- Email: [email protected]
Contamination by slow diffusers in ion implantation processes: The examples of molybdenum and tungsten
Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research B journal homepage: www.elsevier.com/locate/nimb
Contamination by slow diffusers in ion implantation processes: The examples of molybdenum and tungsten M.L. Polignano a,⇑, I. Mica a, F. Barbarossa a, A. Galbiati a, S. Grasso a, V. Soncini b a b
ST Microelectronics, via Olivetti, 2, 20864 Agrate Brianza, MB, Italy Micron, via Olivetti, 2, 20864 Agrate Brianza, MB, Italy
a r t i c l e
i n f o
Article history: Received 22 April 2015 Accepted 27 April 2015 Available online 15 May 2015 Keywords: Contamination Implantation Molybdenum Tungsten DLTS
a b s t r a c t A procedure to measure molybdenum and tungsten contamination in implantation processes by DLTS (Deep Level Transient Spectroscopy) is defined and calibrated for the evaluation of molybdenum and tungsten contaminant dose. The obtained calibrations are used to study molybdenum contamination in BF2 implantations and tungsten contamination by sputtering from a previously contaminated wafer holder. In molybdenum-implanted samples, the molybdenum level located 0.3 eV above valence band is revealed only. In tungsten-implanted samples, two levels are revealed. One of these levels is the tungsten-related hole trap located 0.4 eV above valence band. The other level does not correspond to any tungsten-related level, however it is related to the presence of tungsten and to the sample preparation process. The SPV (Surface Photovoltage) measurement sensitivity to tungsten contamination was also tested, and it was found much lower than the DLTS sensitivity, due to the low tungsten diffusivity. This procedure was used to evaluate contamination in implantation processes. In BF2 implantations, in addition to molybdenum, tungsten contamination is found. Molybdenum and tungsten contamination is found in boron implantation too. The tungsten contamination induced by implantation in a previously contaminated implanter was quantified, and the efficiency of arsenic implantation as a decontamination process was tested. Finally, it was shown that TXRF (Total reflection X-ray Fluorescence) is much less sensitive than DLTS for monitoring tungsten contamination. Ó 2015 Elsevier B.V. All rights reserved.
1. Introduction Metal contamination in ion implantation has been widely studied [1,2]. Schematically, two main contamination mechanisms have been identified in ion implantation, i.e. sputtering from parts close to the wafer during the implantation and mass interference, either with or without changes of the ion mass and charge between the ion extraction and the mass selection. Sputtering was shown to be responsible for iron contamination [3], which can be reduced by implantation through a thin screen layer and can be suppressed by a suitable design of the wafer holder [4]. Well-known examples of mass interference are molybdenum contamination in BF2 implantation [5] or in indium implantation when fluorine is present in the ion source [6], or even in boron implantation [7]. ⇑ Corresponding author. E-mail addresses: [email protected] (M.L. Polignano), isabella.mica@st. com (I. Mica), [email protected] (F. Barbarossa), [email protected] (A. Galbiati), [email protected] (S. Grasso), [email protected] (V. Soncini). http://dx.doi.org/10.1016/j.nimb.2015.04.069 0168-583X/Ó 2015 Elsevier B.V. All rights reserved.
To quantify metal contamination from ion implantation processes, it is necessary to take into account that contamination is frequently located inside the silicon matrix, so surface techniques such as Total Reflection X-ray Fluorescence (TXRF) can miss or underestimate the contaminant amount. In addition, in the case of contamination with slow diffusers such as molybdenum, a very small amount per unit area may result in a significant concentration in the near-surface silicon volume, so this sort of contamination can be difficult to catch by surface techniques and still be harmful for devices. For instance, molybdenum and tungsten contamination were found to be detrimental for imager sensor devices even in very low concentration [8]. In addition, these contaminants are frequently missed by techniques based on recombination lifetime measurements, because the region analyzed by these techniques is usually much deeper than the region involved in molybdenum and tungsten diffusion [9–11]. It was shown that the Deep Level Tungsten Spectroscopy (DLTS) is the best choice to detect contamination by slow diffusers [12].
165
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
However, DLTS measurements give the contaminant concentration in the space charge region of a p–n junction or of a Schottky diode, i.e. in a limited region of the wafer. The formation of the required structures involves some process steps which may result in a loss of contaminant dose, so a calibration is needed to turn DLTS concentration data into contaminant amount per unit area. In this paper, we set up a procedure to measure molybdenum and tungsten contamination in implantation processes by DLTS, and calibrated this procedure for the evaluation of molybdenum and tungsten contaminant dose. The obtained calibrations are used to study molybdenum contamination in BF2 implantations and tungsten contamination by sputtering from a previously contaminated wafer holder.
2. Experimental details 2.1. Sample preparation P-type, (100), 200 mm diameter, 725 lm thick 10 Xcm resistivity Magnetic Czochralski (MCZ) wafers and n-type, 5 Xcm resistivity CZ wafers were used in this study. Some samples were implanted with molybdenum or tungsten to calibrate DLTS measurements in terms of contaminant dose. The implantation conditions for molybdenum and tungsten are reported in Table 1. The wafers for monitoring the contamination introduced by ion implantation were implanted with various dopant ions (BF2, boron, arsenic) in the conditions reported in Table 2. The implantations were performed in an Axcelis NV-GSD200EE/80 High Current ion implanter. In BF2 and boron implantations mass interference is expected to be the dominant contamination mechanism [5,7] and for this reason the impact of the material used for the ion source chamber (molybdenum or tungsten) was investigated. Vice versa, arsenic in known to be responsible for a relevant contamination by sputtering [3,4], and therefore arsenic implantations (1015 cm 2, 60 keV) were used to monitor the contamination of the equipment after previous implantations of wafers with an exposed metal layer (tungsten). Some monitor wafers were implanted before contamination of the equipment, after contamination and after some decontamination cycles consisting of high dose arsenic implantations (1–21016–cm 2, 60 keV, 6.51016 cm 2, 75 keV) of dummy wafers, aimed at removing contamination by sputtering. After implantation, the metal surface concentration was measured by TXRF. The calibration wafers and the wafers for contamination monitoring were treated according to the same process flow. The details of the flow used to prepare the calibration wafers and the samples to study the contamination from implantation in a contaminated implanter are reported in Table 3(a) and (b), respectively. After implantation, the wafers were thermally treated by A Rapid Thermal Process (RTP) at 1100 °C for 3 min in an inert environment. This thermal treatment had the purpose to allow the contaminant diffusion in silicon. Then, the wafers received a Reactive Ion Etching (RIE) of 1.2–1.4 lm silicon with the aim to remove the doped layer. The wafers implanted with molybdenum and tungsten were etched too, so the data obtained in these wafers can be directly compared to those of contamination monitor wafers. The samples for recombination lifetime measurements require no further treatment. To obtain Schottky diodes for DLTS
Table 1 Implantation conditions of the calibration samples. Ion Mo W
Energy (keV) 8 8
Dose (cm 10
2
)
1.510 –1011 1.51010–1.51011
Table 2 Implantation conditions of the samples prepared for contamination monitoring. Ion BF2 B As
Energy (keV) 4 6 60
Dose (cm 14
2
)
Ion source
15
Mo, W Mo, W Mo
310 –310 31015 1015
measurements, the native oxide was etched off and 1000 Å titanium layer was deposited on the silicon surface, masked and etched. Time-of-Flight Secondary Ion Mass Spectroscopy (ToF-SIMS) was used to investigate whether metal surface segregation may happen during the RTP thermal treatment. The samples for this investigation were implanted in the same conditions as in Table 1 and were annealed by the same RTP as the samples prepared for SPV and DLTS measurements. Of course, the samples prepared for ToF-SIMS analysis received no silicon etching. As-implanted and annealed samples were compared to assess metal surface segregation.
