- Email: [email protected]
The thermal stability of iron precipitates in silicon after internal gettering
ARTICLE IN PRESS
Physica B 340–342 (2003) 1051–1055
The thermal stability of iron precipitates in silicon after internal gettering . aa,b, Andrei A. Istratova, Eicke R. Webera Peng Zhanga,*, Hele V.ainol. a
Department of Materials Science and Engineering, University of California at Berkeley and Lawrence Berkeley National Laboratory, MS 62R0203, 1 Cyclotron Road, Berkeley CA 94720, USA b Helsinki University of Technology, Electron Physics Laboratory, P.O.BOX 3500, FIN-02015 HUT, Finland
Abstract The re-dissolution behavior of iron gettered at oxide precipitates serving as internal gettering sites was studied in the temperature range from 750 C to 900 C using p-type Czochralski-grown silicon with a boron doping level of 2.5 1014 cm3 and oxide precipitates density of 5 109 cm3. The concentration of interstitial iron and iron-boron pairs was measured by deep-level transient spectroscopy. It was found that the re-dissolved iron concentration can be fitted by an exponential dependence, and the dissolution rate t1 had an Arrhenius-type dependence on temperature. The activation energy of the dissolution time constant was found to be 1.47 eV. Considering the activation energy of 0.67 eV for iron diffusion in silicon, our result implies an effective binding energy of 0.80 eV between gettered iron impurities and oxide precipitates. The re-dissolved iron concentration during different heat treatments was simulated with our gettering simulator taking into account this effective binding energy. r 2003 Published by Elsevier B.V. PACS: 71.55.Cn; 61.72.Yx Keywords: Internal gettering; Re-dissolution energy barrier; Iron; Silicon
1. Introduction It is well known that transition metal impurities dissolved in the silicon matrix introduce deep electronic levels in the band gap and detrimentally impact the device yield. Among transition metals, iron is one of the most ubiquitous and harmful metal impurities in silicon Integrated Circuit (IC) manufacturing [1,2]. As the device size shrinks into sub-micron scale, the Semiconductor Industry Association Roadmap specifies iron contamina*Corresponding author. Fax: 1-510-486-4995. E-mail address: [email protected] (P. Zhang). 0921-4526/$ - see front matter r 2003 Published by Elsevier B.V. doi:10.1016/j.physb.2003.09.205
tion levels o1 1010 cm3 by the year 2005 [3]. In order to minimize the impact of unintentional contamination on device yield, gettering has been widely used in semiconductor processing [4,5] to remove metal impurities from the device region. Among different gettering techniques, internal gettering [6–8] of metals by oxide precipitates formed in the bulk of silicon wafers was shown to be very efficient for fast diffusers such as iron, and could be implemented with minimal additional costs. During the wafer cooling, supersaturated iron precipitates at heterogeneous nucleation sites offered by oxide precipitates. Therefore, it diffuses out of the denuded zone, where the density of
ARTICLE IN PRESS 1052
P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
precipitation sites is low [7]. The precipitation of iron at oxide precipitates can be theoretically described by the model of diffusion-limited relaxation gettering, mathematically described by Ham’s law [9]. Several reports have shown that the gettered iron can be completely re-dissolved within several minutes if the wafer is heated to a sufficiently high temperature [10–12] after the gettering treatment. The re-dissolution of iron limits the stability of internal gettering. However, very little is known about the kinetics of this dissolution process. It is not clear whether there is a binding energy between the oxide precipitate and iron impurities which would slow down the dissolution of the gettered iron during short anneals at moderate temperatures, inherent in rapid thermal processing (RTP). The stability of gettered iron at oxide precipitates is an important factor for a predictive simulation of optimized gettering. Indeed, if gettered iron is completely re-dissolved during heating within several seconds, only the last cooling in the sequence of heat treatments used for device fabrication has to be optimized to obtain the highest possible gettering efficiency. Besides, the re-dissolved iron will be free to diffuse to device regions and may be collected by heavily doped device areas, which offer a strong competition to the gettering sites in the substrate [13]. In contrast, if there is a dissolution barrier for the gettered iron, the device yield can be significantly increased by the optimization of each processing step. In this paper, we report the kinetics of dissolution of iron gettered at oxide precipitates in p-type Czochralski grown (CZ) silicon as a function of temperature. This approach allowed us to determine the effective energy barrier for redissolution of the gettered iron. The experimental data are used together with gettering simulations to predict how the re-dissolution process affects the gettering stability during RTP and temperature ramping in a conventional furnace.
