Mechanisms of B deactivation control by F co-implantation
N.E.B. Cowern, B. Colombeau, J. Benson, and A.J. Smith Advanced Technology Institute, University of Surrey, Guildford, Surrey GU2 7XH, UK.
W. Lerch, S. Paul and T. Graf Mattson Thermal Products GmbH, Daimlerstr. 10, D-89160 Dornstad, Germany
F. Cristiano and X. Hebras LAAS-CNRS, 7 avenue du Colonel Roche, 31077 Toulouse, France
D. Bolze IHP, Im Technologiepark 25, D-15236 Frankfurt (Oder), Germany
Contact Author:
[email protected] 1
Mechanisms of B deactivation control by F co-implantation
N.E.B. Cowern, B. Colombeau, J. Benson, and A.J. Smith Advanced Technology Institute, University of Surrey, Guildford, Surrey GU2 7XH, UK. W. Lerch, S. Paul and T. Graf Mattson Thermal Products GmbH, Daimlerstr. 10, D-89160 Dornstad, Germany F. Cristiano and X. Hebras LAAS-CNRS, 7 avenue du Colonel Roche, 31077 Toulouse, France D. Bolze IHP, Im Technologiepark 25, D-15236 Frankfurt (Oder), Germany
ABSTRACT
Thermal annealing after preamorphization and solid-phase epitaxy of ultrashallow B implants leads to deactivation and diffusion driven by interstitials released from end-ofrange defects. F inhibits these processes by forming small clusters that trap interstitials. A competing B-F interaction causes deactivation when F and B profiles overlap. Both pathways suppress B transient enhanced diffusion.
PACS numbers: 66.30.Jt, 61.72.Ji, 61.72.Ss, 61.72.Yx, 85.30.Kk Contact Author:
[email protected]
2
In future CMOS technology generations, PMOS source-drain extensions will be just a few nm deep, yet will require sheet resistance (Rs) values below 1000 Ω/sq. A promising way to achieve this is to use preamorphization (PAI) and solid phase epitaxy (SPE). The PAI implant is usually accompanied by F or C co-implantation to prevent post-regrowth deactivation of the ultrashallow B implant. Recently there has been intense interest and controversy over the effect of F on B diffusion during SPE regrowth1-3, and on B diffusion and deactivation during post-regrowth annealing3-6. This paper focuses on the latter (post-regrowth) regime. B deactivation during post-regrowth annealing, without F or C co-implantation, arises from the interaction between substitutional B atoms and interstitial atoms5. Excess self interstitials, Sii, flow down their concentration gradient from the end-of-range (EOR) defect band, where their supersaturation is high due to ripening/dissolution of interstitial clusters and {113} defects in the EOR band, to the surface region containing the B profile, where their concentration is lower due to recombination. Quantitative modeling of EOR defect evolution and the resulting interactions between emitted Sii and B atoms confirms this picture7. Further confirmation comes from recent isochronal annealing experiments8. At present it is unclear how F stabilizes B doping profiles against deactivation. It has been proposed that SPE enables F to form F-V clusters, which then act as traps preventing self interstitials from reaching the surface4. Such clusters have been detected by positron annihilation9, and further experimental evidence for an interstitial trapping mechanism has been reported.3,6 Other experiments suggest that F forms complexes with B atoms, preventing their diffusion and clustering with other B atoms10. It has also been
3
noted that F might increase the stability of EOR defects, thereby reducing the availability of Sii to deactivate B6. These alternative scenarios cannot be discarded easily, as F is an extremely reactive species which may form a variety of complex defects. This paper reports a simple but comprehensive experiment that resolves much of this recent controversy. We clarify the key importance of the Sii trapping mechanism for B deactivation control, and show that direct B-F interactions produce an opposing effect – a degradation in electrical activity. Czochralski -oriented 10-20 Ω.cm silicon wafers were implanted with 30 keV, 1 x 1015 cm-2 Ge and subsequently with 0.5 keV, 1 x 1015 cm-2 B. Some of the remaining wafers were implanted with 0.9 keV, 10 keV, and 22 keV F to doses in the range 3 x 1013 – 1 x 1015 cm-2. These energies place the F projected range (a) at the projected range of the B implant, (b) between the B implant and the amorphouscrystalline interface created by the Ge implant, and (c) at the expected depth of the EOR defect band. SIMS profiles showing the B implant depth distribution, the EOR position, and the three different as-implanted F profiles, are shown in figure 1. Following implantation the wafers were cleaved into samples and annealed in N2 at 650°C or 800°C for 1 sec – 45 min on a recessed wafer in a Mattson rapid thermal processing (RTP) system. Ramp-up and ramp-down rates to/from the set temperature were 50°C/s. Samples were subsequently analyzed by four-point probe (KLA Tencor RS100 type D) to determine Rs, and by secondary ion mass spectrometry (SIMS) at FEI Germany, using O2 or Cs beams to determine atomic profiles of B and Ge, or F, respectively. The Ge profiles, which showed negligible diffusion, were used to enhance the relative depth scale accuracy of the B profiles to a level of ±2%.
