Monitoring metal contamination of silicon by ...

Monitoring metal contamination of silicon by multiwavelength room temperature photoluminescence spectroscopy Shiu-Ko Jang Jian, Chih-Cherng Jeng, and Woo Sik Yoo Citation: AIP Advances 2, 042164 (2012); doi: 10.1063/1.4769746 View online: http://dx.doi.org/10.1063/1.4769746 View Table of Contents: http://aipadvances.aip.org/resource/1/AAIDBI/v2/i4 Published by the AIP Publishing LLC.

Additional information on AIP Advances Journal Homepage: http://aipadvances.aip.org Journal Information: http://aipadvances.aip.org/about/journal Top downloads: http://aipadvances.aip.org/features/most_downloaded Information for Authors: http://aipadvances.aip.org/authors

Downloaded 24 Sep 2013 to 177.101.33.100. All article content, except where otherwise noted, is licensed under a Creative Commons Attribution 3.0 Unported license. See: http://creativecommons.org/licenses/by/3.0/

AIP ADVANCES 2, 042164 (2012)

Monitoring metal contamination of silicon by multiwavelength room temperature photoluminescence spectroscopy Shiu-Ko Jang Jian,1,2 Chih-Cherng Jeng,2 and Woo Sik Yoo3 1

Department of Electrophysics, National Chiayi University, Chiayi 60004, Taiwan Taiwan Semiconductor Manufacturing Company, Ltd., No. 1-1, Nan-Ke Rd., Science-Based Industrial Park, Tainan, 741-44, Taiwan 3 WaferMasters, Inc., 254 East Gish Road, San Jose, CA 95112, USA 2

(Received 1 October 2012; accepted 16 November 2012; published online 27 November 2012)

Thin thermal oxide film (∼36 nm) was grown on p- -Si (100) wafers in a vertical furnace at 950 ◦ C for 90 min in 1 atm dry O2 as a vehicle for monitoring metal contamination. They are annealed in separate vertical furnaces at 1100◦ C for 120 min in N2 and tested for metal contamination using multiwavelength room temperature photoluminescence (RTPL), inductively coupled plasma mass spectroscopy (ICP-MS) and secondary ion mass spectroscopy (SIMS). Significant RTPL intensity and spectral variations, corresponding to the degree of metal contamination, were observed. Nondestructive wafer mapping and virtual depth profiling capabilities of RTPL is a very attractive metal contamination monitoring technique. Copyright 2012 Author(s). This article is distributed under a Creative Commons Attribution 3.0 Unported License. [http://dx.doi.org/10.1063/1.4769746]

Low level metal contaminants such as Al, Fe, Cr, W, Ta on wafers can cause defects in Si substrates and degrade thin gate oxide quality in ultralarge-scale integration (ULSI) circuits causing serious degradation of performance and yield.1, 2 Cu, Fe, and Ni can dissolve in Si and form silicides.3 Metallic contaminants are typically from processing chemicals (used during cleaning, stripping, photolithography, deposition, etching and polishing), process equipment and wafer handling. Fabs manufacturing complementary-metal-oxide-semiconductor image sensors (CIS) and ULSI products are prone to simultaneously cross contamination between products due to the usage of materials with many metal impurities, such as color filters in CIS products.4 Semiconductor manufacturers can significantly increase device yield by reducing contamination. Pixels in CISs have been constantly scaled down in size and are approaching ∼1.0 μm. Improvements in critical device performance parameters, such as dark current per pixel, which suffers from defects and accidental metallic contamination during processing, are required for manufacturing high performance CISs.5 As a device failure analysis technique, the detection and characterization of deep-levels in CISs using dark-current and deep-level transient spectroscopies, have employed gold and tungsten implanted in the CISs to identify the deep-levels responsible for the increase in dark current.5 More practical contamination characterization techniques, which can be implemented on the manufacturing floor, must be developed. Trace metal contamination is typically measured using total reflection x-ray fluorescence (TXRF)6–9 and inductively coupled plasma mass spectrometry (ICP-MS).10 TXRF is a nondestructive characterization technique capable of mapping metal contamination. However, it has limited sensitivity to low Z (atomic number or proton number) elements (such as Na, Mg and Al) and no sensitivity to Li, Be and B. To improve the sensitivity to the low Z elements, a rotating target X-ray tube is used in some commercial systems for generating high intensity X-ray. Some systems equipped with X-ray tubes with different target materials (such as W, Cr, Mo and Au) for analyzing specific elements of interest. Typical lateral resolution is ∼10 mm. ICP-MS can detect a wide 2158-3226/2012/2(4)/042164/8