2.5 1014 cm3. The wafers had an oxide precipitate density of 5 109 cm3 obtained through a sequence of heat treatments [14], which included a nucleation treatment at 650 C for 14 h, followed by a stabilization process at 800 C for 4 h and a growth treatment at 1000 C for 16 h. The samples were intentionally contaminated by evaporating a thin film (B30 nm) of iron on the back side of the sample, followed by a drive-in annealing of 50 min at 950 C in an ambient of 96% N2+4% H2. After that, the set point of the furnace was changed to cool down the samples to 700oC. The average cooling rate of the furnace in this temperature range was approximately 14 C/min. Then the samples were kept at 700 C for 30 min, after which the temperature was decreased to 450 C during the next 30 min. This gettering anneal was terminated by quenching the samples to room temperature by dropping them on an aluminum plate. The gettering efficiency was tested by deeplevel transient spectroscopy (DLTS), which confirmed that the concentration of ungettered interstitial iron was about 5 1010 cm3, which was slightly above the noise level in the DLTS spectra. After etching off approximately 50 mm from the surface by a HNO3+HF+H2O solution, the samples were annealed at 750 C, 800 C, 850 C, or 900 C for different times between several seconds and several minutes (re-dissolution annealing) in the same gas mixture as mentioned above. The dissolution annealing was terminated by quenching the samples on an aluminum plate. The top layer was again removed chemically, after which the aluminum Schottky diodes were deposited by thermal evaporation, and the free iron (interstitial iron (Fei) and iron–boron pair (Fe–B)) concentration in the samples were analyzed by DLTS measurements.
3. Results and discussion 3.1. Binding energy
2. Experimental The samples were cut from p-type 200 mm CZ silicon wafers with a boron doping level of
Fig. 1(a) shows typical DLTS spectra, obtained after iron re-dissolution anneal of the samples at 800 C for different times, from several seconds to several minutes. The peaks around 56 and 255 K
ARTICLE IN PRESS P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
0.0
25 s 70 s 120 s 360 s
1.0 1.5 2.0 2.5 40
(a)
70
100 130 160 190 220 250 280
Temperature (K)
-3
[Fe-B] +[ Fei] (cm )
10
13
(b)
10
12
10
11
0
100 200 300 400 500 600 700 800 900
sponds to the average inter-precipitates distance of 5.8 mm. Measurements of dissolution kinetics of precipitated iron at different annealing temperatures (shown in Fig. 2) revealed an Arrhenius-type dependence of t1 on annealing temperature T: t1 ¼ 4:01 104 expðEA =kB TÞ½s1
ð1Þ
with the energy barrier for iron dissolution of EA ¼ 1:4770:10 eV; significantly larger than the activation energy of ED ¼ 0:67 eV [1] for Fe diffusion in silicon. Hence, our data imply a significant binding between gettered iron and oxide precipitates. It should be noted that past experiments indicated that there is an influence of the carbon content on the stability of gettered iron: in Si with a high carbon content re-dissolution was found to be faster than in low-carbon Si [12]. However, the analysis of the impact of carbon concentration is beyond the scope of this study. 3.2. Simulations
Time (s)
Fig. 1. (a) Typical DLTS spectra showing dissolved iron, mainly in the form of Fe–B pairs, at different annealing times. (b) The dependence of the dissolved iron concentration on annealing times at T=800 C.
