An alternative for shallow doping: gas cluster infusion
12/01/2006
As device dimensions decrease, new methods of efficiently doping to shallow depths are required. Similar manufacturing issues exist for both the precision doping of ultra-shallow junctions (USJ) in the source drain extensions (SDE) and the p+ compensation doping of dual-line DRAM devices. In both applications, conventional beamline implantation may be reaching fundamental limitations, and many companies are exploring alternatives such as plasma doping, large molecular doping, and gas cluster infusion doping. Each of these relatively new technologies has advantages and disadvantages; here, the differentiation of the infusion process from the other techniques will be outlined.
The inherent challenges for traditional ion implanters to deliver sufficiently high beam currents at sufficiently low energies required for shallow doping applications, such as USJ and polysilicon doping for DRAM, have motivated the development of alternative methods for shallow doping. The challenges include large molecule implantation, plasma doping, and infusion doping with ionized gas clusters.
 |
Recent advancements in infusion doping have led to a very high flux (>100mA) of boron at low equivalent monomer implant energies (<1keV). In addition to producing uniquely high equivalent beam currents, the infusion process is physically distinct from implantation (beamline, molecular, or plasma) in the following important ways: 1) the doping depth is independent of the species mass; 2) the doping depth is proportional to the one-third power of the acceleration energy; 3) infusion doping is fully self-amorphizing with no evidence of channeling or end-of-range (EOR) damage; 4) infusion exhibits no self-sputtering at any energy and no dopant build-up on the surface even for very high doses (>5E16/cm2); and 5) the gas cluster beam operates at very low power, requiring no wafer cooling and is fully compatible with photoresist. These differences and the manufacturability of the process will be discussed.
Infusion doping with gas cluster ion beams
The physics of ion implantation has been studied in detail and is relatively well understood. Although they have substantially different methods of delivering the doping species to the substrate, monomer and BF2 beamline, plasma, and molecular doping, all appear to follow traditional ballistic style implantation. Infusion doping with gas cluster ion beams (GCIB), however, is an entirely distinct physical process with behavior and properties significantly different from those of implantation. There are several physical property differences that are relevant to shallow doping applications.
Infusion processing utilizes GCIB technology and has been described in detail elsewhere [1-3]. The GCIB shallow doping applications use gas mixes with a small percentage (1-5%) of a doping gas, (i.e., B2H6) or a mixture of gases (i.e., GeH4+B2H6) diluted in an inert gas such as Ar or He. The gas is delivered to the machine at ~10atm and through adiabatic expansion, clusters of typically >5000 atoms are formed, ionized, accelerated up to 60keV, and transported to the substrate.
The transfer of the cluster energy to the surface results in a rapid heating of a semispherical volume. The cluster molecules are disassociated into their constituent atoms and intermixed, or infused, with the substrate. The gas atoms (H, Ar, He, etc.) will instantaneously leave the surface and all soluble species will infuse to the exact same depth in the substrate independent of atomic mass.
 Figure 1. a) Comparison of 1E15/cm2 B 5keV infusion and 0.5keV traditional beamline doping; b) shows SIMS profiles for the simultaneous infusion of B2H6 and GeH4. |
A typical 5keV infusion doping profile is shown in Fig. 1. The extreme abruptness and lack of channeling or energy contamination compared to a standard 500eV beamline implant is a consequence of the different mechanism of the infusion doping process. The penetration depth of the doping species into the substrate is not determined by the energy of individual atoms or molecules (<5eV in this figure), but rather exclusively by the collective energy of the cluster. The extreme abruptness of the doping profile results in improved leakage and short channel effects (SCE) for USJ, and the lack of energy contamination and very high equivalent beam currents also make infusion doping useful for p+ poly doping.
Mass independent doping
Perhaps the most straightforward and intuitive example of the useful distinction between infusion and all the other doping techniques is the complete independence on mass for doping depths. The profiles and depths are exclusively determined by the collective energy of the cluster which directly determines the super-heating and intermixing depth into the substrate. Any doping species contained within the cluster will infuse to the same depth (Fig 1). The unique capability of simultaneously doping all species to the exact same depth allows for a “cocktail” of multiple gas species such as C, F, As, P, etc., to be used in a single step (no “chained implants” required) with the assurance of all species producing identical doping profiles. The ratio of the retained doses is determined by the gas mix into the machine, and thus the doping ratios, or stoichiometry of the resulting compound materials such as SiGexBy, can be controlled by mass flow controllers and/or the ratio of gas species pre-mixed in the bottle [4, 5].