2, 042164-1

C Author(s) 2012


042164-2

Jang Jian, Jeng, and Yoo

AIP Advances 2, 042164 (2012)

range of elements and is very sensitive to Na, Mg, Al, Li, Be, B and P. However, it is a destructive characterization technique and cannot provide local contamination information. Mapping of Na, Mg, Al, Li, Be and B contamination is not available from TXRF and ICP-MS. New contamination mapping techniques for those elements must be developed for effective contamination monitoring and control in device manufacturing. Surface photo voltage (SPV) and photoconductance decay (PCD) are widely used for the evaluation of metal contamination of lightly doped silicon material.11 Room temperature photoluminescence (RTPL) has been proposed as a non-destructive technique for characterization of metal (Fe, Cu) contamination in silicon material.11–13 In this study, feasibility of multiwavelegth RTPL as an in-line metal contamination monitoring and mapping technique for the low Z elements (such as Na, Mg and Al, with limited sensitivity for TXRF measurements) was studied using metal contamination monitoring wafers processed in commercially available vertical furnaces. Three hundred millimeter (300 mm) diameter, prime p- -Si(100) wafers were used as metal contamination monitoring wafers in CIS manufacturing. For multiwavelength RTPL and vapor phase decomposition (VPD) ICP-MS analysis, very thin (∼36 nm thick) thermal oxide was grown in a commercially available vertical furnace at 950◦ C for 90 min in 1 atm dry O2 . The contamination monitor wafers were annealed with CIS device wafers in several vertical furnaces at 1100◦ C for 120 min in N2 . The wafers were analyzed by TXRF, multiwavelength RTPL, VPD ICP-MS and secondary ion mass spectroscopy (SIMS). SIMS analysis is done to investigate diffusion of metal contaminants into the Si. Visible and infra-red (IR), cw laser lines of 650 and 827 nm, as the excitation source, were irradiated onto the front side of Si wafers, penetrating ∼36 nm thin thermal oxide (SiO2 ) for RTPL measurements. Laser beam size on the wafer surface was approximately 50 μm in diameter. The incident laser excitation power at the wafer surface was fixed at 20 mW for 650 nm excitation and 50 mW for 827 nm excitation. The penetration depths in Si of the excitation wavelengths of 650 nm and 827 nm are ∼4.0 μm and ∼10 μm, respectively (Fig. 1). The incident power at the wafer surface was set lower for the shorter excitation wavelength (650 nm) to avoid wafer heating. The RTPL light was collected through an objective lens and passed through a combination of filters. The RTPL spectrum was analyzed by a thermoelectrically (TE) cooled IR spectrograph system (WaferMasters MPL-300).14–16 The exposure time for RTPL measurements was in the range of 20 ∼ 1000 ms. RTPL line scans and wafer mapping, up to 15,101 points per wafer, were done to gain local contamination information. All metal contamination monitor wafers were analyzed by TXRF for initial screening and VPD ICP-MS, for a wide range of elements. If no metal contaminants are found from the monitor wafers, CIS device wafers go to the next process step. A number of monitoring wafers pass the TXRF test, but fail the VPD ICP-MS, for low Z elements such as Na, Mg and Al. Device performance and yield variations within and between wafers and batches are recorded and correlated with metal contamination levels and elements. There is generally very strong correlation between metal contamination and device yield loss. The metal contamination must be eliminated for manufacturing stable devices. To effectively identify and eliminate the source of metal contamination, the quantity and distribution of contamination (local contamination information or contamination mapping) are very important. The VPD technique scans the entire surface with a water droplet with a very small amount of SiO2 etchant (HF and H2 O2 ). The VPD technique can be integrated in TXRF systems and ICP-MS systems.9, 10 In VPD TXRF, the water droplet, containing surface contaminants, is dried and its residue is analyzed with various X-ray sources. The benefit of VPD TXRF is that concentrated contaminants, from the entire surface, are dissolved and present in the residue, which significantly lowers the detection limits of elements per unit area, but at the cost of losing local information. However, detection of the low Z elements is still not possible. In contrast, the droplet is removed after scanning, and is diluted for ICP-MS analysis. The diluted droplet is then analyzed by ICP-MS. Since the VPD ICP-MS is a destructive characterization technique, measurements were done in the following sequence; TXRF → multiwavelength RTPL → VPD ICP-MS → SIMS. Among many metal contamination monitoring wafers, RTPL measurement results of five selected wafers are described in this paper (Fig. 2). One wafer (A) is the control wafer without metal


