KEYWORDS: MOSFET, reverse short-channel effect, transient enhanced diffusion, arsenic, phosphorus, source, drain, ion implantation

Transcription

1 Jpn. J. Appl. Phys. Vol. 42 (2003) pp Part 1, No. 5A, May 2003 #2003 The Japan Society of Applied Physics -Assisted Low-Energy Arsenic Implantation Technology for N-Channel Metal Oxide Semiconductor Field-Effect Transistor Source/Drain Formation Process Kiyotaka IMAI, SeiichiSHISHIGUCHI 1,KentaroSHIBAHARA 2 andshinyokoyama 2 ULSI Device Development Division, NEC Corporation, 1120 Shimokuzawa, Sagamihara, Kanagawa , Japan 1 Association of Super-Advanced Electronics Technology (ASET), 292 Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa , Japan 2 Research Center for Nanodevices and Systems, Hiroshima University, Kagamiyama, Higashihiroshima, Hiroshima , Japan (Received October 17, 2002; accepted for publication January 17, 2003) In this paper, we present phosphorous-assisted low-energy arsenic implantation technology for forming n þ source/drain regions. Low-energy arsenic implantation suppresses transient enhanced diffusion of boron, and this relieves the reverse shortchannel effect. Combined with phosphorous implantation, this technology minimizes both the junction-leakage current and the gate polysilicon depletion (polydepletion) effect. A 130-nm-gate-length n-channel metal oxide semiconductor field-effect transistor (MOSFET) has been fabricated with this technology, which exhibits improved I on I off and V th roll-off characteristics. [DOI: /JJAP ] KEYWORDS: MOSFET, reverse short-channel effect, transient enhanced diffusion, arsenic, phosphorus, source, drain, ion implantation 1. Introduction One of the most important issues in fabricating a smallsize transistor is how to control the impurity profiles precisely. Boron is widely used as a p-type dopant for the channel and halo regions of n-channel metal oxide semiconductor field-effect transistors (MOSFETs). However, it is difficult to form ideal boron profiles in these regions. This is because point defects due to extension or source/drain ion implantation enhance the redistribution of boron during activation annealing. As a result, boron piles up at the channel surface which leads to a threshold voltage increase. This phenomenon is known as transient enhanced diffusion (TED), and this causes the reverse short-channel effect (RSCE). 1) In transistors with severe RSCE, the threshold voltage initially rolls up with decreasing gate length and after reaching the maximum value, abruptly rolls off. This RSCE and the abrupt roll-off increase the threshold voltage spread of short-channel MOSFETs. Although the use of indium is effective in preventing RSCE, boron is preferable as a main p-type dopant because the electrical solubility of indium is lower, saturating at cm 3. 2) In this paper, we propose a novel source/drain formation process, which minimizes the implantation damage and enables precise boron profile control for n-channel MOS- FETs. First, we investigated the relation between the reverse short-channel effect and arsenic ion-implantation energy. Next, in order to evaluate the boron transient enhanced diffusion due to the source/drain implant damage, we implanted arsenic or phosphorus ions into a boron deltadoped superlattice with a different energy, and quantified TED by the diffusion length of delta-doped marker layers. Then we show how we can obtain suitable junction depth with minimized implantation damage using the phosphorusassisted low-energy arsenic implantation process. Sub- 10 kev arsenic implantation can significantly reduce the reverse short-channel effect caused by TED, and excellent V th roll-off and I on I off characteristics in n-channel MOS- FETs can be obtained. The phosphorus implantation, which forms a buffer around the source/drain junction, can reduce junction leakage current and also improve the polydepletion effect when the source/drain implantation is used for gate 2654 polysilicon doping. 2. Device Fabrication The process sequence for the experiment is based on the 130-nm gate-length complimentary-mosfet (CMOS) technology. 3,4) The channels of an n-channel MOSFET were doped with boron at 30 kev/ cm 2. The gate oxynitride film (EOT = 2.6 nm) was formed using NO gas. Then a nondoped gate polyelectrode was formed. A shallow extension was formed using 2keV or 4 kev arsenic implantation with or without a halo BF 2 implant. Then RTA of 1000 C for 10 s was executed to minimize the influence of TED caused by the extension implantation. After forming an 80-nm-thick sidewall spacer, source/drain implantations were performed with several split conditions. The cap chemical vapor deposition (CVD) SiO 2 was deposited and source/drain activation of 1000 C for 10 s was performed, followed by CoSi 2 formation on the source/drain and gateelectrode regions. 3. Dependence of Reverse Short-Channel Effect on Source/Drain Implantation Conditions Figure 1 shows the threshold voltage versus gate length for devices with different source/drain arsenic implantation Threshold Vd=1.5V (V) S/D As 30 kev S/D As 8 kev Extension As 2 kev w/o halo BF 2 RTA 1000 C 10 s Gate length (µm) V th max Fig. 1. Threshold voltage versus gate length for devices with different source/drain arsenic implantation energies for an n-channel MOSFET. Severe RSCE is observed with arsenic implantation energy of 30 kev.