E1/3 depth dependence
At the lower implant energies required for USJ and DRAM poly doping, all implant techniques are ballistic processes and thus have essentially linear relationships between depth of the implant (determined at 1E18/cm3 concentrations) and the acceleration of the ion. The slope of this relationship is mass dependent and shallower junctions can be produced at a given energy using higher mass molecules; however, the strong relationship between energy and implanted depth results in a problematic susceptibility to energy contamination and thus a problem with tight control of a junction depth. This strong dependence on energy also means all ultra implantation shallow doping processes are necessarily restricted to very low beam energies where productivity becomes problematic and is especially true for beamline tools operating without the benefit of deceleration. On the other hand, for infusion doping, the hemispherical volume of the inter-mixing of the doping species and the substrate is exclusively determined by the collective energy of the cluster. Therefore, the depth with infusion-doping follows a very distinct energy to one third power (i.e., E1/3) dependence (Fig. 2). This weak energy dependence makes GCIB highly efficient for shallow doping allowing for very high equivalent beam currents, extremely abrupt profiles and no risk of energy contamination, but also restricts the infusion process to doping applications of <40nm regardless of doping species.
Self-amorphizing with no EOR damage
When each cluster arrives at the substrate surface, the combination of extremely high pressure (>Mbar) intermixing of gas species and the Si substrate and a high temperature (>5000K) thermal spike effectively produces a fully amorphous layer of the approximate depth of the dopant infusion. The three-dimensional cooling of each of the <10nm diameter a-Si semispheres is too rapid (<5psec) for re-crystallization, and thus all infused surfaces remain fully amorphous and independent of process conditions such as dose, energy, and gas species. The absence of high energy ballistic ions eliminates knock-on effects and pile-up of interstitials at the end of the doping range. The thickness of the infusion-produced amorphous layers follow the E1/3 dependence just as the doping depths and range from ~5nm to ~15nm over the available energy range of 3-60keV. All TEM analyses show complete restoration of the crystal after the anneal with no evidence of the EOR damage typically observed in pre-amorphizing implants (PAI). The lack of EOR damage is likely responsible for the superior leakage performance of infusion-doped USJ devices [9].
No self-sputtering limits
For monomer and molecular implantation, the competing effects between implantation/deposition and etching become more problematic at lower implant energies and higher doses. This effect becomes important for producing junctions shallower than 15nm for USJ and for high concentration compensation doping (>E16/cm2) for poly thicknesses <80nm where low energy implants are required. As shown in Fig. 3, the infusion process has a linear relationship between the GCIB dose and the retained B dose in Si. This linear relationship extends from low doses (<1E15/cm2) up through very high doses (>1E17/cm2). This plot is for 60keV infusion (Xj ≈ 30nm), but the linear relationship holds for all cluster beam energies. For the 30nm deep infusion process, B begins to saturate the Si at a dose between 0.5 and 1E17/cm2. At doses higher than ~1E17/cm2, a pure, fully dense B film will begin to be deposited on the surface whose thickness is also linearly related to GCIB dose. Analysis of as-infused samples with 5E16/cm2 doping levels using high-resolution XPS and SIMS clearly showed no B pile-up on the surface where the concentration of Si on the surface is >30%; additionally, all B was bonded to Si or Si-O, and no B-B bonds were detected.
 Figure 3. Retained B dose vs. GCIB dose. Plot demonstrates no self-sputtering limit for infusion doping levels up to >1E17/cm2. |
Low-beam power
Another important differentiation of gas cluster doping is its compatibility with photoresist. Hydrogen out-gassing, surface hardening or polymerizing, and for plasma doping, dopant deposition, are all photoresist related issues with implantation. The relatively high total power delivered to the wafer necessitates backside cooling for all these technologies to help protect the photoresist. For infusion doping, however, beam currents and power are relatively low even for very high dopant flux because there is less than one ion per thousand atoms. The typical maximum beam power delivered to the substrate for the >100mA boron flux used for the DRAM p+ poly doping is ~6W and <1W for the shallower, lower energy USJ doping applications. For both cases, the wafer remains near room temperature and no wafer cooling is required. These results in combination with the lack of B pile-up on the surface described earlier and no observable hydrogen out-gassing effects allows for a photoresist-friendly process requiring no special or aggressive stripping steps, even for the very high doses (>1E16/cm2).