042164-3


AIP Advances 2, 042164 (2012)

FIG. 1. Excitation wavelength dependence of PL probing depth in Si.

FIG. 2. Location of wafers annealed in a clean furnace and two “suspect” furnaces.

contamination. Four wafers (B, C, D and E) were annealed wafers in two “contamination suspected” vertical furnaces at 1100◦ C for 120 min, in 1 atm N2 . No oxide thickness variation was measured after annealing in N2 . Metal contamination was not detected from any of the five wafers by TXRF. RTPL spectra measured at the center of wafers A ∼ E under 650 nm and 827 nm excitation are plotted in the Fig. 3(a) and 3(b). More then one order of magnitude RTPL intensity difference, between the control wafer A and other wafers B ∼ D was measured. To avoid saturation of the RTPL detector, the exposure time for the measurements was adjusted. To make spectral comparisons easier, the RTPL intensity per unit time (1 s) is used as the vertical axis of the plots. The RTPL intensity of control wafer A was plotted as 1/10 of actual intensity, for easy RTPL spectra comparisons between wafers. Actual RTPL intensity of the control wafer A is 10 times higher than the plotted intensity. The wafers B ∼ E, which were annealed in two different “suspect” furnaces, showed significantly lower RTPL intensity, regardless of excitation wavelength. For 650 nm excitation RTPL intensity, the wafers B and C showed stronger intensity than wafers D and E. Noticeable intensity variation between the wafers B and C was observed, even


042164-4


AIP Advances 2, 042164 (2012)

FIG. 3. RTPL intensity under 650 nm (a) and 827 nm (b) excitation. Normalized RTPL intensity (c) and the ratio of 650 nm excited RTPL intensity to 827 nm excited RTPL intensity (d).

though they are annealed in the same batch, in the same “suspect” furnace. This may be due to the difference between the center and edge of the furnace. For 827 nm excitation, the RTPL intensity difference between wafers B and C are smaller, compared to that under 650 nm excitation. The larger RTPL intensity variations under shorter excitation wavelength, with shallower penetration depth, suggests that the non-radiative recombination centers such as defects, contamination and deep levels, are located near the SiO2 /Si interface. The RTPL intensity difference between wafers annealed in different “suspect” furnaces (namely, Wafer B, C versus Wafers D, E) is very significant under the two different excitation wavelengths. If the RTPL intensity decrease is due to heavy metal contamination, such as Cu or Fe, as reported in references 8 and 9 we should have detected those contaminants by TXRF analysis. Since none of the wafers showed the presence of high Z elements, beyond P, contamination of low Z elements are suspected. Normalized RTPL intensity, and the ratio of 650 nm excitation RTPL intensity to 827 nm excitation RTPL intensity (I650 /I827 ) are plotted in Fig. 3(c) and 3(d). The wafers B, C, D and E showed less than 1/10 of the RTPL intensity of the control wafer A, regardless of excitation wavelength. The 827 nm excited RTPL intensity of wafers B, C, D and E, with deeper penetration depth, are always stronger than those from 650 nm excited RTPL intensity, thus, the presence of non-radiative recombination centers, at or near the surface and SiO2 /Si interface, is suspected. The RTPL intensity ratios (I650 /I827 ) of wafers B and D, which were annealed at the center of the two “suspect” furnaces, were larger than those of the wafers C and E, which were annealed at the end of one of the two “suspect” furnaces. This implies that the annealing at the center of the furnace results in better SiO2 /Si interface quality compared to the edge of the furnaces. Figure 4 shows the 281 point RTPL line scan results of the wafers A, B and D under 650 nm and 827 nm excitation. The line scan measurement was done in 1 mm intervals with the edge exclusion of 10 mm. The intensity was normalized to the maximum RTPL intensity within each wafer, for each excitation wavelength. The RTPL line scan results of each wafer under 650 nm and 827 nm excitations are quite similar in intensity pattern. The control wafer A showed reasonably sharp and uniform RTPL signals across the wafer. The wafer B showed significant within wafer RTPL intensity variations, regardless of excitation wavelengths. The left half showed lower RTPL intensity compared to the right half of the wafer. The RTPL intensity variations are larger under 650 nm excitation suggesting large quality variations at the surface and SiO2 /Si interface. Wafer