2 Jpn. J. Appl. Phys. Vol. 42(2003) Pt. 1, No. 5A K. IMAI et al V th max (mv) extension As 4 kev extension As 2 kev 10 w/o halo BF 2 RTA 1000 C 10 s Source/drain As implantation energy (kev) TED length of Boron (nm) Dose: cm -3 RTA: 1000 C 10 s Arsenic Source/drain implantation energy (kev) Fig. 2. Dependence of Vth max on the source/drain arsenic implantation energy as well as extension arsenic implantation energy. Here, Vth max ¼ V peak th V th at L g ¼ 10 mm as shown in Fig. 1. energies. Halo implantation, which also affects the reverse short-channel effect, was not used in these samples. As shown in Fig. 1, it is obvious that high-energy arsenic implantation causes severe RSCE. In order to quantify the RSCE, we introduce the value Vth max (¼ V peak th V th at L g ¼ 10 mm). Figure 2 shows the dependence of Vth max on source/ drain arsenic implantation energy as well as extension arsenic implantation energy. Here, the arsenic dose in the source/drain was cm 2, while that in the extension was cm 2. The value of Vth max increases almost linearly with the increase in source/drain arsenic implantation energy. Furthermore, Vth max depends on the extension arsenic implantation energy. From these results, we can sufficiently suppress the reverse short-channel effect with 2keV extension arsenic and 8 kev source/drain arsenic implantations for the n-channel MOSFET. 4. Evaluation of TED with Source/Drain Implantation Next, we evaluated the boron transient enhanced diffusion induced by source/drain implantation damage. Boron deltadoped superlattices (nondoped Si/Boron delta-doped layer/ nondoped Si) were formed as marker layers for implantation-damage evaluation using ultrahigh-vacuum chemical vapor deposition (UHV-CVD) on Si substrates. 5,6) The deltadoped layers were 5 nm thick and the nondoped Si layers were 100 nm thick. The diffusion length of the delta-doped marker layers as shown in Fig. 3 can evaluate the implantation damage. Arsenic or phosphorus ions were implanted into boron delta-doped superlattices, in order to evaluate the source/drain implantation damage in the n-channel MOS- FET. The implantation dose for arsenic or phosphorus was Fig. 4. TED length dependence on source/drain implantation energy. Arsenic or phosphorus implantation dose is cm 2. Here, TED length was evaluated by the diffusion length of the delta-doped marker layers as shown in Fig cm 2 and the RTA conditions were 1000 C for 10 s. Figure 4 shows TED length dependence on the implantation energy. The TED increased proportionally with arsenic or phosphorus implantation energy. A very interesting result is that the TED length with phosphorus was longer than that with arsenic implantation, even though the mass number of phosphorus is smaller than that of arsenic. This result can be explained by the following mechanism. The arsenic implantation forms an amorphous region at the surface of the silicon substrate and this works as a sink for the point defects produced by the arsenic implantation. In contrast, the phosphorus implantation does not create an amorphous region and most of the point defects produced by the phosphorous implantation directly enhance the transient diffusion of boron. To confirm this mechanism, we investigated arsenic and phosphorus mixed implantation. Table I shows source/drain formation conditions with arsenic and phosphorus implantations. Total dose of arsenic and phos- Table I. Arsenic and phosphorus mixed ion implantation conditions for source/drain formation. S/D ion implantation conditions Arsenic E (kev) Dose (10 15 cm 2 ) E (kev) Dose (10 15 cm 2 ) Arsenic I/I CVD SiO 2 Boron delta-doped layers undoped Si undoped Si 1000 C 10 s TED length undoped Si Fig. 3. Evaluation method of source/drain implantation damage by using the diffusion length of the delta-doped marker layers.