2.2. Experimental techniques TXRF measurements were obtained by a RigakuTXRF300 instrument in direct mode, with the high energy beam at 0.05° angle and 500 s acquisition time for each point. The tool has a W filament working at 30 kV and 300 mA. 1 mm2 area Schottky diodes were measured by DLTS. A Semilab DLS-83D instrument was used. In this instrument, lock-in integration is used for averaging capacitance transients, and temperature can be scanned in the range 30–300 K. Alternatively, constant temperature spectra can be obtained as a function of the frequency of excitation pulses in the range 0.5 Hz–2 kHz [13]. Both methods were used in this work. The differential DLTS method was used. In this method, the Schottky diode (or the p–n junction) is reverse biased at a voltage Vr and two filling pulses are applied: the first pulse V1 is applied at the beginning of the lock-in integration period; and a second pulse V2 is applied a half period later. In the lock-in integration, the difference DC is obtained between the integrals of the capacitance transients caused by the first pulse and by the second pulse. This method yields the trap concentration Nt in the interval [xd(V1), xd(V2)], where xd is the depletion region edge at a given reverse voltage. In our measurements, samples were reverse biased at 5 V, and filling pulses with amplitudes of 0.5 V and 4.5 V were applied with a pulse width of 20 ls during each integration period. Under these conditions, a region ranging from about 0.8 lm to 2 lm was analyzed. The spectra shown in this paper were obtained with a 23 Hz filling pulse frequency. Schottky diodes were used for these measurements, so majority carrier traps could be revealed only. SPV measurements are carried out by illuminating the sample with light of various wavelengths. Generated minority carriers are collected in the depletion region at the wafer surface and produce a variation in surface potential, which is recorded as a function of light wavelength. In the standard SPV technique [14], the diffusion length Ldiff (or the recombination lifetime s) is extracted from these data by assuming that sample thickness is much larger than diffusion length. This hypothesis often fails in present wafers, and the actual wafer thickness must be taken into account. An ‘‘enhanced SPV’’ [15] technique is available for this purpose. In this work, enhanced SPV was generally used. Both standard and enhanced SPV are only sensitive to bulk recombination, so SPV cannot detect surface-segregated
166
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
Table 3 Flow for the preparation of the calibration wafers (a) and of the samples for the experiments about contamination by implantation in a contaminated implanter (b).
contaminants. In SPV measurements, the probed region is of the order of light penetration depth 1/a (where a is the absorption coefficient) plus carrier diffusion length. This region typically ranges from about 100 lm to the whole wafer thickness, so it is in any case much deeper than the region analyzed by DLTS. SPV measurements can also be associated with optical activation to quantify iron contamination by dissociation of the iron-boron pair [16]. In p-type silicon, the equilibrium state of iron at room temperature is the iron–boron pair. This pair can be dissociated by illumination, and the resulting interstitial metal is more effective as a recombination center than the iron-boron pair, so the carrier lifetime after dissociation (safter) is shorter than before dissociation (sbefore). The quantity 1=safter 1=sbefore is proportional to the iron concentration in the solid solution, so this method can be used to quantify the iron concentration. In addition, the concentration of recombination centers different from iron can also be estimated by assuming a somehow arbitrary capture cross section. In this work the SPV tool in the FAaST SDI system was used. This instrument is equipped with a set of filters producing monochromatic light at seven different wavelengths, ranging from 800 nm to 1 lm, and with a halogen lamp for optical activation. TOF-SIMS data were acquired by an ION TOF IV dual beam TOF-SIMS in the negative modality using a Ga+ primary ion beam operating at 25 keV, 2.2 pA, rastering over 100 100 lm2 area. Sputtering was accomplished by Cs+ at 250 eV energy and 10 nA current for 100 s and the raster was 300 300 lm2. This allows analyzing a region ranging from the surface to about 2 nm depth. This method was chosen in order to increase the WOx secondary ion yield. The integrated spectra are reported instead of the depth profiles to improve the signal-to-noise ratio.
3. Experimental results 3.1. Calibration of DLTS measurements 3.1.1. Molybdenum Fig. 1 reports the DLTS spectra of molybdenum-implanted wafers. One level is revealed only, located at Ev + 0.3 eV (where
Fig. 1. DLTS spectra in molybdenum-implanted p-type samples.
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
167
Fig. 2. Arrhenius plot of ep/T2 obtained in molybdenum-implanted samples. Literature data are shown for a comparison.
Ev is the valence band edge). Fig. 2 compares the Arrhenius plot of ep/T2 (where ep is the hole emission rate and T is the measurement absolute temperature) obtained from these samples with literature data [9] for molybdenum-contaminated p-type silicon. These data agree very well with each other, confirming that the peak reported in Fig. 1 can be identified with molybdenum. The molybdenum concentration measured by DLTS is reported as a function of the implanted dose QMo in Fig. 3, and is found to increase in proportion to the implanted dose, as expected. According to these data, by DLTS we obtain about 81012 cm 3 molybdenum concentration for 1010 cm 2 molybdenum dose.
3.1.2. Tungsten Fig. 4 reports the DLTS spectra of tungsten-implanted wafers. Two peaks can be identified, named H1 and H2. Fig. 5 compares the Arrhenius plots of ep/T2 related to these peaks with literature data [17] for tungsten-contaminated p-type silicon. This comparison allows us to identify H1 with the tungsten hole trap located at Ev + 0.4 eV. The level H2 is located at Ev + 0.36 eV, and does not correspond to any previously reported tungsten-related level [17,18]. The origin of H2 will be discussed in Section 3.3. No deep levels were revealed in n-type silicon, so we did not detect the electron trap reported in Ref. [17]. Fig. 6 reports the concentrations of H1 and H2 as a function of the tungsten dose QW. Here, solid and empty symbols refer to
Fig. 3. Molybdenum concentration measured by DLTS as a function of the molybdenum implanted dose.
Fig. 4. DLTS spectra of tungsten-implanted p-type wafers. These samples were treated by the same process flow as used for arsenic-implanted samples, including the RIE of silicon.
Fig. 5. Arrhenius plot of ep/T2 obtained in tungsten-implanted samples. Literature data are shown for a comparison.
two separate (but nominally identical) set of samples. The concentration of these levels increases in proportion to the tungsten dose up to 41010 cm 2, with about 30% variation between different sets of samples. In these conditions, if we consider the dominant peak H1 only, we obtain 5–6.51012 cm 3 tungsten concentration for 1010 cm 2 implanted dose. When the dose is increased above 51010 cm 2, a saturation is observed, followed by a reduction of the active tungsten concentration. These data show that tungsten deactivates at a concentration of about 2–31013 cm 3, possibly by surface segregation. ToF-SIMS analyses were carried out to confirm tungsten segregation at the wafer surface. Two wafers implanted with the highest dose in this test (1.51011 cm 2) were compared, an as-implanted wafer and a wafer annealed with the same thermal treatment as used for DLTS samples. ToF-SIMS analyses (Fig. 7) revealed tungsten at the wafer surface in the annealed wafer only, thus confirming that in this wafer tungsten segregated at the wafer surface during the thermal treatment. The concentration of impurities different from iron (‘‘Other impurities’’) obtained from SPV measurements shows some correlation with the tungsten dose, as shown in Fig. 8. However, the concentrations reported in Fig. 8 are quite low and close to the
168
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
Fig. 8. Concentration of impurities different from iron obtained from SPV measurements as a function of the tungsten dose.
