- Email: [email protected]
A simple two-step phosphorus doping process for shallow junctions by applying a controlled adsorption and a diffusion in an oxidising ambient
Materials Science and Engineering B 114–115 (2004) 362–366
A simple two-step phosphorus doping process for shallow junctions by applying a controlled adsorption and a diffusion in an oxidising ambient Bodo Kalkofen∗ , Marco Lisker, Edmund P. Burte Institute of Micro and Sensor Systems, Otto-von-Guericke-University Magdeburg, P.O. Box 4120, D-39016 Magdeburg, Germany
Abstract A simple phosphorus doping technique for shallow junctions is presented. The low pressure doping process was carried out in a single RTP reactor chamber by using a two-step process: a controlled adsorption of phosphorus on the silicon surface and a rapid thermal diffusion in an oxidising ambient without the deposition of an oxide capping layer. A low concentration of 50 vpm phosphine diluted in hydrogen allowed a sufficient phosphorus supply while the deposition of phosphorus on the reactor walls was insignificant. The phosphine decomposed on the clean silicon surface at a temperature of 550 ◦ C, at which the silicon surface is saturated by the adsorbed phosphorus. The shallow junctions were defined by successive rapid thermal annealing at temperatures above the adsorption temperature. An oxygen pressure of 4.2 × 103 Pa during the annealing prevented the phosphorus from desorption. Therefore, a deposition of an additional oxide-capping layer was not necessary, allowing more simple processing. This doping method provides shallow junctions of depths below 100 nm with sheet resistances below 1000 /sq. © 2004 Elsevier B.V. All rights reserved. Keywords: Shallow junction; MOSFET; Phosphine adsorption; Phosphours vapor phase doping; Sheet resistance
1. Introduction The progressive development of CMOS-Technology requires a continual shrinking of the device structures with an ever-increasing complexity. Therefore, there is an ongoing process in stretching the limits of manufacturability to find solutions for the increasing demands in precision, reliability and efficiency. One of the important demands for the increase of the performance of sub-100 nm MOSFETs for the USLI technology is the need of shallow source/drain extension junctions to suppress the short channel effects (SCE) while maintaining a low parasitic series resistance. The introduction of dopants by low energy ion implantation is widely used to tackle these challenges. However, not all the problems connected to the implantation have finally been solved. Crystal damage and channelling and the transient enhanced diffusion (TED) during the high temperature annealing for the activation of the dopants do still not make the requirements for the shallow junctions easy to meet. Additionally, ∗
Corresponding author. Tel.: +49 391 67 11631; fax: +49 391 67 12103. E-mail address: [email protected] (B. Kalkofen).
0921-5107/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.mseb.2004.07.064
the doping of trenches for next generation memory devices can hardly be carried out by ion implantation because of shadowing effects of these structures. Furthermore, we saw a need for an alternative doping process existing in smaller research laboratories where the implanter facilities are uneconomical. Some alternatives to the ion implantation doping have been studied and are still in the focus of research, such as plasma immersion ion implantation (PII), gas-immersion laser doping (GILD) and vapor phase doping (VPD) (see [1] and citations herein). We investigated an alternative low pressure rapid thermal doping (RTD) process for the phosphorus doping comparable to the vapor phase doping. Some studies about VPD have been published, e.g. for boron [2,3], arsenic [4], and phosphorus doping as well [5–7]. For those experiments, the silicon surface kept at elevated temperatures was exposed to the doping gas allowing the dopant to diffuse directly into the bulk. The conventional vapor phase doping of phosphorus required a considerably higher partial pressure of the phosphine doping gas compared to the VPD of boron with diborane [8] because of an increased thermal desorption of phosphorus at the elevated doping temperatures. The deposition of an addi-
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
tional capping layer is a method to prevent such a desorption. New doping processes sometimes referred to as atomic-layer doping (ALD) were presented for boron doping [9] and arsenic doping [10], applying an oxide capping layer in order to avoid dopant loss from the surface. In this report a similar low pressure doping process is presented, which is composed of the adsorption of phosphorus on the silicon surface and a rapid thermal diffusion in an oxidising ambient without the deposition of an oxide capping layer.