represent Fe–B pairs and Fei, respectively [15]. Fig. 1(b) shows the dissolved iron concentration vs. dissolution annealing time at 800 C. The experimental data points were fitted with an exponential function: CðtÞ ¼ C0 ½1 expðt=tÞ; where t ¼ ð198783Þs and C0 ¼ 4:58 1012 cm3 ; the solubility of iron in silicon at this temperature [1]. Note that the dissolution time constant of 198 s at 800 C, although short, is much greater than a prediction based on the assumption that there is no dissolution barrier and that the re-dissolution process is only limited by iron diffusivity and solubility in silicon. It is easy to show that if there were no dissolution barrier, the gettered iron would re-dissolve in less than 10 s at the annealing temperature of 800 C in CZ silicon with the oxide precipitates density 5 109 cm3, which corre-
We used the experimental data obtained above to perform numerical simulation of gettering and re-dissolution of iron during different heating cycles. The gettering simulator used in this study was based on the algorithm described in detail in Ref. [16]. Ham’s law [9] was used to describe diffusion-limited precipitation of iron. The
Temperature (K)
Dissolution time constant τ (s)
DLTS signal (a.u.)
0.5
1053
10
3
10
2
1173
1123
1073
τ −1 = 4.01 × 10 4 × exp( − 10
1023
1.47 ± 0.10 eV)s −1 k BT
1
0.84 0.86 0.88 0.90 0.92 0.94 0.96 0.98 1.00
1000/T (K-1) Fig. 2. Arrhenius plot of t1 vs. the reciprocal annealing temperature.
ARTICLE IN PRESS P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
1054
dissolution process was described by the following equation: CðtÞ ¼ Cðt DtÞ þ ½SðtÞ Cðt DtÞ ½1 expðDt=t;
ð2Þ
where C is the dissolved iron concentration, S is the solubility and dissolution time constant t is given by Eq. (1). Fig. 3 shows examples of how iron is predicted to behave during a short RTP treatment at different temperatures with and without the dissolution barrier. In this series of simulation, the total amount of available iron in the silicon wafer is assumed to be 1.0 1011 cm3, the typical accidental contamination level during IC fabrica10
12
(a)
10
11
10
10
10
13
10
12
10
11
10
10
10
15
10
14
10
13
10
12
10
11
10
10
Concentration (cm-3 )
(b)
(c)
0
10
20
30
40
50
60
70
80
Time (s) Fig. 3. Dissolved iron concentration during 30 s RTP anneal at three different temperatures (a) 750 C, (b) 900 C and (c) 1050 C simulated with (solid line) and without (dashed line) binding energy. The dotted line presents the equilibrium solubility during the annealing treatment. The total available amount of iron concentration is assumed to be 1 1011 cm3.
tion. The initial condition is that all the supersaturated iron impurities are gettered, i.e. the dissolved iron concentration equals the solubility 2.59 109 cm3 at the starting temperature of 600 C. The ramp rate is 25 C/s from 600 C to 750–1100 C and the length of the anneal time is 30 s. As we can see from Fig. 3(a), if there were no dissolution barrier the dissolved iron closely follows the solubility and almost instantly reaches the level of 1.0 1011 cm3, whereas the dissolved iron concentration calculated with the dissolution barrier does not exceed 5 1010 cm3 by the end of the 700 C, 30 s anneal. At a higher temperature, for example 900 C (Fig. 3(b)) and 1050 C (Fig. 3(c)), the interstitial iron concentration reaches 1 1011 cm3 (i.e., the precipitated iron dissolves completely) already during the temperature ramping. Thus at high-temperature annealing and low-iron contamination levels, the existence of the dissolution barrier has little or no impact on optimization of the gettering procedure. However, in laboratory experiments, which frequently use relatively high iron contamination levels (as high as 1014 cm3), the dissolution barrier has a significant impact on the total dissolved iron concentration observed after the annealing [17]. In another series of simulations, the same contamination level and initial condition are considered. We simulated the re-dissolved iron concentration during different temperature ramping (heating from 600 C to 850 C), with the ramping rate ranging from 1 to 1500 C/min. As we can see from Fig. 4, the faster the ramping, the lower the dissolved iron concentration is at the same temperature. The difference, which can be as high as one order of magnitude, stems from the fact that the annealing (dissolution) time is shorter for faster ramping. But when contamination is as low as 1011 cm3, for the ramping rates we simulated, all the iron impurities re-dissolve into the silicon matrix before the temperature reaches 850 C. Hence, although there is a significant binding between gettered iron and oxide precipitates, the binding energy has little impact on IC processing once the processing temperature is higher than 850 C, especially when the iron contamination level is low.