Fig. 6. Trap concentration measured by DLTS as a function of tungsten dose. Solid and empty symbols refer to two sets of samples.
measurement sensitivity, so the SPV sensitivity to tungsten contamination is much lower than the DLTS sensitivity. This fact is not surprising, because the region involved in tungsten diffusivity is much thinner [9,10] than the region analyzed by SPV, which is of the order of the whole wafer thickness in our experimental conditions. 3.2. Contamination in implantation processes 3.2.1. Molybdenum contamination in BF2 and boron implanted silicon Fig. 9 shows the DLTS spectra of wafers implanted with 31014 cm 2 BF2 ions in implanters equipped with a molybdenum or a tungsten source chamber. In both wafers the molybdenum and the tungsten peak are revealed, however the molybdenum concentration is strongly reduced if the tungsten source is used. This fact is expected since mass interference of doubly ionized molybdenum in BF2 implantation is very well known [5], so it is expected that molybdenum from the source chamber may reach
Fig. 7. ToF-SIMS spectra of two samples implanted with 1.51011 cm dose. , as implanted, , annealed sample.
2
tungsten
the wafers. Tungsten contamination in BF2 implantation was previously observed [19] and ascribed to a mass interference mechanism, however the data in Fig. 9 show that the tungsten concentration is almost independent of the material used for the source chamber. Therefore, in our experiment mass interference is not likely to be the dominant mechanism for tungsten contamination. Molybdenum and tungsten concentration data were converted into dose data by using the calibration in Section 3.1. The so-obtained molybdenum and tungsten doses are reported as a function of the BF2 dose in Fig. 10, and are found to increase in proportion to the BF2 implanted dose. From these data, the molybdenum and tungsten contamination doses are a fraction of 2.610 5 and 0.610 6 respectively of the BF2 dose, if a molybdenum source is used. Fig. 11 shows the DLTS spectra of wafers implanted with 31015 cm 2 boron ions in implanters equipped with a molybdenum or a tungsten source chamber. Also in this case, both the molybdenum and the tungsten peak are revealed. As in BF2 implantation, the molybdenum concentration is strongly reduced by using a tungsten source. It was also shown that molybdenum
Fig. 9. DLTS spectra of wafers implanted with BF2 by using implanters equipped with molybdenum or tungsten sources.
169
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
Fig. 10. Molybdenum (a) and tungsten (b) contamination dose as a function of the BF2 dose, in an implanter equipped with a molybdenum source.
Fig. 11. DLTS spectra of wafers implanted with boron by using implanters equipped with molybdenum or tungsten sources.
Fig. 12. DLTS spectra of samples implanted with arsenic before contamination of the implanter, after tungsten contamination and after some decontamination processes.
mass interference can also affect boron implantations [7], so also in boron implantation molybdenum from the source may contaminate the implanted wafers. However in boron implantations the contamination mechanism is much less effective than in BF2 implantations, so the ratio between contaminant dose and implanted dose is lower in boron than in BF2 implantations. For what concerns tungsten, in boron implantation tungsten contamination seems to be sensitive to the source material, however the available data are not enough to formulate hypotheses about the origin of this contamination.
the wafer with very low energy [3,4]. Therefore, the tungsten dose estimated from the data in Fig. 12 can be compared with TXRF measurements of surface contamination after implantation. This comparison is reported in Table 4, and shows that TXRF detects tungsten contamination in the sample implanted after contamination of the implanter only. The lowest dose decontamination process is enough to reduce contamination below the TXRF detection limit, however DLTS measurements show that some
3.2.2. Tungsten contamination from a contaminated implanter Fig. 12 shows the DLTS spectra of arsenic-implanted samples (1015 cm 2, 60 keV) before contamination of the implanter, after contamination and after some decontamination processes. The same peaks as in tungsten implanted samples are observed. In Section 3.1.2 peak H1 was identified with tungsten and calibrated to obtain the tungsten dose, so the data in Fig. 12 can be used to estimate the tungsten dose diffused in silicon. In this experiment, contamination is mainly due to sputtering from parts close to the wafer during implantation. This mechanism is usually responsible for near-surface contamination, because contaminant atoms reach
Table 4 Tungsten concentration per unit area measured by TXRF and estimated from DLTS spectra, in samples implanted with arsenic before contamination of the implanter, after tungsten contamination and after some decontamination processes. Arsenic implantation conditions Before contamination After contamination, no decontamination Decontamination, 1016 cm 2 Decontamination, 21016 cm
2
[W]TXRF (1010 cm
2
)
QW,DLTS (1010 cm
2
3
<1.5 8.1 ± 0.8
<410 1.4 ± 0.1
<1.5 <1.5
0.87 ± 0.08 0.32 ± 0.03
)
170
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
tungsten contamination is still present in these wafer. On the other hand, the tungsten concentration at the surface measured by TXRF in the sample implanted after contamination of the implanter is higher than the datum estimated from DLTS spectra, possibly because of partial tungsten inactivation in this sample (see Fig. 6). It can be noted that in this example DLTS has excellent sensitivity per unit area. This sensitivity is obtained because DLTS analyzes a rather shallow region (2–3 lm deep), and the tungsten diffusion profile is limited to this region due to its low diffusivity. According to DLTS data, a very high arsenic dose is required to achieve an almost complete decontamination of the implanter.
3.3. Investigation about the H2 level The identification of the H2 level requires some more analysis. As shown in the previous sections, this level is observed both in tungsten-implanted wafers and in wafer contaminated with tungsten by arsenic implantation in a contaminated implanter. While level H1 can be identified with a well-known tungsten level, level H2 does not correspond to any tungsten-related level [17,18]. To confirm that H2 too is related to tungsten, in Fig. 13 we report the concentration of H2 as a function of the concentration of H1. Data from both tungsten-implanted samples and samples contaminated with tungsten by implantation in a contaminated implanter were included in this plot. The concentration of H2 increases in proportion to the concentration of H1, indicating that H2 too is related to tungsten contamination. To investigate whether the preparation used for these samples is involved in the formation of the level H2, some additional tungsten-implanted samples were prepared according to the same flow as in Table 3(a), with the exception that the silicon etching step was skipped. Fig. 14 shows the DLTS spectrum of a wafer implanted with 31010 cm 2 tungsten and not etched. In this spectrum, the level H1 only is found, showing that the level H2 is formed upon etching. On the other hand, Fig. 13 shows that H2 is related to tungsten, so H2 is probably a complex between tungsten and some species introduced by RIE. The wafers receive no thermal treatment after etching, and the maximum temperature reached in the metallization and masking process is about 100 °C for a few minutes. Tungsten is a slow diffuser, and cannot diffuse at such low temperatures. For this reason, the species introduced by RIE must be a very fast diffuser, which can diffuse through the silicon lattice and reach tungsten atoms even at the low temperatures involved in this part of the sample preparation flow.