2. Experimental The substrates used for all the doping experiments were p-type boron-doped 150 mm diameter silicon wafers (1 0 0) with a resistivity of about 10 cm. Initially, the wafers were cleaned by a conventional chemical etch procedure, a Caro etch at 120 ◦ C for 5 min and a HF dip. They were rinsed in deionized water and spin dried in a N2 -atmosphere at elevated temperature. Immediately after the cleaning, the wafers were loaded into a load lock chamber that was evacuated subsequently. All the handling and the following processing were carried out under permanent vacuum conditions. The doping was carried out in a modified ‘Hector’ RTPCluster-Module from Steag AST (Mattson Technology, Inc.). This tool allowed a low-pressure single wafer processing. The wafer was placed on three quartz pins, one of which had an integrated thermo-couple for direct and fast response temperature measurement [11]. This method of temperature measurement limited the temperature range for the experiments up to 950 ◦ C. The wafer was heated by radiation of 16 tungsten halogen lamps through a quartz pane. The doping gases were 52.3 vpm PH3 in H2 for the two-step phosphorus doping and 9.91% by volume PH3 in N2 for a conventional vapor phase doping process, respectively. The entire doping process ran automatically following a pre-defined recipe. Two different process sequences were examined: a conventional vapor phase doping as reference process and the two-step doping. The high temperature annealing steps were always carried out under static pressure conditions. For that, the pumping port was closed and the chamber was filled with the doping gas or the oxygen, respectively, until the desired pressure was reached. After that, the temperature cycle was run. The chamber was purged with nitrogen after the high temperature annealing step was finished. The experiments were carried out under different annealing conditions. The steady state annealing temperature was varied between 900 and 950 ◦ C and the steady state time between 30 and 60 s. The process sequence for the two-step doping will be discussed in more detail in the next section. The electrical properties of the doped layers were characterized by four point probe measurements (FPP) on a CDE ResMap 168 and the chemical composition was analysed by secondary ion mass spectroscopy (SIMS) using an IMS 4F of CAMECA. The impact energy of the Cs+ primary ions that were used for the analyses was set to 3.5 keV.
363
3. Results and discussion 3.1. Conventional vapor phase doping with phosphorus The first doping experiments were carried out in a nitrogen-containing atmosphere by using the phosphine-rich gas diluted in nitrogen. A mean sheet resistance of 527 /sq was achieved for a ramp cycle of 60 s ramp-up time to 900 ◦ C, a steady state of 30 s and a ramp-down for 60 s. The doping pressure was 1 × 103 Pa which corresponds to a partial pressure of the phosphine gas of about 100 Pa. SIMS measurements of wafer samples revealed maximum concentrations of phosphorus of 1.2 × 1020 cm−3 . Sato et al. [7] reported of similar experiments with phosphine in H2 . They proposed doping at temperatures above 1000 ◦ C with a higher partial pressure of phosphine. Kiyota et al. [5] achieved for their atmospheric pressure doping a maximum concentration of 1.3 × 1019 cm−3 . Their maximum phosphine partial pressure was limited to about 6 Pa. They also proposed higher phosphine concentrations because of the low doping efficiency they found for the VPD. However, a severe drawback was already observed after about ten runs with the phosphine-rich doping gas. A reddish coating appeared inside the reactor on the cooled walls as well as on the air-cooled quartz pane. This coating was most probably red phosphorus from the thermal decomposition of phosphine. Hsueh [12] reported such an observation in a conventional open tube diffusion system, where PH3 of a partial pressure of about 30 Pa in a nitrogen carrier was used. He proposed a diffusion from a local P2 O5 source. Kiyota et al. [5] did not report about phosphorus contamination, but Sato et al. [7] gave an indication of a phosphorus contamination of the chamber walls. Such contamination layers cannot be tolerated for a lamp-heated RTP tool. A coating of the quartz window adversely affects its transmittance leading to a poor controllability of the transmitted lamp power. A frequent cleaning of the reactor as by a thermal HCl process would increase tool cost and complexity and might be detrimental to the doping process. Hence, the rapid vapor phase doping in the phosphine-rich ambient was technologically unfavorable and was no longer followed. However, the high phosphine concentration is of fundamental importance in the ordinary VPD process sequence to obtain the desired phosphorus concentration in the silicon. This was proved by our further experiments. The use of the diluted 52.3 vpm PH3 in H2 in the same process sequence as described above resulted in no measurable doping of the silicon, as was found by sheet resistance measurements, even though the pressure of the doping gas was increased to about 1 × 104 Pa. Apparently, a partial pressure of PH3 up to 5 × 10−1 Pa is not sufficient for this process sequence. There are a number of publications about the interaction of phosphine with the silicon surface (see [13,14] and citations therein). It has been reported that PH3 easily adsorbs on (1 0 0) Si; a non-dissociative adsorption of PH3 on the dangling bonds of the (2 × 1)-reconstructed surface was found
364
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
at room temperature. At elevated temperatures, dissociation of PH3 takes place depending on the initial PH3 coverage and hence, the number of available free dangling bonds to be occupied by the dissociated hydrogen. Hydrogen desorption from the H-terminated bonds is observed above 400 ◦ C vacating further sites for the adsorption of phosphine and phosphorus, respectively. The fully dissociatively adsorbed state is energetically more favorable than the system consisting of the phosphine gas and the free silicon surface. Hence, the phosphorus is chemisorbed on the silicon surface. One monolayer of phosphorus at full coverage is composed of P–P dimers and corresponds to a sheet concentration of 6.78 × 1014 cm−2 , which is the density of dangling bonds of a free (1 0 0) Si surface. The phosphorus coverage terminates the silicon surface chemically and electrically [15,16], as the phosphorus keeps a non-binding electron pair, resulting in a saturation of the phosphorus coverage as reported by Yu and Meyerson [17]. A maximum coverage was found at 550 ◦ C by thermal desorption measurements, the further increase in temperature leads to desorption of the phosphorus, in a temperature range where the diffusion of phosphorus in silicon is not significant. This explains the difficulties with the common vapor phase doping approach. 3.2. Two-step vapor phase doping with phosphorus The preceding results led to an alternative approach: a doping process was developed that consisted of two steps. Similar methods were described in [9,10] for the doping of silicon with boron and arsenic, respectively. However, a cover layer of silicon oxide was necessary in order to prevent the desorption and the outdiffusion of the dopants. We investigated, whether the need of the capping layer could be avoided. 3.3. Controlled adsorption Fig. 1 shows the trace of the temperature of the wafer measured with the Pin-TC and of the chamber pressure that
Fig. 1. The trace of the wafer temperature and the chamber pressure during the controlled adsorption.