ARTICLE IN PRESS
-3
Concentration (cm )
P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
10
13
10
12
10
11
10
10
Acknowledgements This work was supported by SiWEDS (the NSFI/UCRC Si Wafer Engineering and Defect Science consortium (www.siweds.org)). H.V. acknowledges support by Finnish National Technology Agency, Academy of Finland, Okmetic Oyj, Micro Analog Systems Oy, and VTI Technologies Oy. The authors acknowledge Dr. R. Falster from MEMC for providing the silicon wafers. Part of the work was carried out in the UC Berkeley Microfabrication Laboratory.
(3) (2) (1)
10
9
600
1055
650
700
750
800
850
Temperature (oC)
Fig. 4. Dissolved iron concentration during temperature ramping from 600 C to 850 C with ramping rates: (1) 1 C/ min, (2) 60 C/min and (3) 1500 C/min. The total available amount of iron concentration is assumed to be 1 1011 cm3. The dashed line presents the equilibrium solubility during the annealing treatment.
4. Conclusions The analysis of the dissolution kinetics of iron gettered by oxide precipitates in Cz-silicon revealed that there is a strong binding between the gettered iron and oxide precipitates. The effective dissolution energy barrier was determined to be EA ¼ 1:4770:10 eV; as determined in the temperature range from 750 C to 900 C. Considering the diffusion barrier of 0.67 eV for iron diffusion in silicon, our results give an effective binding energy of 0.80 eV between oxide precipitates and the gettered iron impurities. The existence of this binding energy may be used to improve the efficiency of internal gettering during rapid thermal annealing, particularly at temperatures below 800 C or at high-iron contamination levels. But for a high processing temperature (as high as 850 C) and low-iron contamination level (o1011 cm3), the impact of the dissolution barrier is negligible and it is still true that only the last cooling procedure has to be adjusted to optimize internal gettering.
References [1] A.A. Istratov, H. Hieslmair, E.R. Weber, Appl. Phys. A 69 (1999) 13. [2] A.A. Istratov, H. Hieslmair, E.R. Weber, Appl. Phys. A 70 (2000) 489. [3] SIA, International technology roadmap for semiconductors 2002, http://public.itrs.net or http://www.sematech.org. . [4] S.M. Myers, M. Seibt, W. Schroter, J. Appl. Phys. 88 (2000) 3795. [5] A.A. Istratov, H. Hieslmair, E.R. Weber, MRS Bulletin 25 (2000) 33. [6] T.Y. Tan, E.E. Garnder, W.K. Tice, Appl. Phys. Lett. 30 (1977) 175. [7] D. Gilles, E.R. Weber, S.K. Hahn, Phys. Rev. Lett. 64 (1990) 196. [8] M. Aoki, A. Hara, A. Ohsawa, J. Appl. Phys. 72 (1992) 895. [9] F.S. Ham, J. Phys. Chem. Solids 6 (1958) 335. [10] E.G. Colas, E.R. Weber, Appl. Phys. Lett. 48 (1986) 1371. [11] M. Aoki, A. Hara, J. Appl. Phys. 74 (1993) 1440. [12] S.A. McHugo, E.R. Weber, M. Mizuno, F.G. Kirscht, Appl. Phys. Lett. 66 (1995) 2840. [13] A.A. Istratov, W. Huber, E.R. Weber, J. Electrochem. Soc. 150 (2003) G244. [14] K.F. Kelton, R. Falster, D. Gambaro, M. Olma, M. Cornara, P.F. Wei, J. Appl. Phys. 85 (1999) 8097. [15] S.D. Brotherton, P. Bradley, A. Gill, J. Appl. Phys. 57 (1985) 1941. [16] H. Hieslmair, S. Balasubramanian, A.A. Istratov, E.R. Weber, Semicond. Sci. Technol. 15 (2001) 567. . a, A.A. Istratov, E.R. Weber, Appl. [17] P. Zhang, H. V.ainol. Phys. Lett. 83 (2003) 4324.