4
At 800°C the PAI layer regrows epitaxially in a fraction of a second during the heating ramp, even when co-implanted with F. The consequences of diffusion and clustering in the amorphous phase thus appear in the 1s anneal data, while events in the post-regrowth crystalline phase show up in a comparison between the 1s data and results for the longer anneal times. The results for this post-regrowth regime show several interesting features, indicating that F acts on B through at least two mechanisms, one of which is deleterious for B activation and the other strongly beneficial. To demonstrate this, and identify the likely mechanisms involved, we present first SIMS results with and without F co-implantation, and then Rs data showing de-re-activation trends as a function of F co-implant and annealing conditions. Figure 2 (a) shows SIMS profiles of a 0.5 keV, 1×1015cm-2 B implant, preamorphized by 30 keV Ge, but with no F co-implant, before and after rapid thermal annealing for 1s, 15s, 120s and 2700s at 800°C. The profile after 1s (almost identical to that measured immediately after SPE regrowth at 650°C, not shown) indicates that extremely rapid diffusion of B has taken place in the amorphous phase, as seen in earlier studies1,2. Each annealed profile in figure 2(a) exhibits a static peak, ‘kink’ and fastdiffusing tail, the kink concentration decreasing from ~2×1020cm-3 after 1s to ~2×1019cm3
after 120s, followed by an slower increase up to 2700s. This kink is a well known
feature, marking the concentration level above which B forms immobile clusters. Its initial level after 1s reflects B clustering that has occurred in the amorphous phase before regrowth. The subsequent decrease in this level is caused by interstitial-driven B clustering in the regrown crystal.
5
Figure 2 (b) shows corresponding results where F has been co-implanted at 10 keV to 1×1015cm-2. The impact of F on B diffusion in the amorphous phase is to increase the profile broadening after 1s relative to that in Fig 2(a). This increased broadening, known from previous studies, was first attributed to an effect of F on B diffusion in amorphous silicon1, and later to the role of F in slowing SPE regrowth2. Concerning the behavior of B in the recrystallized silicon at times after 1s, the B SIMS profiles in figure 2(b) show no significant decrease in the kink concentration, indicating that F is preventing the interstitial-driven deactivation process. Since the F has been implanted well beyond the B projected range, its concentration in the B peak region is too low to allow significant capture of B into B-F complexes. We therefore propose that the 10 keV F implant suppresses B deactivation by preventing the arrival of Sii in the B-doped region. Another noticeable feature of figure 2 is the rate of B diffusion after the 10 keV F implant, which is much lower than in the case without F. This is again consistent with a reduction in the Sii supersaturation. The profile depth at a concentration of 3 x 1018/cm3 is within 10% of the value predicted by the process simulator SUPREM4, using a model that assumes point defects are in thermodynamic equilibrium. This agreement shows that the F co-implant has reduced the Sii concentration in the B region to near-equilibrium values. Figure 3 presents Rs data showing B deactivation during annealing at 800°C, with and without high-dose (1 x 1015 cm-2) F implantation with the energies and depth distributions shown in Fig. 1. These results show several key features, discussed in the following paragraphs.
6
In the F-free control case Rs rises to a peak value of 1580 Ω/sq. and then falls at longer annealing times. This behavior, expected from earlier work5,7,8, can be understood in terms of Sii trapping to form B-I clusters, followed by reactivation and diffusion of B, as we also saw in figure 2(a). The addition of a F implant at 0.9 keV causes a deleterious increase in Rs at short annealing times, suggesting that B interacts directly with F to produce complexes with a lower level of electrical activity than the substitutional B atoms that formed them. On further annealing the situation reverses, the value of Rs with the 0.9 keV F implant falling below that found in the control case. The addition of a 10 keV F implant has no deleterious impact at short times, and almost completely suppresses the deactivation observed in the control case. A slight increase in Rs is observed, peaking at a somewhat shorter anneal time than in the case of the 0.9 keV F implant. Since the 10 keV F implant has almost no overlap with the B implant profile, we attribute the improvement in activation at short times to the absence of direct B-F reactions. Since the 10 keV implant also has relatively little overlap with the EOR band, it is unlikely that the suppression of deactivation in this case is related to a change in stability of the EOR band. We propose that deactivation is suppressed by trapping of Sii as they flow from the EOR band towards the surface. Interstitial trapping also accounts for the observed cross-over between the 0.9 keV Rs curve and the control curve at longer annealing times, as the initial deactivation caused by the direct B-F interaction is compensated by a reduction in deactivation due to trapping. In this case there is no overlap whatever between the F implant and the EOR band, thus confirming the dominant role of the Sii trapping mechanism.
