vacancy is involved.
Both acceptor and donor-type vacancies are known to exist at high temperature [12].
In the past, the determination of oxygen precipita- tion in lightly-doped silicon has been performed by measu- ring the change in the interstitial oxygen [Oi] con- centration using the Fourier Transform Infrared (FTIR) spectrophotometer prior to and after various thermal cycles [13-14]. It was found that reduction of [Oj] con- centration after a given thermal cycle could be directly correlated to the initial [Oi] concentration in the sili- con matrix [6, 15]. Plots of the reduction of [0-j] concen- tration as a function of initial [Oi] concentration have recently been used as generic curves [16] for various oxygen precipitate thermal cycles. An unknown factor involving this method is the influence of the oxygen depleted zone, termed the "denuded zone" (DNZ) on the FTIR reading [17]. This can eventually be prevented by removing the oxygen denuded layer prior to measurement [18].
In heavily-doped silicon, determination of the oxygen precipitation rate by IR measurement is not possible. This is because the [Oi] concentration cannot be determined directly by the FTIR due to free carrier absorption. SIMS, (Secondary Ion Mass Spectrometry), which determines the total concentration of oxygen, cannot distingush [Oi]
concentration from the oxygen concentration which is incorporated in the form of a solid solution. Therefore, any reduction in interstitial oxygen concentration cannot be quantitatively determined. At the present time, evalua- tion of the nucleation and growth of bulk defects asso- ciated with oxygen precipitation in heavily-doped silicon Is relied primarily upon observation of the defect forma- tion.
The demand for epitaxial silicon material for the application of IC devices such as high density CMOS [19- 20] and CCD [21] has increased significantly during the past several years. Devices fabricated on epitaxial sili- con material have shown superb characteristics (e.g.
latch-up prevention and the uniformity of resistivity) [22-23] when compared to devices fabricated on lightly-
doped silicon. Despite the abundant research on oxygen precipitation and internal gettering in epitaxial silicon material [ 2 4 - 2 9 ] , a fundamental understanding of the nucleation and growth of bulk defects and control of the precipitation process is considered to be lacking. Exten- sive studies were conducted on the following subjects and will be discussed in this paper:
a. Heterogeneous nucleation of the oxide micropreci- pitates.
b. Growth kinetics of bulk defects in epitaxial silicon wafers.
c. Impact of the epitaxial deposition process on oxygen precipitation.
EXPERIMENTAL PROCEDURES. RESULTS, AND DISCUSSIONS
Heterogeneous Nucleation of the Oxide Microprecipitates The starting materials were P"*" and N + d O O ) substrate wafers, 100 mm in diameter and heavily-doped with boron and antimony (0.01 and 0.025 o h m - c m ) . In order to prevent a possible fluctuation in the oxygen precipitation rate due to the thermal history effect [30-31] and the crystal growth conditions [ 3 2 - 3 3 ] , substrate wafers were prepared only from a very short section of fully grown crystals.
The length of the sections were between 15 and 30 millime- ters. The oxygen content in each section was determined by SIMS because the FTIR cannot be used when the resistivity is lower than 0.1 ohm-cm. Only sections that contained oxygen in the range between 1.45xl0l8 and 1.55xl0l8 atoms/cm3 (ASTM F121-79) were used.
Isochronal pre-annealing of the substrate wafers was carried out in nitrogen ambient at specific temperatures ranging between 500 and 900°C. Mixed gases (nitrogen + 5%
oxygen in volume) were used to prevent "nitrogen" pitting when pre-annealing was performed at temperatures above 750°C. This pre-annealing was necessary to induce hete- rogeneous nucleation of the "grown-in" microprecipitates.
Substrate wafers with no pre-annealing were also set aside and used as the control samples. P"*" substrate wafers were pre-annealed for 3 hours, whereas N"*" substrate wafers were pre-annealed for 24 hours so that sufficient bulk defects could be obtained for this study. After pre-annealing, wafers were processed through the polishing and epitaxial deposition steps. During epitaxial deposition, epitaxial layers of approximately 6 and 13.5 microns thick were deposited. Epitaxial deposition was performed at 1150°C using trichlorosi1ane as the gas source. Boron and phos- phorus were used as the doping species. After epitaxial deposition, the wafers were annealed at 1050°C in dry oxygen ambient for 16 hours. This annealing step would represent the well drive-in step of the CMOS IC fabrica-
tion process. Bulk defects were examined on the (110) cleaved surface after Wright etch [34]. P/P+ samples were etched for 1 1/4 minutes, while 2 1/2 minutes were re- quired to etch the N/N"*" samples. Extended etching of P/P + samples results in overetching and the formation of arti- facts.
Figs. 1 and 2 show a series of photomicrographs of the cross sections of P/P+ and N/N"*" samples. It is observed that the bulk defect size and density are strongly influenced by the pre-annealing temperature. The isolated bulk defects in a heavily-precipitated area are
• •' 4 , '
-> f •.-.
(a)
\wm^m
SUBSTRATE--'
i*;'.-..'>.f
< ? ằ N ằ " ằ .
^ * V - \ < : . ; ' <
< ^ - ^ 4 H K > * •
'< ằ ằ •
• v ^ /
' - : y
• r:f:
..** ; '
• •'• .
rô
# / t
*
•*•
fr,
(c)
FIG. 1 -- Cross section optical photomicrographs for P/P+(100) wafers preannealed at a) 525, b) 650, c) 750, and d) SOO'C.