Physica B 340–342 (2003) 1051–1055
The thermal stability of iron precipitates in silicon after internal gettering . aa,b, Andrei A. Istratova, Eicke R. Webera Peng Zhanga,*, Hele V.ainol. a
Department of Materials Science and Engineering, University of California at Berkeley and Lawrence Berkeley National Laboratory, MS 62R0203, 1 Cyclotron Road, Berkeley CA 94720, USA b Helsinki University of Technology, Electron Physics Laboratory, P.O.BOX 3500, FIN-02015 HUT, Finland
Abstract The re-dissolution behavior of iron gettered at oxide precipitates serving as internal gettering sites was studied in the temperature range from 750 C to 900 C using p-type Czochralski-grown silicon with a boron doping level of 2.5 1014 cm3 and oxide precipitates density of 5 109 cm3. The concentration of interstitial iron and iron-boron pairs was measured by deep-level transient spectroscopy. It was found that the re-dissolved iron concentration can be fitted by an exponential dependence, and the dissolution rate t1 had an Arrhenius-type dependence on temperature. The activation energy of the dissolution time constant was found to be 1.47 eV. Considering the activation energy of 0.67 eV for iron diffusion in silicon, our result implies an effective binding energy of 0.80 eV between gettered iron impurities and oxide precipitates. The re-dissolved iron concentration during different heat treatments was simulated with our gettering simulator taking into account this effective binding energy. r 2003 Published by Elsevier B.V. PACS: 71.55.Cn; 61.72.Yx Keywords: Internal gettering; Re-dissolution energy barrier; Iron; Silicon
1. Introduction It is well known that transition metal impurities dissolved in the silicon matrix introduce deep electronic levels in the band gap and detrimentally impact the device yield. Among transition metals, iron is one of the most ubiquitous and harmful metal impurities in silicon Integrated Circuit (IC) manufacturing [1,2]. As the device size shrinks into sub-micron scale, the Semiconductor Industry Association Roadmap specifies iron contamina*Corresponding author. Fax: 1-510-486-4995. E-mail address: [email protected] (P. Zhang). 0921-4526/$ - see front matter r 2003 Published by Elsevier B.V. doi:10.1016/j.physb.2003.09.205
tion levels o1 1010 cm3 by the year 2005 [3]. In order to minimize the impact of unintentional contamination on device yield, gettering has been widely used in semiconductor processing [4,5] to remove metal impurities from the device region. Among different gettering techniques, internal gettering [6–8] of metals by oxide precipitates formed in the bulk of silicon wafers was shown to be very efficient for fast diffusers such as iron, and could be implemented with minimal additional costs. During the wafer cooling, supersaturated iron precipitates at heterogeneous nucleation sites offered by oxide precipitates. Therefore, it diffuses out of the denuded zone, where the density of
ARTICLE IN PRESS 1052
P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
precipitation sites is low [7]. The precipitation of iron at oxide precipitates can be theoretically described by the model of diffusion-limited relaxation gettering, mathematically described by Ham’s law [9]. Several reports have shown that the gettered iron can be completely re-dissolved within several minutes if the wafer is heated to a sufficiently high temperature [10–12] after the gettering treatment. The re-dissolution of iron limits the stability of internal gettering. However, very little is known about the kinetics of this dissolution process. It is not clear whether there is a binding energy between the oxide precipitate and iron impurities which would slow down the dissolution of the gettered iron during short anneals at moderate temperatures, inherent in rapid thermal processing (RTP). The stability of gettered iron at oxide precipitates is an important factor for a predictive simulation of optimized gettering. Indeed, if gettered iron is completely re-dissolved during heating within several seconds, only the last cooling in the sequence of heat treatments used for device fabrication has to be optimized to obtain the highest possible gettering efficiency. Besides, the re-dissolved iron will be free to diffuse to device regions and may be collected by heavily doped device areas, which offer a strong competition to the gettering sites in the substrate [13]. In contrast, if there is a dissolution barrier for the gettered iron, the device yield can be significantly increased by the optimization of each processing step. In this paper, we report the kinetics of dissolution of iron gettered at oxide precipitates in p-type Czochralski grown (CZ) silicon as a function of temperature. This approach allowed us to determine the effective energy barrier for redissolution of the gettered iron. The experimental data are used together with gettering simulations to predict how the re-dissolution process affects the gettering stability during RTP and temperature ramping in a conventional furnace.