Manufacturability
Significant advances have been made at Epion to demonstrate the readiness of infusion processing for high-volume manufacturing. A comparison of equivalent beam currents as a function of equivalent monomer beam energies is shown in Fig. 4. Assumptions were made that equivalent monomer beam energies of 0.1, 0.2, 0.4, and 0.6keV result in Xj values of ~7.5, 12, 20, and 28nm at 1E18/cm3 B levels, respectively. The appropriate infusion energies were used to match these Xj values. Other process performance parameters such as uniformity of <0.5%, metal contamination of <5E9 for all metals, and uptime of >200hrs with wafer-to-wafer repeatability of <2% (1σ) and low particle levels all appear production-worthy and are described in detail elsewhere [11].
Device performance
The advantages of infusion’s ultra-abrupt profiles and lack of EOR are observed in device properties. Renesas has measured improved short channel effects (SCE) for 50nm gate length pMOSFETS using 5keV B infusion in comparison to 200eV implants [12, 13]. Ho Lee and colleagues at Samsung [9] have observed improved SCE using GCIB infusion. They also measure a 4× reduction in STI-bounded p+/well junction leakage of infusion-doped samples compared with the BF2+Ge-PAI implantation for a standard 90nm logic process with 50nm physical gate lengths. Using a hybrid laser spike anneal (LSA) and a reduced temperature spike rapid thermal anneal (RTA), infusion-doped ultra-shallow junctions with Xj ≈ 18nm and Rs ≈ 800Ω/sq were produced exhibiting superior activation, lower drain induced barrier lowering (DIBL) (120mV/V), and a systematic inversion Tox reduction compared to the control implants. Specific halo and spacer schemes were needed for preserving low parasitic series resistance, gate induced leakage, and Vth (threshold voltage) variation in the MOSFETs designed for shallow and abrupt extension junctions.
Conclusion
The distinct physical differences between the doping mechanism of infusion and the other implant technologies represent several advantages for USJ and DRAM manufacturing for GCIB technology. Among the advantages are: 1) mass-independent doping for simultaneous co-doping of multiple species to the exact same depths; 2) E1/3 doping depth dependence resulting in uniquely high (>100mA) B fluxes and extremely shallow and abrupt doping profiles; 3) complete self-amorphization with no evidence of EOR damage eliminating the need for PAI and improving leakage performance; 4) no self-sputtering effects allowing for a one-to-one correlation between GCIB dose and retained B dose up to 1E17/cm2 doses; and 5) a low power ion beam allowing for a room temperature process with photoresist compatibility.
References
- M.E. Mack, Nucl. Inst. Meth. B237, p. 235-239, 2005.
- M.E. Mack, et al., IIT 2002 Proceedings, IEEE, Piscataway, p.665, 2003.
- J. Bachand, et al., IIT 2002 Proceedings, IEEE, Piscataway, p. 669, 2003.
- J. Borland, J. Hautala, M. Gwinn, T.G. Tetrault, W. Skinner, “USJ and Strained-Si Formation Using Infusion Doping And Deposition,” Solid State Technology, p. 53, May 2004.
- J. Borland, Nucl. Inst. Meth. B237, p. 6-11, 2004.
- J. Borland et al., “Shallow and Abrupt Junction Formation: Paradigm Shift at 65-70nm,” Solid State Technology, p. 83, June 2002.
- J. Borland, et al., Semiconductor International, January 2005.
- D. Jacobson, Ext. Abs. IWJT, S2-1, 2005.
- Ho Lee et al., IWJT 2006, Shanghai, China, IEEE Press.
- J. Borland et al., 8th Intl Work. on Ultra-Shallow Doping, p. 201-208, June 2005.
- W. Skinner, J. Hautala, M. Gwinn, T. Kuroi, Advanced Semiconductor Manufacturing Conference Boston, MA, IEEE Press, 2006.
- T. Yamashita, et al., Ext. Abs. IWJT 2005.
- T. Kuroi, Y. Kawasaki, 8th International Workshop on Ultra-Shallow Doping, p. 4-9, June 5-8, 2005.
Contact John Hautala, CTO, at Epion Corp., 37 Manning Rd., Billerica, MA 01821; ph 978/215-6238, e-mail [email protected].