7
The result of implanting F at 22 keV is very similar to that obtained at 10 keV, and can be attributed to the same mechanism. The slightly lower peak Rs values obtained with the 22 keV F implant may arise from the increased separation between the B and F profiles, which minimizes direct B-F interactions in the deeper parts of the diffused B profiles. To assess the Sii trapping efficiency, figure 4 plots Rs as a function of F dose at 10 keV, after 45s annealing at 800°C. The strongest deactivation occurs in the absence of F, the Rs value rising from its initial value of ~750 Ω/sq. after SPE regrowth to 1250 Ω/sq. after annealing. The amount of deactivation decreases with F dose, closely following an exponential decay curve. Evidently the B implant is increasingly stabilized against interstitial-driven BIC formation, as Sii diffusing from the EOR band are captured by the intervening field of F-related traps. In general, within such a trapping field, the concentration distribution of Sii quickly reaches a steady state, decreasing with distance x from the interstitial source according to exp(-x/L), where L=(4πacCT)-1/2, ac is the capture radius of the trap and CT its concentration11. The observed exponential dependence on F dose can be derived from this formula, if and only if we assume that CT increases with the square of the implant dose. This suggests that the trap defect contains more than one F atom (applying a simple mass-action argument to our result would imply two). At the same time, assuming reasonable values for ac in the range 0.3-0.5 nm, we calculate that at our highest F dose of 1015/cm2, the number of F-related traps is not more than a factor 2-3 below the initially implanted F dose, since lower values do not produce sufficient trapping. These
8
constraints suggest that the traps are small clusters, with a most probable size in the range of 2-3 F atoms. One possible candidate is the F3V cluster predicted by Diebel4. In summary, implantation of F to depths intermediate between a shallow B implant and a deeper preamorphization end-of-range defect band helps to stabilize the B against deactivation, whilst minimizing deactivation by B-F complexes. The stabilization arises predominantly from the trapping of interstitials at small F-related defects, possibly F-V clusters. The possible role of F in stabilizing EOR defects has neither been established nor ruled out, but does not appear necessary or sufficient to explain our results. Our conclusions, based on the use of ion-implanted B, Ge and F under commercial implantation conditions, are readily applicable to the design and fabrication of advanced high-performance CMOS source-drain structures. This work was partly funded by the CEC projects FRENDTECH and ARTEMIS. We thank D. Carey, R. Duffy and B.J. Pawlak for useful discussions.
9
REFERENCES 1
J. M. Jacques et al., Appl. Phys. Lett. 82, 3469 (2003).
2
R. Duffy et al., Appl. Phys. Lett. 84, 4283 (2004).
3
D. F. Downey, J. W. Chow, E. Ishida, and K. S. Jones, Appl. Phys. Lett. 73, 1263
(1998). 4
M. Diebel et al., Mat. Res. Soc. Symp. Proc. 765, D6.15.1 (2003).
5
B.J. Pawlak et al., Appl. Phys. Lett., 84, 2005 (2004).
6
G. Impellizzeri et al., Appl. Phys. Lett. 84, 1862 (2004).
7
B. Colombeau et al., Mat. Res. Soc. Symp. Proc. 810, C3.6 (2004)
8
W. Lerch et al., Proc. Electrochem. Soc. 2004-01 (2004) pp. 90-105.
9
P.J. Simpson et al., Appl. Phys. Lett. 85, 1538 (2004)
10
A. Mokhberi, R. Kasnavi, P. B. Griffin, and J. D. Plummer, Appl. Phys.Lett. 80, 3530
(2002). 11
N. E. B. Cowern, Appl. Phys. Lett. 64, 2646 (1994).
10
Figure Captions
Figure 1: SIMS depth profiles for the 0.5 keV B implant and the 0.9 keV, 10 keV, and 22 keV F implants used in this study, together with the expected depth range of the EOR defect band from the 30 keV Ge PAI implant (vertical dotted lines).
Figure 2: (a) SIMS profiles of a 0.5 keV, 1×1015cm-2 B implant, before and after rapid thermal annealing. (b) Corresponding results in the case where F has been co-implanted at 10 keV to a dose of 1×1015cm-2.
Figure 3: Sheet resistance values for a 0.5 keV B implant after annealing at 800°C (a) without F co-implant, (b-d) with different 1×1015cm-2 F co-implants.
Figure 4: Impact of 10 keV F implant dose on Rs after annealing for 45 min at 800°C. Error bars are given at the 2σ level. Deactivation of the B implant decreases exponentially with F dose, suggesting that the trap concentration is proportional to the square of the F dose.
11
-3 Concentration (cm )
Fig. 1
10
21
10
20
10
19
10
18
0
20
40
60
Depth (nm)
12
80
100
Fig. 2
22
B conc.(cm-3)
10
B 500eV as-implanted o 800 C 1s o 800 C 15s o 800 C 120s o 800 C 2700s
21
10
20
10
19
10
22
B conc.(cm-3)
10
B 500eV as-implanted o 800 C 1s o 800 C 15s o 800 C 120s o 800 C 2700s
21
10
20
10
19
10
0
20
40
Depth (nm)
13
60
Sheet Resistance (Ω/sq.)
Fig. 3
No F 0.9keV F 10keV F 22keV F
1500 1200 900 600 1
10
100
1000
Annealing time (sec)
14
Fig. 4
Sheet resistance (Ω/sq.)
1400 1200
Data Trapping Model
1000 800 600
0.0
14
5.0x10
F dose (ions/cm2)
15
15
1.0x10