2.5 1014 cm3. The wafers had an oxide precipitate density of 5 109 cm3 obtained through a sequence of heat treatments [14], which included a nucleation treatment at 650 C for 14 h, followed by a stabilization process at 800 C for 4 h and a growth treatment at 1000 C for 16 h. The samples were intentionally contaminated by evaporating a thin film (B30 nm) of iron on the back side of the sample, followed by a drive-in annealing of 50 min at 950 C in an ambient of 96% N2+4% H2. After that, the set point of the furnace was changed to cool down the samples to 700oC. The average cooling rate of the furnace in this temperature range was approximately 14 C/min. Then the samples were kept at 700 C for 30 min, after which the temperature was decreased to 450 C during the next 30 min. This gettering anneal was terminated by quenching the samples to room temperature by dropping them on an aluminum plate. The gettering efficiency was tested by deeplevel transient spectroscopy (DLTS), which confirmed that the concentration of ungettered interstitial iron was about 5 1010 cm3, which was slightly above the noise level in the DLTS spectra. After etching off approximately 50 mm from the surface by a HNO3+HF+H2O solution, the samples were annealed at 750 C, 800 C, 850 C, or 900 C for different times between several seconds and several minutes (re-dissolution annealing) in the same gas mixture as mentioned above. The dissolution annealing was terminated by quenching the samples on an aluminum plate. The top layer was again removed chemically, after which the aluminum Schottky diodes were deposited by thermal evaporation, and the free iron (interstitial iron (Fei) and iron–boron pair (Fe–B)) concentration in the samples were analyzed by DLTS measurements.
3. Results and discussion 3.1. Binding energy
2. Experimental The samples were cut from p-type 200 mm CZ silicon wafers with a boron doping level of
Fig. 1(a) shows typical DLTS spectra, obtained after iron re-dissolution anneal of the samples at 800 C for different times, from several seconds to several minutes. The peaks around 56 and 255 K
ARTICLE IN PRESS P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
0.0
25 s 70 s 120 s 360 s
1.0 1.5 2.0 2.5 40
(a)
70
100 130 160 190 220 250 280
Temperature (K)
-3
[Fe-B] +[ Fei] (cm )
10
13
(b)
10
12
10
11
0
100 200 300 400 500 600 700 800 900
sponds to the average inter-precipitates distance of 5.8 mm. Measurements of dissolution kinetics of precipitated iron at different annealing temperatures (shown in Fig. 2) revealed an Arrhenius-type dependence of t1 on annealing temperature T: t1 ¼ 4:01 104 expðEA =kB TÞ½s1
ð1Þ
with the energy barrier for iron dissolution of EA ¼ 1:4770:10 eV; significantly larger than the activation energy of ED ¼ 0:67 eV [1] for Fe diffusion in silicon. Hence, our data imply a significant binding between gettered iron and oxide precipitates. It should be noted that past experiments indicated that there is an influence of the carbon content on the stability of gettered iron: in Si with a high carbon content re-dissolution was found to be faster than in low-carbon Si [12]. However, the analysis of the impact of carbon concentration is beyond the scope of this study. 3.2. Simulations
Time (s)
Fig. 1. (a) Typical DLTS spectra showing dissolved iron, mainly in the form of Fe–B pairs, at different annealing times. (b) The dependence of the dissolved iron concentration on annealing times at T=800 C.