was set for the adsorption of the phosphorus on the silicon surface. Firstly, the wafer was heated with a fast ramp to a temperature of 900 ◦ C, while the flow of PH3 /H2 was held constant and the pressure of the doping gas in the chamber reached an equilibrium. The RTP tool allowed a maximum pressure of 250 Pa at a maximum flow of dopant gas of 100 sccm and a minimum pumping speed. The initial annealing was carried out in order to condition the wafer surface reproducibly. After the initial annealing, the wafer temperature was kept at 550 ◦ C, the temperature of the maximum of phosphor coverage at the silicon surface. This is the actual deposition step by the thermal decomposition of PH3 . Saturation of the surface coverage with phosphorus is already reached at very low PH3 exposures of 12–16 L [17,18] which allows the use of the highly diluted doping gas. 3.4. Annealing The annealing was carried out in an oxygen-containing ambient. The presence of oxygen during the annealing is known to affect the diffusivities of dopants. Oxidationenhanced diffusion (OED) is observed for dopant atoms, which diffuse predominantly by an interstitial mechanism like phosphorus because of the positive impact of the oxygen on the generation of excess of self-interstitials. In Fig. 2, a typical process example with the trace of the wafer temperature and the oxygen pressure is shown. The sheet resistance measurements proved that the oxygen is indispensable for the annealing step, since the phosphorus doping failed when the same sequence was performed but no oxygen was used. The first experiments proved that the doping of silicon with phosphorus is possible using the low phosphine partial pressure of 1.25 × 10−2 Pa during the adsorption. Contamination of the reactor walls with reddish phosphorus was not found. A minimum sheet resistance of 950 /sq for an oxygen annealing at a temperature of 950 ◦ C for 30 s was reached and of 2250 /sq for an annealing at 900 ◦ C for 60 s. The
Fig. 2. The trace of the wafer temperature and the chamber pressure during the annealing step.
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
Fig. 3. Sheet resistance depending on the time a wafer was in contact to clean room air after the doping process.
sheet resistance measurements were carried out in the clean room ambient. It was observed that the sheet resistance values increased during the first 100 min the wafers were set to air before constant values were reached. The change of the sheet resistance measured on one site for different times is illustrated in Fig. 3. The increase of the sheet resistance indicates that a native oxide grew on the silicon after the doping process during the initial exposure of the wafer to the clean room air. The native oxide growth was accompanied by a phosphorus segregation and is known to be self limited for a wafer storage in air. Apparently, an oxide thickness saturation occurred after the first 100 min in air. Results of the SIMS profile measurements of samples of the doped wafers are presented in Fig. 4. The slope of the phosphorus concentration in the silicon could be approximated by a Gaussian distribution for a wide range of the profile. The Gaussian shape of the profile indicates that the conditions of a drive-in redistribution were present during the annealing. The junction depth which is defined here as the depth were the concentration of the phosphorus falls to a level of 1 × 1017 cm−3 was calculated from
365
the SIMS profiles to be 60 nm for the 900 ◦ C annealing and 98 nm for the 950 ◦ C annealing, respectively. A pile-up of the phosphorus concentration close to the surface is visible in the profiles. It was found that the piledup phosphorus did not contribute to the active doping. Sheet resistance calculations of the profiles were carried out using ATHENA simulation software of SILVACO. The calculated sheet resistances of the SIMS profiles were noticeably lower than the measured ones while the calculated values of the Gaussian distributions were only slightly lower than the measured results. But the congruence was almost exact when the first 5–6 nm of the profile were not included into the calculations. This could be explained with regard to the sheet resistance increase presented above. A phosphorus pile-up during a native oxide growth would diminish the number of active dopants close to the interface that had still been active before the native oxide started to grow. The large peaks in the SIMS profiles show an interface effect indicating that the secondary ion yield and the sputtering rate in the native oxide and close to the interface are different from the bulk. The phosphorus pile-up and the enhanced oxidation on phosphorus doped surfaces is known from different studies (e.g. [19,20]). More detailed investigation about the wafer surface conditions that were predominant during the different steps of our process is subject to further work. The sheet concentrations of the dopants were calculated from the Gaussian distribution and were 6.7 × 1013 cm−2 for the 950 ◦ C annealing and 2.6 × 1013 cm−2 for the 900 ◦ C annealing, respectively. Assuming a coverage of a monolayer of phosphorus, at most 10% of the adsorbed phosphorus diffused into the silicon onto active sites. For the time being, the observed combined junction depth and sheet resistance behaviour of the junctions made by the two-step doping process did not fall within the requirements of the ITRS roadmap for ultra shallow junctions for the current technology node as the concentration of the introduced phosphorus appeared to be too low. However, our initial experiments were restricted in terms of the doping temperature. Further studies are under way to optimize the doping sequence with respect to an improved phosphorus supply. The observed oxidation of the silicon after the doping sequence indicates that the low oxygen concentration in the doping ambient did not result in an oxide coverage of the surface that was hence not capped and still highly reactive. This should allow a multi-step doping by a sequence of repeated adsorption and short thermal doping cycles permitting a new atomic-layer doping.