Fig. 14. DLTS spectrum of a sample implanted with 31010 cm annealed, but not etched.
2
tungsten dose and
Fig. 15. DLTS spectrum of a sample implanted with 31010 cm annealed, before and after 250 °C 45 min baking.
2
tungsten dose and
The effect of a low temperature thermal treatment on the level H2 was also studied. Fig. 15 reports the DLTS spectra of a sample implanted with 31010 cm 2 tungsten, RTP annealed and etched, before and after 45 min baking at 250 °C. The baking treatment suppresses the H2 level, indicating that the H2 level can be easily destroyed by a moderate temperature treatment. In addition, the baking treatment increases the concentration of H1 up to the concentration measured in the not etched sample. A further thermal treatment (up to 5 h) at the same temperature yields no further modification of the DLTS spectrum. Chlorine and HBr are used for RIE silicon etching. Hydrogen is a very fast diffuser and easily forms complexes with many metals, so it is a possible candidate as the responsible for the formation of the level H2. However, our data are not enough for a positive identification of H2. 4. Conclusions
Fig. 13. Concentration of the level H2 as a function of the concentration of the level H1 in samples implanted with tungsten or contaminated with tungsten by arsenic implantation in a contaminated implanter.
A procedure for monitoring molybdenum and tungsten contamination in ion implantation processes by DLTS was set up and calibrated by using molybdenum and tungsten implanted samples.
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
The molybdenum concentration measured by DLTS was found to be proportional to the molybdenum dose over the whole range under study (up to 1011 cm 2). In molybdenum-implanted samples, the molybdenum level at Ev + 0.3 eV is revealed only. In tungsten-implanted samples, two levels are revealed, named H1 and H2. The level H1 is the tungsten-related hole trap, the level H2 does not correspond to any tungsten-related level, however it is related to tungsten and it is induced by the silicon dry etching included in the flow for monitoring contamination from ion implantation. The SPV sensitivity to tungsten contamination was also tested, and it was found much lower than the DLTS sensitivity, due to the low tungsten diffusivity. The concentration measured by DLTS saturates and then decreases starting from 51010 cm 2 tungsten dose. ToF-SIMS analyses revealed tungsten at the wafer surface, indicating that the observed tungsten inactivation is due to surface segregation. This procedure was used to evaluate contamination in implantation processes. In BF2 implantations, in addition to molybdenum, tungsten contamination is found. The molybdenum contamination in BF2 implantation is a well-known example of mass interference, and as expected the molybdenum concentration strongly depends on the material of the source chamber. Vice versa the tungsten contamination is roughly independent of the source material. Molybdenum and tungsten contamination is found in boron implantation too. The tungsten contamination induced by implantation in a previously contaminated implanter was quantified, and the efficiency of arsenic implantation as a decontamination process was tested. Finally, it was shown that TXRF is much less sensitive than DLTS for monitoring tungsten contamination.
References [1] F.A. Stevie, R.G. Wilson, D.S. Simons, M.I. Current, P.C. Zalm, Review of secondary ion mass spectroscopy characterization of contamination associated with ion implantation, J. Vac. Sci. Technol., B 12 (1994) 2263. [2] H. Ryssel, L. Frey, Contamination problems in ion implantation, in: J.F. Ziegler (Ed.), Handbook of Ion Implantation Technology, Elsevier Science Publisher B.V., Amsterdam, 1992, p. 675.
171
[3] M.L. Polignano, C. Bresolin, F. Cazzaniga, A. Sabbadini, G. Queirolo, Investigation of metal contamination by photocurrent measurements: validation and application to ion implantation processes in: J.K. Lowell, R.T. Chen, J.P. Mathur (Eds.), Optical Characterization Techniques for High Performance Microelectronic Device Manufacturing II, Proc. SPIE 2638, 1995, p. 14. [4] M.L. Polignano, D. Caputo, A. Giussani, V. Soncini, G. Di Toma, Metal Contamination Reduction in the Evolution of Ion Implantation Technology, in: Proc. of the XIII International Conference on Ion Implantation Technology, IEEE, 2000, p. 686. [5] A. Cubina, M. Frost, Effects of molybdenum contamination resulting from BF2 implantation, Nucl. Instr. Methods Phys. Res. B55 (1991) 160. [6] D. Codegoni, M.L. Polignano, V. Soncini, C. Bresolin, Molybdenum contamination in indium and boron implantation processes, Electrochem. Soc. Trans. 10 (2007) 85. [7] V. Haublein, L. Frey, H. Ryssel, Additional peaks in mass spectra due to charge exchange events and dissociation of molecular ions during extraction, Nucl. Instr. Methods Phys. Res., Sect. B 237 (2005) 346. [8] F. Russo, G. Moccia, G. Nardone, R. Alfonsetti, G. Polsinelli, A. D’Angelo, A. Patacchiola, M. Liverani, P. Pianezza, T. Lippa, M. Carlini, M.L. Polignano, I. Mica, E. Cazzini, M. Ceresoli, D. Codegoni, Proximity gettering of slow diffuser contaminants in CMOS image sensors, Solid-State Electron. 91 (2014) 91. [9] J.L. Benton, D.C. Jacobson, B. Jackson, J.A. Johnson, T. Boone, D.J. Eaglesham, F.A. Stevie, J. Becerro, Behavior of molybdenum in silicon evaluated for integrated circuit processing, J. Electrochem. Soc. 146 (1999) 1929. [10] K. Graff, Metal Impurities in Silicon-Device Fabrication, Springer, Berlin, 2000. ch. 4 and 5. [11] A. De Luca, A. Portavoce, M. Texier, C. Grosjean, N. Burle, V. Oison, B. Pichaud, Tungsten diffusion in silicon, J. Appl. Phys. 115 (2014) 013501. [12] M.L. Polignano, D. Codegoni, S. Grasso, I. Mica, G. Borionetti, A. Nutsch, A comparative analysis of different measurement techniques to monitor metal and organic contamination in silicon device processing, Phys. Status Solidi A 212 (2015) 495. [13] G. Ferenczi, C.A. Londos, T. Pavelka, M. Somogyi, A. Mertend, Correlation of the concentration of the carbon-associated radiation damage levels with the total carbon concentration in silicon, J. Appl. Phys. 63 (1988) 183. [14] A.M. Goodman, A method for the measurement of short minority carrier diffusion lengths in semiconductors, J. Appl. Phys. 32 (1961) 2550. [15] J. Lagowski, A.M. Kontkiewicz, L. Jastrzebski, P. Edelman, Method for the measurement of long minority carrier diffusion lengths exceeding wafer thickness, Appl. Phys. Lett. 63 (1993) 2902. [16] G. Zoth, W. Bergholz, A fast, preparation-free method to detect iron in silicon, J. Appl. Phys. 67 (11) (1990) 6764. [17] Y. Fujisaki, T. Ando, H. Kozuka, Y. Takano, Characterization of tungsten-related deep levels in bulk silicon crystal, J. Appl. Phys. 63 (1988) 2304. [18] S. Boughaba, D. Mathiot, Deep level transient spectroscopy characterization of tungsten-related deep levels in silicon, J. Appl. Phys. 69 (1) (1991) 278. [19] R.B. Liebert, G.C. Angel and M. Kase, Tungsten Contamination in BF2 Implants, in: Proc. of the 11th International Conference on Ion Implantation Technology, IEEE, 1997, p. 135.