represent Fe–B pairs and Fei, respectively [15]. Fig. 1(b) shows the dissolved iron concentration vs. dissolution annealing time at 800 C. The experimental data points were fitted with an exponential function: CðtÞ ¼ C0 ½1 expðt=tÞ; where t ¼ ð198783Þs and C0 ¼ 4:58 1012 cm3 ; the solubility of iron in silicon at this temperature [1]. Note that the dissolution time constant of 198 s at 800 C, although short, is much greater than a prediction based on the assumption that there is no dissolution barrier and that the re-dissolution process is only limited by iron diffusivity and solubility in silicon. It is easy to show that if there were no dissolution barrier, the gettered iron would re-dissolve in less than 10 s at the annealing temperature of 800 C in CZ silicon with the oxide precipitates density 5 109 cm3, which corre-
We used the experimental data obtained above to perform numerical simulation of gettering and re-dissolution of iron during different heating cycles. The gettering simulator used in this study was based on the algorithm described in detail in Ref. [16]. Ham’s law [9] was used to describe diffusion-limited precipitation of iron. The
Temperature (K)
Dissolution time constant τ (s)
DLTS signal (a.u.)
0.5
1053
10
3
10
2
1173
1123
1073
τ −1 = 4.01 × 10 4 × exp( − 10
1023
1.47 ± 0.10 eV)s −1 k BT
1
0.84 0.86 0.88 0.90 0.92 0.94 0.96 0.98 1.00
1000/T (K-1) Fig. 2. Arrhenius plot of t1 vs. the reciprocal annealing temperature.
ARTICLE IN PRESS P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
1054
dissolution process was described by the following equation: CðtÞ ¼ Cðt DtÞ þ ½SðtÞ Cðt DtÞ ½1 expðDt=t;
ð2Þ
where C is the dissolved iron concentration, S is the solubility and dissolution time constant t is given by Eq. (1). Fig. 3 shows examples of how iron is predicted to behave during a short RTP treatment at different temperatures with and without the dissolution barrier. In this series of simulation, the total amount of available iron in the silicon wafer is assumed to be 1.0 1011 cm3, the typical accidental contamination level during IC fabrica10
12
(a)
10
11
10
10
10
13
10
12
10
11
10
10
10
15
10
14
10
13
10
12
10
11
10
10
Concentration (cm-3 )
(b)
(c)
0
10
20
30
40
50
60
70
80
Time (s) Fig. 3. Dissolved iron concentration during 30 s RTP anneal at three different temperatures (a) 750 C, (b) 900 C and (c) 1050 C simulated with (solid line) and without (dashed line) binding energy. The dotted line presents the equilibrium solubility during the annealing treatment. The total available amount of iron concentration is assumed to be 1 1011 cm3.
tion. The initial condition is that all the supersaturated iron impurities are gettered, i.e. the dissolved iron concentration equals the solubility 2.59 109 cm3 at the starting temperature of 600 C. The ramp rate is 25 C/s from 600 C to 750–1100 C and the length of the anneal time is 30 s. As we can see from Fig. 3(a), if there were no dissolution barrier the dissolved iron closely follows the solubility and almost instantly reaches the level of 1.0 1011 cm3, whereas the dissolved iron concentration calculated with the dissolution barrier does not exceed 5 1010 cm3 by the end of the 700 C, 30 s anneal. At a higher temperature, for example 900 C (Fig. 3(b)) and 1050 C (Fig. 3(c)), the interstitial iron concentration reaches 1 1011 cm3 (i.e., the precipitated iron dissolves completely) already during the temperature ramping. Thus at high-temperature annealing and low-iron contamination levels, the existence of the dissolution barrier has little or no impact on optimization of the gettering procedure. However, in laboratory experiments, which frequently use relatively high iron contamination levels (as high as 1014 cm3), the dissolution barrier has a significant impact on the total dissolved iron concentration observed after the annealing [17]. In another series of simulations, the same contamination level and initial condition are considered. We simulated the re-dissolved iron concentration during different temperature ramping (heating from 600 C to 850 C), with the ramping rate ranging from 1 to 1500 C/min. As we can see from Fig. 4, the faster the ramping, the lower the dissolved iron concentration is at the same temperature. The difference, which can be as high as one order of magnitude, stems from the fact that the annealing (dissolution) time is shorter for faster ramping. But when contamination is as low as 1011 cm3, for the ramping rates we simulated, all the iron impurities re-dissolve into the silicon matrix before the temperature reaches 850 C. Hence, although there is a significant binding between gettered iron and oxide precipitates, the binding energy has little impact on IC processing once the processing temperature is higher than 850 C, especially when the iron contamination level is low.