4. Conclusions
Fig. 4. SIMS profiles of samples of two different wafers that were annealed in oxygen at different temperatures and the respective Gaussian approximations.
A low pressure phosphorus doping process for shallow junctions in silicon using low-concentration phosphine in hydrogen as doping source was presented. Low phosphine partial pressures of less than 5 × 10−1 Pa were not sufficient for a conventional vapor phase doping because of the
366
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
high desorption rate of phosphorus at the doping temperature. The required high partial pressures of phosphine of about 100 Pa for those processes resulted in a rapid contamination of the chamber walls with phosphorus. The presented two-step process sequence applied a separate step for the adsorption of phosphorus on the silicon surface and succeeding high temperature annealing in an oxidising ambient without additional phosphorus supply. This allowed to maintain a low phosphine partial pressure of 1.25 × 10−2 Pa during the adsorption at 550 ◦ C. The oxidising ambient to a significant extent prevented adsorbed phosphorus from desorption, so that an additional deposition of a capping layer could be avoided. Acknowledgements The work was supported by the Deutsche Forschungsgemeinschaft, DFG (German Research Foundation).
References [1] E.C. Jones, E. Ishida, Mater. Sci. Eng. R 24 (1998) 1. [2] Y. Kiyota, T. Nakamura, T. Inada, A. Kuranouchi, Y. Hirano, J. Electrochem. Soc. 140 (1993) 1117.
[3] M.J.J. Theunissen, C.E. Timmering, J.G.V. Berkum, Y.J. Mergler, Jpn. J. Appl. Phys. Part 1 38 (1999) 5805. [4] C. Ransom, T.N. Jackson, J.F. DeGelormo, C. Zeller, D.E. Kotecki, C. Graimann, D.K. Sadana, J. Benedict, J. Electrochem. Soc. 141 (1994) 1378. [5] Y. Kiyota, T. Nakamura, K. Muraki, T. Inada, J. Electrochem. Soc. 141 (1994) 2241. [6] S. Lourdudoss, S.-.L. Zhang, Appl. Phys. Lett. 64 (1994) 3461. [7] T. Sato, I. Mizushima, N. Aoki, Y. Tsunashima, Jpn. J. Appl. Phys. Part 1 37 (1998) 1162. [8] Y. Kiyota, T. Inada, J. Vac. Sci. Technol. A 19 (2001) 2441. [9] K.-S. Kim, Y.-H. Song, K.-T. Park, H. Kurino, T. Matsuura, K. Hane, M. Koyanagi, Thin Solid Films 369 (2000) 207. [10] K.W. Koh, H.J. Oh, H. Choi, H. Kurino, M. Koyanagi, Mater. Sci. Eng. B 89 (2002) 435. [11] W. Lerch, W. Blersch, S. Yanagawa, IEEE Transact. Semicond. Manufact. 11 (1998) 598. [12] Y.W. Hsueh, J. Electrochem. Soc. 117 (1970) 807. [13] R. Miotto, G.P. Srivastava, A.C. Ferraz, Phys. Rev. B 63 (2001) 125321 (nine pages). [14] R. Miotto, G.P. Srivastava, R.H. Miwa, A.C. Ferraz, J. Chem. Phys. 114 (2001) 9549. [15] F. Bozso, P. Avouris, Phys. Rev. B 43 (1991) 1847. [16] G.P. Srivastava, Vacuum 67 (2002) 11. [17] M.L. Yu, B.S. Meyerson, J. Vac. Sci. Technol. A 2 (1984) 446. [18] Y. Tsukidate, M. Suemitsu, Appl. Surf. Sci. 151 (1999) 148. [19] Y. Kamiura, Y. Mizokawa, M. Iida, Y. Isobe, K. Kawamoto, Jpn. J. Appl. Phys. Part 1 32 (1993) 4863. [20] Y. Sato, M. Watanabe, K. Imai, J. Electrochem. Soc. 140 (1993) 2679.