3 2656 Jpn. J. Appl. Phys. Vol. 42 (2003) Pt. 1, No. 5A K. IMAI et al. Fig. 5. V th max (mv) implantation energy:10kev 7.5keV 5keV Dependence of V max th on source/drain implantation conditions. phorus was kept constant at cm 2. Halo implantation was used to suppress the short-channel effect in these experiments. Figure 5 shows the dependence of Vth max on arsenic and phosphorus mixed implantations. Vth max decreases with arsenic dose decrease from cm 2 to cm 2 and phosphorus dose increase from 0 to cm 2. This result indicates that phosphorus implantation into an amorphous region introduced by arsenic implantation can reduce TED. Lower phosphorus energy is preferable for reducing the RSCE because the number of point defects will be smaller. 5. Concept of -Assisted Low-Energy Arsenic Implantation From evaluation of RSCE and TED with different extension and source/drain implant energies, it is obvious that low-energy arsenic implantation is effective for suppressing the TED and subsequent RSCE. Reduction of extension energy is preferable for obtaining a shallow junction, resulting in an excellent V th roll off. However, we have a problem with 8 kev source/drain arsenic implantation because the junction is not deep enough to suppress the junction leakage with the CoSi 2 process. In order to minimize both the RSCE and junction-leakage current, we evaluated the phosphorus-assisted low-energy arsenic implantation for forming n þ source/drain regions. The concept behind this implantation technology is shown in Fig. 6. First, low-energy (<10 kev) arsenic ions are implanted to amorphize the silicon surface. Then phosphorus ions are implanted into the amorphous layer. Point defects produced by the phosphorus implantation are consumed in the amorphous layer. Therefore, phosphorus implantation does not affect the boron TED. Deep junctions can be formed due to the high diffusivity of phosphorus. As a result, we can obtain suitable junction depth with minimized implantation damage using the phosphorus-assisted lowenergy arsenic implantation process. Figure 7 shows the SIMS profile of the source/drain region formed by 8 kev arsenic and 5 kev phosphorus implantation. Here, the dose of arsenic is cm 2 and that of phosphorus is cm 2. The depth of the buffer region produced by phosphorus implantation is relatively shallower compared with phosphorus solitary implantation. Fig. 7. SIMS profile of the source/drain region formed by 8 kev arsenic and 5 kev phosphorus implantation. Amorphous layer 1. Amorphous layer formation with low-energy (E<10keV) arsenic ion-implantation Arsenic profile after implantation 2. ion implantation into amorphous layer profile after implantation 3. RTA 1000 C 10 s. Ideal junction depth with reduced TED after diffusion profile after diffusion Fig. 6. Concept of the phosphorus-assisted low-energy arsenic implantation.

4 Jpn. J. Appl. Phys. Vol. 42(2003) Pt. 1, No. 5A K. IMAI et al Fig. 8. Junction leakage characteristics of n þ /p diode with different source/drain implantation conditions. In the case of phosphorus solitary implantation, point defects induced by phosphorus implantation also enhance diffusion of the implanted phosphorus itself. In contrast, the case of phosphorus implantation combined with high dose arsenic, point defects are consumed in the amorphous region created by arsenic implantation and phosphorus profile remains relatively shallow. Therefore we can obtain suitable junction depth by controlling the energy and dose of phosphorus. The leakage currents of the n þ /p junction with cobalt silicide are shown in Fig. 8. The 8 kev arsenic implantation induces large leakage current. Combined with cm 2 /5 kev phosphorus implantation, however, the leakage current remains low at the reverse bias of 1.5 V. 6. Optimization of -Assisted Low-Energy Arsenic Ion-Implantation In this session, we demonstrate the optimization of phosphorous-assisted low-energy arsenic implantation for 130-nm-gate-length n-channel MOSFET. The evaluation items are, 1) junction leakage, 2) junction capacitance, 3) V th roll-off characteristics, and 4) polydepletion effect. The total dose of arsenic and phosphorus was kept constant at cm 2. Halo implantation was used to suppress the shortchannel effect in these experiments. Figure 9 shows the dependence of n þ junction leakage on source/drain implantation conditions. The 8 kev/ cm 2 arsenic solitary implantation (condition B) suppresses the RSCE as shown in Fig. 5. However, this causes significant leakage increase of n þ /p junction. On the contrary, phosphorus-assisted 8 kev arsenic implantation effectively suppresses junction leakage. Junction leakage current with 8 kev/ cm 2 arsenic and 7.5 kev/ cm 2 phosphorus is the same level as that with 30 kev/ cm 2 arsenic. A higher energy implantation of phosphorus is preferable for reducing the junction leakage. Figure 10 shows the dependence of n þ junction capacitance on source/drain implantation conditions. The 8 kev/ cm 2 arsenic solitary implantation causes significant increase of n þ /p junction capacitance, because channel concentration at such a shallow n þ /p junction is still high. On the other hand, phosphorus-assisted 8 kev arsenic implantation successfully suppresses the junction capacitance. The junction capacitance with 8 kev/ cm 2 arsenic and 10 kev/ cm 2 phosphorus is the same level as that with 30 kev/ cm 2 arsenic. Fig. 11 shows the dependence of L min on source/drain implantation conditions. Here, L min is defined as L poly at V th =L g ¼ 4 V/mm. The RSCE is significantly reduced with arsenic dose decrease from cm 2 to 3 Junction capacitance (ff/µm 2 ) kev 5 kev Fig. 10. Dependence of n þ junction capacitance on source/drain implantation conditions Junction leakage (A/µm 2 ) ~ kev 5 kev Lmin (µm) kev 5 kev Fig. 9. Dependence of n þ junction leakage on with source/drain implantation conditions. Fig. 11. Dependence of L min on source/drain implantation conditions. Here, L min is defined as L poly at V th =L g ¼ 4 V/mm.