042164-5


AIP Advances 2, 042164 (2012)

FIG. 4. RTPL line scan spectra from wafers A, B and D under 650 nm and 827 nm excitation. Line scan (281 points) measurements were done in 1 mm intervals across each 300 mm wafer.


042164-6


AIP Advances 2, 042164 (2012)

FIG. 5. RTPL wafer maps (15,101 points per wafer) under 650 nm and 827 nm excitation. Gas flow direction from nozzle is indicated by red arrows. Relative location of “ring”-shaped boat is indicated by blue arrows.

D showed weaker RTPL intensity, as well as, two characteristic weak points near the wafer edges under all excitation wavelengths. The large RTPL intensity dips, near the wafer edge are not very obvious from the figure, due to the overlap with neighboring RTPL spectra. The difference between RTPL line scan results from wafers B and D strongly suggests that there is significant difference in distribution of non-radiative recombination centers in both plane and depth directions. It may represent the local contamination information. To investigate this further, high resolution RTPL mapping of the entire surface was done for all five wafers under 650 nm and 827 nm excitation. Figure 5 shows the multiwavelength high resolution (15,101 points/wafer/wavelength) RTPL wafer mapping results. In general, 650 nm excitation RTPL wafer maps show large within wafer variations. Auto color scale was used for wafer mapping. The measurement points with the lowest and highest RTPL intensity, within wafer, are in purple and red color, respectively. The wafer maps must be compared with RTPL line scan results shown in Fig. 4. Since the color of the wafer map represents the range of variations within wafer, the control wafer, with small RTPL intensity variation, still shows noticeable variation in color. The color scale in Fig. 5 is somewhat exaggerated. In reality, the RTPL intensity variation of the control wafer A is the smallest. The wafer stage pattern (stripes and radial lines) is visible from the RTPL wafer maps of the control wafer A. Since the RTPL intensity is very uniform, the wafer stage pattern became apparent. The range (maximum – minimum) of RTPL intensity variations within wafer is in order, A < B < C < D < E. Wafers B and C which were annealed in the same “suspect” furnace showed almost identical RTPL maps. Wafers D and E also show very similar maps. The presence of a “ring”-shaped dark RTPL region is very apparent. Due to the strong contrast near the “ring” area, the RTPL intensity variations in the center of wafers D and E appeared to be less obvious. This pattern difference in RTPL intensity maps may indicate the difference in the origin and mechanism of the contamination of low Z elements in the two different “suspect” furnaces. The horizontal stripe patterns on wafers were the actual RTPL signature from the wafers. The stability of RTPL system was confirmed by measuring prime Si


042164-7


AIP Advances 2, 042164 (2012)

wafers with native oxide before and after measuring the five contamination monitoring wafers. No horizontal stripe patters were observed from the prime Si wafers with native oxide. After RTPL measurements, metal contamination level of all five wafers were measured by VPD ICP-MS. No contaminants were detected from control wafer A. Wafers B and C showed Al concentration ∼2.5 × 1010 atoms/cm2 and Mg concentration ∼1.0 × 1010 atoms/cm2 . Wafers D and E showed Al concentration ∼6.0 × 1010 atoms/cm2 and Mg concentration ∼3.5 × 1010 atoms/cm2 . The desired contamination control level for CISs is