A simple two-step phosphorus doping process for shallow junctions by applying a controlled adsorption and a diffusion in an oxidising ambient Bodo Kalkofen∗ , Marco Lisker, Edmund P. Burte Institute of Micro and Sensor Systems, Otto-von-Guericke-University Magdeburg, P.O. Box 4120, D-39016 Magdeburg, Germany
Abstract A simple phosphorus doping technique for shallow junctions is presented. The low pressure doping process was carried out in a single RTP reactor chamber by using a two-step process: a controlled adsorption of phosphorus on the silicon surface and a rapid thermal diffusion in an oxidising ambient without the deposition of an oxide capping layer. A low concentration of 50 vpm phosphine diluted in hydrogen allowed a sufficient phosphorus supply while the deposition of phosphorus on the reactor walls was insignificant. The phosphine decomposed on the clean silicon surface at a temperature of 550 ◦ C, at which the silicon surface is saturated by the adsorbed phosphorus. The shallow junctions were defined by successive rapid thermal annealing at temperatures above the adsorption temperature. An oxygen pressure of 4.2 × 103 Pa during the annealing prevented the phosphorus from desorption. Therefore, a deposition of an additional oxide-capping layer was not necessary, allowing more simple processing. This doping method provides shallow junctions of depths below 100 nm with sheet resistances below 1000 /sq. © 2004 Elsevier B.V. All rights reserved. Keywords: Shallow junction; MOSFET; Phosphine adsorption; Phosphours vapor phase doping; Sheet resistance
1. Introduction The progressive development of CMOS-Technology requires a continual shrinking of the device structures with an ever-increasing complexity. Therefore, there is an ongoing process in stretching the limits of manufacturability to find solutions for the increasing demands in precision, reliability and efficiency. One of the important demands for the increase of the performance of sub-100 nm MOSFETs for the USLI technology is the need of shallow source/drain extension junctions to suppress the short channel effects (SCE) while maintaining a low parasitic series resistance. The introduction of dopants by low energy ion implantation is widely used to tackle these challenges. However, not all the problems connected to the implantation have finally been solved. Crystal damage and channelling and the transient enhanced diffusion (TED) during the high temperature annealing for the activation of the dopants do still not make the requirements for the shallow junctions easy to meet. Additionally, ∗
Corresponding author. Tel.: +49 391 67 11631; fax: +49 391 67 12103. E-mail address: [email protected] (B. Kalkofen).
0921-5107/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.mseb.2004.07.064
the doping of trenches for next generation memory devices can hardly be carried out by ion implantation because of shadowing effects of these structures. Furthermore, we saw a need for an alternative doping process existing in smaller research laboratories where the implanter facilities are uneconomical. Some alternatives to the ion implantation doping have been studied and are still in the focus of research, such as plasma immersion ion implantation (PII), gas-immersion laser doping (GILD) and vapor phase doping (VPD) (see [1] and citations herein). We investigated an alternative low pressure rapid thermal doping (RTD) process for the phosphorus doping comparable to the vapor phase doping. Some studies about VPD have been published, e.g. for boron [2,3], arsenic [4], and phosphorus doping as well [5–7]. For those experiments, the silicon surface kept at elevated temperatures was exposed to the doping gas allowing the dopant to diffuse directly into the bulk. The conventional vapor phase doping of phosphorus required a considerably higher partial pressure of the phosphine doping gas compared to the VPD of boron with diborane [8] because of an increased thermal desorption of phosphorus at the elevated doping temperatures. The deposition of an addi-
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
tional capping layer is a method to prevent such a desorption. New doping processes sometimes referred to as atomic-layer doping (ALD) were presented for boron doping [9] and arsenic doping [10], applying an oxide capping layer in order to avoid dopant loss from the surface. In this report a similar low pressure doping process is presented, which is composed of the adsorption of phosphorus on the silicon surface and a rapid thermal diffusion in an oxidising ambient without the deposition of an oxide capping layer.