Contents lists available at ScienceDirect
Nuclear Instruments and Methods in Physics Research B journal homepage: www.elsevier.com/locate/nimb
Contamination by slow diffusers in ion implantation processes: The examples of molybdenum and tungsten M.L. Polignano a,⇑, I. Mica a, F. Barbarossa a, A. Galbiati a, S. Grasso a, V. Soncini b a b
ST Microelectronics, via Olivetti, 2, 20864 Agrate Brianza, MB, Italy Micron, via Olivetti, 2, 20864 Agrate Brianza, MB, Italy
a r t i c l e
i n f o
Article history: Received 22 April 2015 Accepted 27 April 2015 Available online 15 May 2015 Keywords: Contamination Implantation Molybdenum Tungsten DLTS
a b s t r a c t A procedure to measure molybdenum and tungsten contamination in implantation processes by DLTS (Deep Level Transient Spectroscopy) is defined and calibrated for the evaluation of molybdenum and tungsten contaminant dose. The obtained calibrations are used to study molybdenum contamination in BF2 implantations and tungsten contamination by sputtering from a previously contaminated wafer holder. In molybdenum-implanted samples, the molybdenum level located 0.3 eV above valence band is revealed only. In tungsten-implanted samples, two levels are revealed. One of these levels is the tungsten-related hole trap located 0.4 eV above valence band. The other level does not correspond to any tungsten-related level, however it is related to the presence of tungsten and to the sample preparation process. The SPV (Surface Photovoltage) measurement sensitivity to tungsten contamination was also tested, and it was found much lower than the DLTS sensitivity, due to the low tungsten diffusivity. This procedure was used to evaluate contamination in implantation processes. In BF2 implantations, in addition to molybdenum, tungsten contamination is found. Molybdenum and tungsten contamination is found in boron implantation too. The tungsten contamination induced by implantation in a previously contaminated implanter was quantified, and the efficiency of arsenic implantation as a decontamination process was tested. Finally, it was shown that TXRF (Total reflection X-ray Fluorescence) is much less sensitive than DLTS for monitoring tungsten contamination. Ó 2015 Elsevier B.V. All rights reserved.
1. Introduction Metal contamination in ion implantation has been widely studied [1,2]. Schematically, two main contamination mechanisms have been identified in ion implantation, i.e. sputtering from parts close to the wafer during the implantation and mass interference, either with or without changes of the ion mass and charge between the ion extraction and the mass selection. Sputtering was shown to be responsible for iron contamination [3], which can be reduced by implantation through a thin screen layer and can be suppressed by a suitable design of the wafer holder [4]. Well-known examples of mass interference are molybdenum contamination in BF2 implantation [5] or in indium implantation when fluorine is present in the ion source [6], or even in boron implantation [7]. ⇑ Corresponding author. E-mail addresses: [email protected] (M.L. Polignano), isabella.mica@st. com (I. Mica), [email protected] (F. Barbarossa), [email protected] (A. Galbiati), [email protected] (S. Grasso), [email protected] (V. Soncini). http://dx.doi.org/10.1016/j.nimb.2015.04.069 0168-583X/Ó 2015 Elsevier B.V. All rights reserved.
To quantify metal contamination from ion implantation processes, it is necessary to take into account that contamination is frequently located inside the silicon matrix, so surface techniques such as Total Reflection X-ray Fluorescence (TXRF) can miss or underestimate the contaminant amount. In addition, in the case of contamination with slow diffusers such as molybdenum, a very small amount per unit area may result in a significant concentration in the near-surface silicon volume, so this sort of contamination can be difficult to catch by surface techniques and still be harmful for devices. For instance, molybdenum and tungsten contamination were found to be detrimental for imager sensor devices even in very low concentration [8]. In addition, these contaminants are frequently missed by techniques based on recombination lifetime measurements, because the region analyzed by these techniques is usually much deeper than the region involved in molybdenum and tungsten diffusion [9–11]. It was shown that the Deep Level Tungsten Spectroscopy (DLTS) is the best choice to detect contamination by slow diffusers [12].
165
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
However, DLTS measurements give the contaminant concentration in the space charge region of a p–n junction or of a Schottky diode, i.e. in a limited region of the wafer. The formation of the required structures involves some process steps which may result in a loss of contaminant dose, so a calibration is needed to turn DLTS concentration data into contaminant amount per unit area. In this paper, we set up a procedure to measure molybdenum and tungsten contamination in implantation processes by DLTS, and calibrated this procedure for the evaluation of molybdenum and tungsten contaminant dose. The obtained calibrations are used to study molybdenum contamination in BF2 implantations and tungsten contamination by sputtering from a previously contaminated wafer holder.
2. Experimental details 2.1. Sample preparation P-type, (100), 200 mm diameter, 725 lm thick 10 Xcm resistivity Magnetic Czochralski (MCZ) wafers and n-type, 5 Xcm resistivity CZ wafers were used in this study. Some samples were implanted with molybdenum or tungsten to calibrate DLTS measurements in terms of contaminant dose. The implantation conditions for molybdenum and tungsten are reported in Table 1. The wafers for monitoring the contamination introduced by ion implantation were implanted with various dopant ions (BF2, boron, arsenic) in the conditions reported in Table 2. The implantations were performed in an Axcelis NV-GSD200EE/80 High Current ion implanter. In BF2 and boron implantations mass interference is expected to be the dominant contamination mechanism [5,7] and for this reason the impact of the material used for the ion source chamber (molybdenum or tungsten) was investigated. Vice versa, arsenic in known to be responsible for a relevant contamination by sputtering [3,4], and therefore arsenic implantations (1015 cm 2, 60 keV) were used to monitor the contamination of the equipment after previous implantations of wafers with an exposed metal layer (tungsten). Some monitor wafers were implanted before contamination of the equipment, after contamination and after some decontamination cycles consisting of high dose arsenic implantations (1–21016–cm 2, 60 keV, 6.51016 cm 2, 75 keV) of dummy wafers, aimed at removing contamination by sputtering. After implantation, the metal surface concentration was measured by TXRF. The calibration wafers and the wafers for contamination monitoring were treated according to the same process flow. The details of the flow used to prepare the calibration wafers and the samples to study the contamination from implantation in a contaminated implanter are reported in Table 3(a) and (b), respectively. After implantation, the wafers were thermally treated by A Rapid Thermal Process (RTP) at 1100 °C for 3 min in an inert environment. This thermal treatment had the purpose to allow the contaminant diffusion in silicon. Then, the wafers received a Reactive Ion Etching (RIE) of 1.2–1.4 lm silicon with the aim to remove the doped layer. The wafers implanted with molybdenum and tungsten were etched too, so the data obtained in these wafers can be directly compared to those of contamination monitor wafers. The samples for recombination lifetime measurements require no further treatment. To obtain Schottky diodes for DLTS
Table 1 Implantation conditions of the calibration samples. Ion Mo W
Energy (keV) 8 8
Dose (cm 10
2
)
1.510 –1011 1.51010–1.51011
Table 2 Implantation conditions of the samples prepared for contamination monitoring. Ion BF2 B As
Energy (keV) 4 6 60
Dose (cm 14
2
)
Ion source
15
Mo, W Mo, W Mo
310 –310 31015 1015
measurements, the native oxide was etched off and 1000 Å titanium layer was deposited on the silicon surface, masked and etched. Time-of-Flight Secondary Ion Mass Spectroscopy (ToF-SIMS) was used to investigate whether metal surface segregation may happen during the RTP thermal treatment. The samples for this investigation were implanted in the same conditions as in Table 1 and were annealed by the same RTP as the samples prepared for SPV and DLTS measurements. Of course, the samples prepared for ToF-SIMS analysis received no silicon etching. As-implanted and annealed samples were compared to assess metal surface segregation.