5 2658 Jpn. J. Appl. Phys. Vol. 42 (2003) Pt. 1, No. 5A K. IMAI et al. Tox_inv (nm) kev 5 kev Fig. 12. Dependence of inversion capacitance of n-channel MOSFET on source/drain implantation conditions. Ioff (Vg=0, Vd=1.5V) (A/µm) kev As P-assisted 8 kev As Extension: 2 kev As with halo BF 2 RTA 1000 C 10 s I on (V g =V d =1.5V) (µa/µm) Fig. 14. I on versus I off characteristics of the n-channel MOSFET with halo BF 2 implantation cm 2 as shown in Fig. 5, but L min degrades when the dose of phosphorus reaches cm 2 as shown in Fig. 11. Suppression of the RSCE using phosphorus-assisted low-energy arsenic implantation improves L min, but excess phosphorus in the source/drain region diffuses laterally and degrades L min, with the activation conditions of 1000 C for 10 s. The phosphorus implantation also suppresses the polydepletion effect when source/drain implantation is used for gate polysilicon doping. Figure 12 shows the dependence of inversion capacitance of the n-channel MOSFET on source/ drain implantation conditions. The polydepletion effect with phosphorus-assisted 8 kev arsenic implantation is even better than that with 30 kev arsenic. From Fig. 5 and Figs. 9 12, we concluded the best implantation condition for 130 nm gate-length n-channel MOSFET is 8 kev/ cm 2 arsenic with 7.5 kev/ cm 2 phosphorus. 7. Transistor Characteristics The V th roll-off characteristics of n-channel MOSFET with halo BF 2 implantation are shown in Fig. 13. The V th roll-off characteristics are improved by using low-energy arsenic implantation. Figure 14 shows the I on versus I off Threshold voltage (V) kev As P-assisted 8 kev As 0.1 Extension: 2 kev As with halo BF 2 RTA 1000 C 10 s Gate length (µm) Fig. 13. V th roll-off characteristic of n-channel MOSFET with optimized phosphorus-assisted low-energy arsenic implantation as compared with that of conventional arsenic-only implantation. Fig. 15. I d V g and I d V d characteristics of n-channel MOSFET with optimized phosphorus-assisted low-energy arsenic implantation. characteristics of the n-channel MOSFET with halo BF 2 implantation. The I on =I off ratio improves by using phosphorus-assisted low-energy implantation. These results indicate that boron profiles in the channel and the halo region are well controlled due to suppressed TED. Figure 15 shows I d V g and I d V d characteristics of the n-channel MOSFET with optimized phosphorus-assisted low-energy arsenic implantation for source/drain formation. The subthreshold slopes is 81.0 mv/decade and the saturation current is 840 ma/mm at V g ¼ V d ¼ 1:5 V, and off current of 6.4 na/mm. 8. Conclusion We propose phosphorus-assisted low-energy arsenic source/drain implantation. Low-energy arsenic implantation effectively reduces the TED of boron in the channel and halo regions of an n-channel MOSFET. Our experiment shows that phosphorus mixed implantation suppresses both junction leakage and the polydepletion effect. The source/drain implantation of 8 kev/ cm 2 arsenic with 7.5 kev/ cm 2 phosphorus effectively suppresses the RSCE and improves I on versus I off characteristics. Acknowledgments The authors would like to thank K. Ando and S. Koyama for helping with sample preparation.

6 Jpn. J. Appl. Phys. Vol. 42(2003) Pt. 1, No. 5A K. IMAI et al ) A. Kamgar, H.-H. Vuong, C. T. Liu, C. S. Rafferty and J. T. Clemens: IEDM Tech. Dig., 1997, p ) J. Liu and S. Mehta: Proc. of SPIE, Microelectronic Device Technology II, 1998, p ) K. Imai, S. Shishiguchi, K. Yamaguchi, N. Kimizuka, H. Onishi and T. Horiuchi: Symp. on VLSI Tech., 1999, p ) K. Imai, K. Yamaguchi, N. Kimizuka, H. Onishi, T. Kudo, A. Ono, K. Noda, Y. Goto, H. Fujii, M. Ikeda, K. Kazama, S. Maruyama, T. Kuwata and T. Horiuchi: IEDM Tech. Dig., 1999, p ) S. Shishiguchi, A. Mineji., T. yasunaga and S. saito: Symp. on VLSI Tech., 1998, p ) S. Shishiguchi, A. Mineji, T. Matsuda and H. Kitajima: Electrochem. Soc. Symp. Proc., 99 10, 1999, p. 105.