2. Experimental The substrates used for all the doping experiments were p-type boron-doped 150 mm diameter silicon wafers (1 0 0) with a resistivity of about 10 cm. Initially, the wafers were cleaned by a conventional chemical etch procedure, a Caro etch at 120 ◦ C for 5 min and a HF dip. They were rinsed in deionized water and spin dried in a N2 -atmosphere at elevated temperature. Immediately after the cleaning, the wafers were loaded into a load lock chamber that was evacuated subsequently. All the handling and the following processing were carried out under permanent vacuum conditions. The doping was carried out in a modified ‘Hector’ RTPCluster-Module from Steag AST (Mattson Technology, Inc.). This tool allowed a low-pressure single wafer processing. The wafer was placed on three quartz pins, one of which had an integrated thermo-couple for direct and fast response temperature measurement [11]. This method of temperature measurement limited the temperature range for the experiments up to 950 ◦ C. The wafer was heated by radiation of 16 tungsten halogen lamps through a quartz pane. The doping gases were 52.3 vpm PH3 in H2 for the two-step phosphorus doping and 9.91% by volume PH3 in N2 for a conventional vapor phase doping process, respectively. The entire doping process ran automatically following a pre-defined recipe. Two different process sequences were examined: a conventional vapor phase doping as reference process and the two-step doping. The high temperature annealing steps were always carried out under static pressure conditions. For that, the pumping port was closed and the chamber was filled with the doping gas or the oxygen, respectively, until the desired pressure was reached. After that, the temperature cycle was run. The chamber was purged with nitrogen after the high temperature annealing step was finished. The experiments were carried out under different annealing conditions. The steady state annealing temperature was varied between 900 and 950 ◦ C and the steady state time between 30 and 60 s. The process sequence for the two-step doping will be discussed in more detail in the next section. The electrical properties of the doped layers were characterized by four point probe measurements (FPP) on a CDE ResMap 168 and the chemical composition was analysed by secondary ion mass spectroscopy (SIMS) using an IMS 4F of CAMECA. The impact energy of the Cs+ primary ions that were used for the analyses was set to 3.5 keV.
363
3. Results and discussion 3.1. Conventional vapor phase doping with phosphorus The first doping experiments were carried out in a nitrogen-containing atmosphere by using the phosphine-rich gas diluted in nitrogen. A mean sheet resistance of 527 /sq was achieved for a ramp cycle of 60 s ramp-up time to 900 ◦ C, a steady state of 30 s and a ramp-down for 60 s. The doping pressure was 1 × 103 Pa which corresponds to a partial pressure of the phosphine gas of about 100 Pa. SIMS measurements of wafer samples revealed maximum concentrations of phosphorus of 1.2 × 1020 cm−3 . Sato et al. [7] reported of similar experiments with phosphine in H2 . They proposed doping at temperatures above 1000 ◦ C with a higher partial pressure of phosphine. Kiyota et al. [5] achieved for their atmospheric pressure doping a maximum concentration of 1.3 × 1019 cm−3 . Their maximum phosphine partial pressure was limited to about 6 Pa. They also proposed higher phosphine concentrations because of the low doping efficiency they found for the VPD. However, a severe drawback was already observed after about ten runs with the phosphine-rich doping gas. A reddish coating appeared inside the reactor on the cooled walls as well as on the air-cooled quartz pane. This coating was most probably red phosphorus from the thermal decomposition of phosphine. Hsueh [12] reported such an observation in a conventional open tube diffusion system, where PH3 of a partial pressure of about 30 Pa in a nitrogen carrier was used. He proposed a diffusion from a local P2 O5 source. Kiyota et al. [5] did not report about phosphorus contamination, but Sato et al. [7] gave an indication of a phosphorus contamination of the chamber walls. Such contamination layers cannot be tolerated for a lamp-heated RTP tool. A coating of the quartz window adversely affects its transmittance leading to a poor controllability of the transmitted lamp power. A frequent cleaning of the reactor as by a thermal HCl process would increase tool cost and complexity and might be detrimental to the doping process. Hence, the rapid vapor phase doping in the phosphine-rich ambient was technologically unfavorable and was no longer followed. However, the high phosphine concentration is of fundamental importance in the ordinary VPD process sequence to obtain the desired phosphorus concentration in the silicon. This was proved by our further experiments. The use of the diluted 52.3 vpm PH3 in H2 in the same process sequence as described above resulted in no measurable doping of the silicon, as was found by sheet resistance measurements, even though the pressure of the doping gas was increased to about 1 × 104 Pa. Apparently, a partial pressure of PH3 up to 5 × 10−1 Pa is not sufficient for this process sequence. There are a number of publications about the interaction of phosphine with the silicon surface (see [13,14] and citations therein). It has been reported that PH3 easily adsorbs on (1 0 0) Si; a non-dissociative adsorption of PH3 on the dangling bonds of the (2 × 1)-reconstructed surface was found
364
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
at room temperature. At elevated temperatures, dissociation of PH3 takes place depending on the initial PH3 coverage and hence, the number of available free dangling bonds to be occupied by the dissociated hydrogen. Hydrogen desorption from the H-terminated bonds is observed above 400 ◦ C vacating further sites for the adsorption of phosphine and phosphorus, respectively. The fully dissociatively adsorbed state is energetically more favorable than the system consisting of the phosphine gas and the free silicon surface. Hence, the phosphorus is chemisorbed on the silicon surface. One monolayer of phosphorus at full coverage is composed of P–P dimers and corresponds to a sheet concentration of 6.78 × 1014 cm−2 , which is the density of dangling bonds of a free (1 0 0) Si surface. The phosphorus coverage terminates the silicon surface chemically and electrically [15,16], as the phosphorus keeps a non-binding electron pair, resulting in a saturation of the phosphorus coverage as reported by Yu and Meyerson [17]. A maximum coverage was found at 550 ◦ C by thermal desorption measurements, the further increase in temperature leads to desorption of the phosphorus, in a temperature range where the diffusion of phosphorus in silicon is not significant. This explains the difficulties with the common vapor phase doping approach. 3.2. Two-step vapor phase doping with phosphorus The preceding results led to an alternative approach: a doping process was developed that consisted of two steps. Similar methods were described in [9,10] for the doping of silicon with boron and arsenic, respectively. However, a cover layer of silicon oxide was necessary in order to prevent the desorption and the outdiffusion of the dopants. We investigated, whether the need of the capping layer could be avoided. 3.3. Controlled adsorption Fig. 1 shows the trace of the temperature of the wafer measured with the Pin-TC and of the chamber pressure that
Fig. 1. The trace of the wafer temperature and the chamber pressure during the controlled adsorption.