2.2. Experimental techniques TXRF measurements were obtained by a RigakuTXRF300 instrument in direct mode, with the high energy beam at 0.05° angle and 500 s acquisition time for each point. The tool has a W filament working at 30 kV and 300 mA. 1 mm2 area Schottky diodes were measured by DLTS. A Semilab DLS-83D instrument was used. In this instrument, lock-in integration is used for averaging capacitance transients, and temperature can be scanned in the range 30–300 K. Alternatively, constant temperature spectra can be obtained as a function of the frequency of excitation pulses in the range 0.5 Hz–2 kHz [13]. Both methods were used in this work. The differential DLTS method was used. In this method, the Schottky diode (or the p–n junction) is reverse biased at a voltage Vr and two filling pulses are applied: the first pulse V1 is applied at the beginning of the lock-in integration period; and a second pulse V2 is applied a half period later. In the lock-in integration, the difference DC is obtained between the integrals of the capacitance transients caused by the first pulse and by the second pulse. This method yields the trap concentration Nt in the interval [xd(V1), xd(V2)], where xd is the depletion region edge at a given reverse voltage. In our measurements, samples were reverse biased at 5 V, and filling pulses with amplitudes of 0.5 V and 4.5 V were applied with a pulse width of 20 ls during each integration period. Under these conditions, a region ranging from about 0.8 lm to 2 lm was analyzed. The spectra shown in this paper were obtained with a 23 Hz filling pulse frequency. Schottky diodes were used for these measurements, so majority carrier traps could be revealed only. SPV measurements are carried out by illuminating the sample with light of various wavelengths. Generated minority carriers are collected in the depletion region at the wafer surface and produce a variation in surface potential, which is recorded as a function of light wavelength. In the standard SPV technique [14], the diffusion length Ldiff (or the recombination lifetime s) is extracted from these data by assuming that sample thickness is much larger than diffusion length. This hypothesis often fails in present wafers, and the actual wafer thickness must be taken into account. An ‘‘enhanced SPV’’ [15] technique is available for this purpose. In this work, enhanced SPV was generally used. Both standard and enhanced SPV are only sensitive to bulk recombination, so SPV cannot detect surface-segregated
166
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
Table 3 Flow for the preparation of the calibration wafers (a) and of the samples for the experiments about contamination by implantation in a contaminated implanter (b).
contaminants. In SPV measurements, the probed region is of the order of light penetration depth 1/a (where a is the absorption coefficient) plus carrier diffusion length. This region typically ranges from about 100 lm to the whole wafer thickness, so it is in any case much deeper than the region analyzed by DLTS. SPV measurements can also be associated with optical activation to quantify iron contamination by dissociation of the iron-boron pair [16]. In p-type silicon, the equilibrium state of iron at room temperature is the iron–boron pair. This pair can be dissociated by illumination, and the resulting interstitial metal is more effective as a recombination center than the iron-boron pair, so the carrier lifetime after dissociation (safter) is shorter than before dissociation (sbefore). The quantity 1=safter 1=sbefore is proportional to the iron concentration in the solid solution, so this method can be used to quantify the iron concentration. In addition, the concentration of recombination centers different from iron can also be estimated by assuming a somehow arbitrary capture cross section. In this work the SPV tool in the FAaST SDI system was used. This instrument is equipped with a set of filters producing monochromatic light at seven different wavelengths, ranging from 800 nm to 1 lm, and with a halogen lamp for optical activation. TOF-SIMS data were acquired by an ION TOF IV dual beam TOF-SIMS in the negative modality using a Ga+ primary ion beam operating at 25 keV, 2.2 pA, rastering over 100 100 lm2 area. Sputtering was accomplished by Cs+ at 250 eV energy and 10 nA current for 100 s and the raster was 300 300 lm2. This allows analyzing a region ranging from the surface to about 2 nm depth. This method was chosen in order to increase the WOx secondary ion yield. The integrated spectra are reported instead of the depth profiles to improve the signal-to-noise ratio.
3. Experimental results 3.1. Calibration of DLTS measurements 3.1.1. Molybdenum Fig. 1 reports the DLTS spectra of molybdenum-implanted wafers. One level is revealed only, located at Ev + 0.3 eV (where
Fig. 1. DLTS spectra in molybdenum-implanted p-type samples.
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
167
Fig. 2. Arrhenius plot of ep/T2 obtained in molybdenum-implanted samples. Literature data are shown for a comparison.
Ev is the valence band edge). Fig. 2 compares the Arrhenius plot of ep/T2 (where ep is the hole emission rate and T is the measurement absolute temperature) obtained from these samples with literature data [9] for molybdenum-contaminated p-type silicon. These data agree very well with each other, confirming that the peak reported in Fig. 1 can be identified with molybdenum. The molybdenum concentration measured by DLTS is reported as a function of the implanted dose QMo in Fig. 3, and is found to increase in proportion to the implanted dose, as expected. According to these data, by DLTS we obtain about 81012 cm 3 molybdenum concentration for 1010 cm 2 molybdenum dose.
3.1.2. Tungsten Fig. 4 reports the DLTS spectra of tungsten-implanted wafers. Two peaks can be identified, named H1 and H2. Fig. 5 compares the Arrhenius plots of ep/T2 related to these peaks with literature data [17] for tungsten-contaminated p-type silicon. This comparison allows us to identify H1 with the tungsten hole trap located at Ev + 0.4 eV. The level H2 is located at Ev + 0.36 eV, and does not correspond to any previously reported tungsten-related level [17,18]. The origin of H2 will be discussed in Section 3.3. No deep levels were revealed in n-type silicon, so we did not detect the electron trap reported in Ref. [17]. Fig. 6 reports the concentrations of H1 and H2 as a function of the tungsten dose QW. Here, solid and empty symbols refer to
Fig. 3. Molybdenum concentration measured by DLTS as a function of the molybdenum implanted dose.
Fig. 4. DLTS spectra of tungsten-implanted p-type wafers. These samples were treated by the same process flow as used for arsenic-implanted samples, including the RIE of silicon.
Fig. 5. Arrhenius plot of ep/T2 obtained in tungsten-implanted samples. Literature data are shown for a comparison.
two separate (but nominally identical) set of samples. The concentration of these levels increases in proportion to the tungsten dose up to 41010 cm 2, with about 30% variation between different sets of samples. In these conditions, if we consider the dominant peak H1 only, we obtain 5–6.51012 cm 3 tungsten concentration for 1010 cm 2 implanted dose. When the dose is increased above 51010 cm 2, a saturation is observed, followed by a reduction of the active tungsten concentration. These data show that tungsten deactivates at a concentration of about 2–31013 cm 3, possibly by surface segregation. ToF-SIMS analyses were carried out to confirm tungsten segregation at the wafer surface. Two wafers implanted with the highest dose in this test (1.51011 cm 2) were compared, an as-implanted wafer and a wafer annealed with the same thermal treatment as used for DLTS samples. ToF-SIMS analyses (Fig. 7) revealed tungsten at the wafer surface in the annealed wafer only, thus confirming that in this wafer tungsten segregated at the wafer surface during the thermal treatment. The concentration of impurities different from iron (‘‘Other impurities’’) obtained from SPV measurements shows some correlation with the tungsten dose, as shown in Fig. 8. However, the concentrations reported in Fig. 8 are quite low and close to the
168
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
Fig. 8. Concentration of impurities different from iron obtained from SPV measurements as a function of the tungsten dose.