ARTICLE IN PRESS
-3
Concentration (cm )
P. Zhang et al. / Physica B 340–342 (2003) 1051–1055
10
13
10
12
10
11
10
10
Acknowledgements This work was supported by SiWEDS (the NSFI/UCRC Si Wafer Engineering and Defect Science consortium (www.siweds.org)). H.V. acknowledges support by Finnish National Technology Agency, Academy of Finland, Okmetic Oyj, Micro Analog Systems Oy, and VTI Technologies Oy. The authors acknowledge Dr. R. Falster from MEMC for providing the silicon wafers. Part of the work was carried out in the UC Berkeley Microfabrication Laboratory.
(3) (2) (1)
10
9
600
1055
650
700
750
800
850
Temperature (oC)
Fig. 4. Dissolved iron concentration during temperature ramping from 600 C to 850 C with ramping rates: (1) 1 C/ min, (2) 60 C/min and (3) 1500 C/min. The total available amount of iron concentration is assumed to be 1 1011 cm3. The dashed line presents the equilibrium solubility during the annealing treatment.
4. Conclusions The analysis of the dissolution kinetics of iron gettered by oxide precipitates in Cz-silicon revealed that there is a strong binding between the gettered iron and oxide precipitates. The effective dissolution energy barrier was determined to be EA ¼ 1:4770:10 eV; as determined in the temperature range from 750 C to 900 C. Considering the diffusion barrier of 0.67 eV for iron diffusion in silicon, our results give an effective binding energy of 0.80 eV between oxide precipitates and the gettered iron impurities. The existence of this binding energy may be used to improve the efficiency of internal gettering during rapid thermal annealing, particularly at temperatures below 800 C or at high-iron contamination levels. But for a high processing temperature (as high as 850 C) and low-iron contamination level (o1011 cm3), the impact of the dissolution barrier is negligible and it is still true that only the last cooling procedure has to be adjusted to optimize internal gettering.
References [1] A.A. Istratov, H. Hieslmair, E.R. Weber, Appl. Phys. A 69 (1999) 13. [2] A.A. Istratov, H. Hieslmair, E.R. Weber, Appl. Phys. A 70 (2000) 489. [3] SIA, International technology roadmap for semiconductors 2002, http://public.itrs.net or http://www.sematech.org. . [4] S.M. Myers, M. Seibt, W. Schroter, J. Appl. Phys. 88 (2000) 3795. [5] A.A. Istratov, H. Hieslmair, E.R. Weber, MRS Bulletin 25 (2000) 33. [6] T.Y. Tan, E.E. Garnder, W.K. Tice, Appl. Phys. Lett. 30 (1977) 175. [7] D. Gilles, E.R. Weber, S.K. Hahn, Phys. Rev. Lett. 64 (1990) 196. [8] M. Aoki, A. Hara, A. Ohsawa, J. Appl. Phys. 72 (1992) 895. [9] F.S. Ham, J. Phys. Chem. Solids 6 (1958) 335. [10] E.G. Colas, E.R. Weber, Appl. Phys. Lett. 48 (1986) 1371. [11] M. Aoki, A. Hara, J. Appl. Phys. 74 (1993) 1440. [12] S.A. McHugo, E.R. Weber, M. Mizuno, F.G. Kirscht, Appl. Phys. Lett. 66 (1995) 2840. [13] A.A. Istratov, W. Huber, E.R. Weber, J. Electrochem. Soc. 150 (2003) G244. [14] K.F. Kelton, R. Falster, D. Gambaro, M. Olma, M. Cornara, P.F. Wei, J. Appl. Phys. 85 (1999) 8097. [15] S.D. Brotherton, P. Bradley, A. Gill, J. Appl. Phys. 57 (1985) 1941. [16] H. Hieslmair, S. Balasubramanian, A.A. Istratov, E.R. Weber, Semicond. Sci. Technol. 15 (2001) 567. . a, A.A. Istratov, E.R. Weber, Appl. [17] P. Zhang, H. V.ainol. Phys. Lett. 83 (2003) 4324.