was set for the adsorption of the phosphorus on the silicon surface. Firstly, the wafer was heated with a fast ramp to a temperature of 900 ◦ C, while the flow of PH3 /H2 was held constant and the pressure of the doping gas in the chamber reached an equilibrium. The RTP tool allowed a maximum pressure of 250 Pa at a maximum flow of dopant gas of 100 sccm and a minimum pumping speed. The initial annealing was carried out in order to condition the wafer surface reproducibly. After the initial annealing, the wafer temperature was kept at 550 ◦ C, the temperature of the maximum of phosphor coverage at the silicon surface. This is the actual deposition step by the thermal decomposition of PH3 . Saturation of the surface coverage with phosphorus is already reached at very low PH3 exposures of 12–16 L [17,18] which allows the use of the highly diluted doping gas. 3.4. Annealing The annealing was carried out in an oxygen-containing ambient. The presence of oxygen during the annealing is known to affect the diffusivities of dopants. Oxidationenhanced diffusion (OED) is observed for dopant atoms, which diffuse predominantly by an interstitial mechanism like phosphorus because of the positive impact of the oxygen on the generation of excess of self-interstitials. In Fig. 2, a typical process example with the trace of the wafer temperature and the oxygen pressure is shown. The sheet resistance measurements proved that the oxygen is indispensable for the annealing step, since the phosphorus doping failed when the same sequence was performed but no oxygen was used. The first experiments proved that the doping of silicon with phosphorus is possible using the low phosphine partial pressure of 1.25 × 10−2 Pa during the adsorption. Contamination of the reactor walls with reddish phosphorus was not found. A minimum sheet resistance of 950 /sq for an oxygen annealing at a temperature of 950 ◦ C for 30 s was reached and of 2250 /sq for an annealing at 900 ◦ C for 60 s. The
Fig. 2. The trace of the wafer temperature and the chamber pressure during the annealing step.
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
Fig. 3. Sheet resistance depending on the time a wafer was in contact to clean room air after the doping process.
sheet resistance measurements were carried out in the clean room ambient. It was observed that the sheet resistance values increased during the first 100 min the wafers were set to air before constant values were reached. The change of the sheet resistance measured on one site for different times is illustrated in Fig. 3. The increase of the sheet resistance indicates that a native oxide grew on the silicon after the doping process during the initial exposure of the wafer to the clean room air. The native oxide growth was accompanied by a phosphorus segregation and is known to be self limited for a wafer storage in air. Apparently, an oxide thickness saturation occurred after the first 100 min in air. Results of the SIMS profile measurements of samples of the doped wafers are presented in Fig. 4. The slope of the phosphorus concentration in the silicon could be approximated by a Gaussian distribution for a wide range of the profile. The Gaussian shape of the profile indicates that the conditions of a drive-in redistribution were present during the annealing. The junction depth which is defined here as the depth were the concentration of the phosphorus falls to a level of 1 × 1017 cm−3 was calculated from
365
the SIMS profiles to be 60 nm for the 900 ◦ C annealing and 98 nm for the 950 ◦ C annealing, respectively. A pile-up of the phosphorus concentration close to the surface is visible in the profiles. It was found that the piledup phosphorus did not contribute to the active doping. Sheet resistance calculations of the profiles were carried out using ATHENA simulation software of SILVACO. The calculated sheet resistances of the SIMS profiles were noticeably lower than the measured ones while the calculated values of the Gaussian distributions were only slightly lower than the measured results. But the congruence was almost exact when the first 5–6 nm of the profile were not included into the calculations. This could be explained with regard to the sheet resistance increase presented above. A phosphorus pile-up during a native oxide growth would diminish the number of active dopants close to the interface that had still been active before the native oxide started to grow. The large peaks in the SIMS profiles show an interface effect indicating that the secondary ion yield and the sputtering rate in the native oxide and close to the interface are different from the bulk. The phosphorus pile-up and the enhanced oxidation on phosphorus doped surfaces is known from different studies (e.g. [19,20]). More detailed investigation about the wafer surface conditions that were predominant during the different steps of our process is subject to further work. The sheet concentrations of the dopants were calculated from the Gaussian distribution and were 6.7 × 1013 cm−2 for the 950 ◦ C annealing and 2.6 × 1013 cm−2 for the 900 ◦ C annealing, respectively. Assuming a coverage of a monolayer of phosphorus, at most 10% of the adsorbed phosphorus diffused into the silicon onto active sites. For the time being, the observed combined junction depth and sheet resistance behaviour of the junctions made by the two-step doping process did not fall within the requirements of the ITRS roadmap for ultra shallow junctions for the current technology node as the concentration of the introduced phosphorus appeared to be too low. However, our initial experiments were restricted in terms of the doping temperature. Further studies are under way to optimize the doping sequence with respect to an improved phosphorus supply. The observed oxidation of the silicon after the doping sequence indicates that the low oxygen concentration in the doping ambient did not result in an oxide coverage of the surface that was hence not capped and still highly reactive. This should allow a multi-step doping by a sequence of repeated adsorption and short thermal doping cycles permitting a new atomic-layer doping.