Fig. 6. Trap concentration measured by DLTS as a function of tungsten dose. Solid and empty symbols refer to two sets of samples.
measurement sensitivity, so the SPV sensitivity to tungsten contamination is much lower than the DLTS sensitivity. This fact is not surprising, because the region involved in tungsten diffusivity is much thinner [9,10] than the region analyzed by SPV, which is of the order of the whole wafer thickness in our experimental conditions. 3.2. Contamination in implantation processes 3.2.1. Molybdenum contamination in BF2 and boron implanted silicon Fig. 9 shows the DLTS spectra of wafers implanted with 31014 cm 2 BF2 ions in implanters equipped with a molybdenum or a tungsten source chamber. In both wafers the molybdenum and the tungsten peak are revealed, however the molybdenum concentration is strongly reduced if the tungsten source is used. This fact is expected since mass interference of doubly ionized molybdenum in BF2 implantation is very well known [5], so it is expected that molybdenum from the source chamber may reach
Fig. 7. ToF-SIMS spectra of two samples implanted with 1.51011 cm dose. , as implanted, , annealed sample.
2
tungsten
the wafers. Tungsten contamination in BF2 implantation was previously observed [19] and ascribed to a mass interference mechanism, however the data in Fig. 9 show that the tungsten concentration is almost independent of the material used for the source chamber. Therefore, in our experiment mass interference is not likely to be the dominant mechanism for tungsten contamination. Molybdenum and tungsten concentration data were converted into dose data by using the calibration in Section 3.1. The so-obtained molybdenum and tungsten doses are reported as a function of the BF2 dose in Fig. 10, and are found to increase in proportion to the BF2 implanted dose. From these data, the molybdenum and tungsten contamination doses are a fraction of 2.610 5 and 0.610 6 respectively of the BF2 dose, if a molybdenum source is used. Fig. 11 shows the DLTS spectra of wafers implanted with 31015 cm 2 boron ions in implanters equipped with a molybdenum or a tungsten source chamber. Also in this case, both the molybdenum and the tungsten peak are revealed. As in BF2 implantation, the molybdenum concentration is strongly reduced by using a tungsten source. It was also shown that molybdenum
Fig. 9. DLTS spectra of wafers implanted with BF2 by using implanters equipped with molybdenum or tungsten sources.
169
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
Fig. 10. Molybdenum (a) and tungsten (b) contamination dose as a function of the BF2 dose, in an implanter equipped with a molybdenum source.
Fig. 11. DLTS spectra of wafers implanted with boron by using implanters equipped with molybdenum or tungsten sources.
Fig. 12. DLTS spectra of samples implanted with arsenic before contamination of the implanter, after tungsten contamination and after some decontamination processes.
mass interference can also affect boron implantations [7], so also in boron implantation molybdenum from the source may contaminate the implanted wafers. However in boron implantations the contamination mechanism is much less effective than in BF2 implantations, so the ratio between contaminant dose and implanted dose is lower in boron than in BF2 implantations. For what concerns tungsten, in boron implantation tungsten contamination seems to be sensitive to the source material, however the available data are not enough to formulate hypotheses about the origin of this contamination.
the wafer with very low energy [3,4]. Therefore, the tungsten dose estimated from the data in Fig. 12 can be compared with TXRF measurements of surface contamination after implantation. This comparison is reported in Table 4, and shows that TXRF detects tungsten contamination in the sample implanted after contamination of the implanter only. The lowest dose decontamination process is enough to reduce contamination below the TXRF detection limit, however DLTS measurements show that some
3.2.2. Tungsten contamination from a contaminated implanter Fig. 12 shows the DLTS spectra of arsenic-implanted samples (1015 cm 2, 60 keV) before contamination of the implanter, after contamination and after some decontamination processes. The same peaks as in tungsten implanted samples are observed. In Section 3.1.2 peak H1 was identified with tungsten and calibrated to obtain the tungsten dose, so the data in Fig. 12 can be used to estimate the tungsten dose diffused in silicon. In this experiment, contamination is mainly due to sputtering from parts close to the wafer during implantation. This mechanism is usually responsible for near-surface contamination, because contaminant atoms reach
Table 4 Tungsten concentration per unit area measured by TXRF and estimated from DLTS spectra, in samples implanted with arsenic before contamination of the implanter, after tungsten contamination and after some decontamination processes. Arsenic implantation conditions Before contamination After contamination, no decontamination Decontamination, 1016 cm 2 Decontamination, 21016 cm
2
[W]TXRF (1010 cm
2
)
QW,DLTS (1010 cm
2
3
<1.5 8.1 ± 0.8
<410 1.4 ± 0.1
<1.5 <1.5
0.87 ± 0.08 0.32 ± 0.03
)
170
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
tungsten contamination is still present in these wafer. On the other hand, the tungsten concentration at the surface measured by TXRF in the sample implanted after contamination of the implanter is higher than the datum estimated from DLTS spectra, possibly because of partial tungsten inactivation in this sample (see Fig. 6). It can be noted that in this example DLTS has excellent sensitivity per unit area. This sensitivity is obtained because DLTS analyzes a rather shallow region (2–3 lm deep), and the tungsten diffusion profile is limited to this region due to its low diffusivity. According to DLTS data, a very high arsenic dose is required to achieve an almost complete decontamination of the implanter.
3.3. Investigation about the H2 level The identification of the H2 level requires some more analysis. As shown in the previous sections, this level is observed both in tungsten-implanted wafers and in wafer contaminated with tungsten by arsenic implantation in a contaminated implanter. While level H1 can be identified with a well-known tungsten level, level H2 does not correspond to any tungsten-related level [17,18]. To confirm that H2 too is related to tungsten, in Fig. 13 we report the concentration of H2 as a function of the concentration of H1. Data from both tungsten-implanted samples and samples contaminated with tungsten by implantation in a contaminated implanter were included in this plot. The concentration of H2 increases in proportion to the concentration of H1, indicating that H2 too is related to tungsten contamination. To investigate whether the preparation used for these samples is involved in the formation of the level H2, some additional tungsten-implanted samples were prepared according to the same flow as in Table 3(a), with the exception that the silicon etching step was skipped. Fig. 14 shows the DLTS spectrum of a wafer implanted with 31010 cm 2 tungsten and not etched. In this spectrum, the level H1 only is found, showing that the level H2 is formed upon etching. On the other hand, Fig. 13 shows that H2 is related to tungsten, so H2 is probably a complex between tungsten and some species introduced by RIE. The wafers receive no thermal treatment after etching, and the maximum temperature reached in the metallization and masking process is about 100 °C for a few minutes. Tungsten is a slow diffuser, and cannot diffuse at such low temperatures. For this reason, the species introduced by RIE must be a very fast diffuser, which can diffuse through the silicon lattice and reach tungsten atoms even at the low temperatures involved in this part of the sample preparation flow.
Fig. 14. DLTS spectrum of a sample implanted with 31010 cm annealed, but not etched.
