Properties of octadecaborane-based ULE boron implants
08/01/2007
The materials properties of junctions formed by ULE implants of B18H22+ provide an opportunity to significantly increase productivity for all low-energy p-type implants without the use of deceleration and the associated risk of energy contamination. B18H22 implants also form an amorphous layer due to the large mass of the material. We have found that the thickness of the amorphous layer is dependent on the beam current of the implant and that, upon millisecond annealing, results in a shallow junction that is free of end-of-range (EOR) damage.
Other issues critical to integrating B18H22 into standard CMOS processing include dose retention effects during implantation and subsequent photoresist strip. Dose retention and sputtering of silicon have been examined using secondary ion mass spectrometry (SIMS) for implants ranging from 50eV to 2keV. The challenge for photoresist strip is to not remove dopant that is placed within 1-2nm of the surface. The use of fluorine-free downstream ashing in combination with proper wet-cleaning techniques is shown to minimize undesirable dopant loss from the cleaning process.
Introduction
Molecular implants for n-type doping applications have been in production for several years now [1] and are known to yield equivalent device performance while improving beam currents for low-energy applications. For example, As2+ implanted at 2keV will have nearly double the throughput of the monomer As+ implanted at 1keV due to advantages realized in beam transport.
Ultra-shallow junction (USJ) requirements at the 45nm node and below have driven the need for production-worthy p-type implants at energies below 1keV. Implants of B18H22 can be helpful in meeting productivity requirements for all applications requiring low-energy boron implants.
Because each B18H22+ ion contains 18 boron atoms, the particle beam current is 18× the electrical beam current, so this simultaneously increases productivity while minimizing the risk of charging damage. Additionally, the high molecular weight of B18H22 means that the extraction energy is ~20× the desired equivalent B energy, making possible implants as low as 100eV B equivalent without using decel. Figure 1 compares the relative particle beam currents available today using B18H22 and atomic boron.
We studied the implanting, cleaning, and activating implants of B18H22+ and investigated the junction properties using TEM after millisecond anneals. We also quantified sheet resistance/junction depth (R/X) characteristics, dopant retention, and properties of cleaning photoresist from ultra-shallow implants.
Experiments
Implants were performed using B18H22 as the precursor [2]. The solid material is delivered through a heated manifold into an electron capture-type ion source and extracted into a standard implanter beam line capable of mass resolving up to ~4.0keV equivalent boron energy.
We implanted B18H22 in the range of 5×1014-2×1015 boron atoms/cm2 at energies ranging from 1.0-40keV (50eV-2.0keV equivalent boron energy). Samples were analyzed using SIMS for the dopant profiles, ThermaWave (TW) for evaluation of implant damage, and cross-sectional transmission electron microscopy (XTEM) for amorphous layer thickness and EOR damage evaluation.
For the ashing studies, a downstream microwave asher using O2 chemistry was used [3]. This asher may also be configured for fluorine chemistries in order to evaluate aggressive ashing on the dopant. Additionally, standard wet-chemistry techniques including APM and SPM dips were used to evaluate wafer cleanliness, following implant and ash steps. Millisecond anneals were performed at 950°C and 1050°C using proprietary flash techniques.
Results
Implants were carried out on 300mm n-type wafers and analyzed for as-implanted profiles using SIMS. Figure 2 shows the measured profiles for B18 implants ranging from 2keV down to 50eV in energy. The profiles are well-behaved and consistent with previously reported profiles of B18 implanted into crystalline silicon. We believe that the 50eV profile has been artificially deepened due to ion beam mixing effects during the SIMS analysis and that the true profile is even shallower. The range (R) of the implants was calculated using the stopping range of ions in matter (SRIM) [4]. Good agreement between the model and extracted values was observed over the entire energy range of the implants.
 Figure 2. SIMS profiles of B18H22+ implanted into c-Si. |
The B18 implant tails are shallower than equivalent boron implant due to formation of an amorphous layer during the initial stages of implant that suppresses channeling, leading to a shallower junction (X) [5]. If the silicon is pre-amorphized prior to implant, the B18 and B profiles appear and behave identically upon annealing.