4. Conclusions
Fig. 4. SIMS profiles of samples of two different wafers that were annealed in oxygen at different temperatures and the respective Gaussian approximations.
A low pressure phosphorus doping process for shallow junctions in silicon using low-concentration phosphine in hydrogen as doping source was presented. Low phosphine partial pressures of less than 5 × 10−1 Pa were not sufficient for a conventional vapor phase doping because of the
366
B. Kalkofen et al. / Materials Science and Engineering B 114–115 (2004) 362–366
high desorption rate of phosphorus at the doping temperature. The required high partial pressures of phosphine of about 100 Pa for those processes resulted in a rapid contamination of the chamber walls with phosphorus. The presented two-step process sequence applied a separate step for the adsorption of phosphorus on the silicon surface and succeeding high temperature annealing in an oxidising ambient without additional phosphorus supply. This allowed to maintain a low phosphine partial pressure of 1.25 × 10−2 Pa during the adsorption at 550 ◦ C. The oxidising ambient to a significant extent prevented adsorbed phosphorus from desorption, so that an additional deposition of a capping layer could be avoided. Acknowledgements The work was supported by the Deutsche Forschungsgemeinschaft, DFG (German Research Foundation).
References [1] E.C. Jones, E. Ishida, Mater. Sci. Eng. R 24 (1998) 1. [2] Y. Kiyota, T. Nakamura, T. Inada, A. Kuranouchi, Y. Hirano, J. Electrochem. Soc. 140 (1993) 1117.
[3] M.J.J. Theunissen, C.E. Timmering, J.G.V. Berkum, Y.J. Mergler, Jpn. J. Appl. Phys. Part 1 38 (1999) 5805. [4] C. Ransom, T.N. Jackson, J.F. DeGelormo, C. Zeller, D.E. Kotecki, C. Graimann, D.K. Sadana, J. Benedict, J. Electrochem. Soc. 141 (1994) 1378. [5] Y. Kiyota, T. Nakamura, K. Muraki, T. Inada, J. Electrochem. Soc. 141 (1994) 2241. [6] S. Lourdudoss, S.-.L. Zhang, Appl. Phys. Lett. 64 (1994) 3461. [7] T. Sato, I. Mizushima, N. Aoki, Y. Tsunashima, Jpn. J. Appl. Phys. Part 1 37 (1998) 1162. [8] Y. Kiyota, T. Inada, J. Vac. Sci. Technol. A 19 (2001) 2441. [9] K.-S. Kim, Y.-H. Song, K.-T. Park, H. Kurino, T. Matsuura, K. Hane, M. Koyanagi, Thin Solid Films 369 (2000) 207. [10] K.W. Koh, H.J. Oh, H. Choi, H. Kurino, M. Koyanagi, Mater. Sci. Eng. B 89 (2002) 435. [11] W. Lerch, W. Blersch, S. Yanagawa, IEEE Transact. Semicond. Manufact. 11 (1998) 598. [12] Y.W. Hsueh, J. Electrochem. Soc. 117 (1970) 807. [13] R. Miotto, G.P. Srivastava, A.C. Ferraz, Phys. Rev. B 63 (2001) 125321 (nine pages). [14] R. Miotto, G.P. Srivastava, R.H. Miwa, A.C. Ferraz, J. Chem. Phys. 114 (2001) 9549. [15] F. Bozso, P. Avouris, Phys. Rev. B 43 (1991) 1847. [16] G.P. Srivastava, Vacuum 67 (2002) 11. [17] M.L. Yu, B.S. Meyerson, J. Vac. Sci. Technol. A 2 (1984) 446. [18] Y. Tsukidate, M. Suemitsu, Appl. Surf. Sci. 151 (1999) 148. [19] Y. Kamiura, Y. Mizokawa, M. Iida, Y. Isobe, K. Kawamoto, Jpn. J. Appl. Phys. Part 1 32 (1993) 4863. [20] Y. Sato, M. Watanabe, K. Imai, J. Electrochem. Soc. 140 (1993) 2679.