2
tungsten dose and
Fig. 15. DLTS spectrum of a sample implanted with 31010 cm annealed, before and after 250 °C 45 min baking.
2
tungsten dose and
The effect of a low temperature thermal treatment on the level H2 was also studied. Fig. 15 reports the DLTS spectra of a sample implanted with 31010 cm 2 tungsten, RTP annealed and etched, before and after 45 min baking at 250 °C. The baking treatment suppresses the H2 level, indicating that the H2 level can be easily destroyed by a moderate temperature treatment. In addition, the baking treatment increases the concentration of H1 up to the concentration measured in the not etched sample. A further thermal treatment (up to 5 h) at the same temperature yields no further modification of the DLTS spectrum. Chlorine and HBr are used for RIE silicon etching. Hydrogen is a very fast diffuser and easily forms complexes with many metals, so it is a possible candidate as the responsible for the formation of the level H2. However, our data are not enough for a positive identification of H2. 4. Conclusions
Fig. 13. Concentration of the level H2 as a function of the concentration of the level H1 in samples implanted with tungsten or contaminated with tungsten by arsenic implantation in a contaminated implanter.
A procedure for monitoring molybdenum and tungsten contamination in ion implantation processes by DLTS was set up and calibrated by using molybdenum and tungsten implanted samples.
M.L. Polignano et al. / Nuclear Instruments and Methods in Physics Research B 356–357 (2015) 164–171
The molybdenum concentration measured by DLTS was found to be proportional to the molybdenum dose over the whole range under study (up to 1011 cm 2). In molybdenum-implanted samples, the molybdenum level at Ev + 0.3 eV is revealed only. In tungsten-implanted samples, two levels are revealed, named H1 and H2. The level H1 is the tungsten-related hole trap, the level H2 does not correspond to any tungsten-related level, however it is related to tungsten and it is induced by the silicon dry etching included in the flow for monitoring contamination from ion implantation. The SPV sensitivity to tungsten contamination was also tested, and it was found much lower than the DLTS sensitivity, due to the low tungsten diffusivity. The concentration measured by DLTS saturates and then decreases starting from 51010 cm 2 tungsten dose. ToF-SIMS analyses revealed tungsten at the wafer surface, indicating that the observed tungsten inactivation is due to surface segregation. This procedure was used to evaluate contamination in implantation processes. In BF2 implantations, in addition to molybdenum, tungsten contamination is found. The molybdenum contamination in BF2 implantation is a well-known example of mass interference, and as expected the molybdenum concentration strongly depends on the material of the source chamber. Vice versa the tungsten contamination is roughly independent of the source material. Molybdenum and tungsten contamination is found in boron implantation too. The tungsten contamination induced by implantation in a previously contaminated implanter was quantified, and the efficiency of arsenic implantation as a decontamination process was tested. Finally, it was shown that TXRF is much less sensitive than DLTS for monitoring tungsten contamination.
References [1] F.A. Stevie, R.G. Wilson, D.S. Simons, M.I. Current, P.C. Zalm, Review of secondary ion mass spectroscopy characterization of contamination associated with ion implantation, J. Vac. Sci. Technol., B 12 (1994) 2263. [2] H. Ryssel, L. Frey, Contamination problems in ion implantation, in: J.F. Ziegler (Ed.), Handbook of Ion Implantation Technology, Elsevier Science Publisher B.V., Amsterdam, 1992, p. 675.
171
[3] M.L. Polignano, C. Bresolin, F. Cazzaniga, A. Sabbadini, G. Queirolo, Investigation of metal contamination by photocurrent measurements: validation and application to ion implantation processes in: J.K. Lowell, R.T. Chen, J.P. Mathur (Eds.), Optical Characterization Techniques for High Performance Microelectronic Device Manufacturing II, Proc. SPIE 2638, 1995, p. 14. [4] M.L. Polignano, D. Caputo, A. Giussani, V. Soncini, G. Di Toma, Metal Contamination Reduction in the Evolution of Ion Implantation Technology, in: Proc. of the XIII International Conference on Ion Implantation Technology, IEEE, 2000, p. 686. [5] A. Cubina, M. Frost, Effects of molybdenum contamination resulting from BF2 implantation, Nucl. Instr. Methods Phys. Res. B55 (1991) 160. [6] D. Codegoni, M.L. Polignano, V. Soncini, C. Bresolin, Molybdenum contamination in indium and boron implantation processes, Electrochem. Soc. Trans. 10 (2007) 85. [7] V. Haublein, L. Frey, H. Ryssel, Additional peaks in mass spectra due to charge exchange events and dissociation of molecular ions during extraction, Nucl. Instr. Methods Phys. Res., Sect. B 237 (2005) 346. [8] F. Russo, G. Moccia, G. Nardone, R. Alfonsetti, G. Polsinelli, A. D’Angelo, A. Patacchiola, M. Liverani, P. Pianezza, T. Lippa, M. Carlini, M.L. Polignano, I. Mica, E. Cazzini, M. Ceresoli, D. Codegoni, Proximity gettering of slow diffuser contaminants in CMOS image sensors, Solid-State Electron. 91 (2014) 91. [9] J.L. Benton, D.C. Jacobson, B. Jackson, J.A. Johnson, T. Boone, D.J. Eaglesham, F.A. Stevie, J. Becerro, Behavior of molybdenum in silicon evaluated for integrated circuit processing, J. Electrochem. Soc. 146 (1999) 1929. [10] K. Graff, Metal Impurities in Silicon-Device Fabrication, Springer, Berlin, 2000. ch. 4 and 5. [11] A. De Luca, A. Portavoce, M. Texier, C. Grosjean, N. Burle, V. Oison, B. Pichaud, Tungsten diffusion in silicon, J. Appl. Phys. 115 (2014) 013501. [12] M.L. Polignano, D. Codegoni, S. Grasso, I. Mica, G. Borionetti, A. Nutsch, A comparative analysis of different measurement techniques to monitor metal and organic contamination in silicon device processing, Phys. Status Solidi A 212 (2015) 495. [13] G. Ferenczi, C.A. Londos, T. Pavelka, M. Somogyi, A. Mertend, Correlation of the concentration of the carbon-associated radiation damage levels with the total carbon concentration in silicon, J. Appl. Phys. 63 (1988) 183. [14] A.M. Goodman, A method for the measurement of short minority carrier diffusion lengths in semiconductors, J. Appl. Phys. 32 (1961) 2550. [15] J. Lagowski, A.M. Kontkiewicz, L. Jastrzebski, P. Edelman, Method for the measurement of long minority carrier diffusion lengths exceeding wafer thickness, Appl. Phys. Lett. 63 (1993) 2902. [16] G. Zoth, W. Bergholz, A fast, preparation-free method to detect iron in silicon, J. Appl. Phys. 67 (11) (1990) 6764. [17] Y. Fujisaki, T. Ando, H. Kozuka, Y. Takano, Characterization of tungsten-related deep levels in bulk silicon crystal, J. Appl. Phys. 63 (1988) 2304. [18] S. Boughaba, D. Mathiot, Deep level transient spectroscopy characterization of tungsten-related deep levels in silicon, J. Appl. Phys. 69 (1) (1991) 278. [19] R.B. Liebert, G.C. Angel and M. Kase, Tungsten Contamination in BF2 Implants, in: Proc. of the 11th International Conference on Ion Implantation Technology, IEEE, 1997, p. 135.