We have also quantified the dose retention of the boron as a function of implant energy. It is well known that the implantation process may become convoluted with competing processes that can reduce the total dose retained in the sample. These processes, which include sputtering and backscattering, have been reported in the literature for low-energy B+ and BF2+ [6], as well as sputtering rates for higher energy B18H22+ implants [7]. As the energy of the process becomes lower, these competing effects will nominally increase the required dose, and could ultimately limit the total dose achievable for that energy. At energies >1keV, 100% of the implanted dose is retained in the sample, but shows a significant reduction as the energies approach 50eV. Approximately 55-80% of the dose is retained for 50, 200, and 500eV implants, respectively, so this, coupled with typically small beam currents at these values, will remain a process control and productivity challenge for implementing all extremely low implant processes.
We observed a dependence of TW and R as a function of beam current for the 500eV implant. ThermaWave is an indicator of damage introduced into the Si lattice by the implant. There was a 4.8% decrease in R and 5.5% increase in TW when going from 1.8mA to 16mA of boron current. While the ThermaWave response may not be reliable when an amorphous layer is formed, the trend correlates to the R value for the range of beam currents investigated. We believe this response is due to the damage and resulting amorphization of the lattice during implant, with higher beam currents producing more damage and a slightly thicker amorphous layer.
 Figure 3. XTEM examination of EOR damage following a flash anneal for B18H22+ implants at 1E15B/cm2 dose. |
Figure 3 shows a series of XTEM measurements examining EOR damage after a flash anneal, compared to implants preceded by a pre-amorphizing Ge implant. The figures indicate that the flash anneal can successfully anneal away damage introduced by the B18H22, while deeper damage introduced by the Ge still contains EOR damage.
Results from ashing a 500eV 1E15/cm2 dose B18H22+ implant in a downstream microwave asher and following up with either an oxidative H2O2/H2SO4 or reducing HF clean were done to investigate retained boron dose. The sheet resistance decreased by 8% for oxygen ashing followed by oxidative clean chemistry indicating a higher concentration of boron remains in the sample compared to the as-implanted and annealed sample. A 16% increase in R was observed for oxygen ash followed by HF clean, a 46% increase was seen with CF4 ash and HF dip, and a 63% increase was seen with CF4 ash followed by H2O2/H2SO4 clean. Since the oxygen plasma treatment is known to form a thin oxide layer, we postulate that this layer prevents out-diffusion of the boron upon annealing, resulting in a lower R value. The HF-dip post ash removes the oxide layer (along with some implanted B); subsequent annealing results in lower B concentrations in the sample. In the case of fluorine containing ashes, no oxide layer is formed, and Si is removed during the ashing, thus a higher R value is observed for this treatment.
 Figure 4. SIMS profiles indicating ashing chemistry effects on low-energy boron profiles. |
Figure 4 shows the SIMS profiles of post-implant ashed and annealed samples compared to the as-implanted samples. Again, we note that the oxygen-ashed sample has a slightly higher concentration of B than the anneal only sample, validating the role of a thin oxide layer in boron retention. For the fluorine-containing ashes, which are often used for implanted photoresist crust breakthrough and residue removal, a strong dependency on F concentration is found. For high F concentrations, almost all of the B has been removed from the sample. These studies indicate the importance of properly designing the post-implant clean sequence: formation of a thin oxide protective cap to prevent B out-diffusion during anneals, elimination of F-chemistries during the ash, and use of only oxidizing cleans during wet treatment.
Conclusion
We investigated the use of molecular implants of B18H22 for the formation of USJs. For very low energies, the dose retention was <80%, consistent with previous findings. The amorphous layer formed during implant caused differences in the annealing behavior when compared to standard B+ implants. Flash anneals yielded shallow junctions without the presence of EOR damage; however, this damage was observed in samples pre-amorphized with Ge. We found that ultra-shallow implants are very sensitive to the post-implant ash and wet clean treatments, though good results could be obtained through the use of O2-based ashing and oxidative wet cleans. The high level of productivity offered by the molecule coupled with good process results make B18H22 an excellent candidate for sub-65nm junction formation.
References
- B. Chang, et al., “Arsenic Dimer Implants for Shallow Extension in 0.13µm Logic Devices,” 14th International Conference on Ion Implantation Technology, edited by B. Brown et al., p. 111, 2004.
- D. Adams, et al., “A Vaporizer for Decaborane and Octadecaborane,” 16th International Conference on Ion Implantation Technology, edited by K. J. Kirby et al., AIP Conf. Proc. 866, p.178, 2006.
- A. Srivastava, et al., “Ultra-shallow Junction Cleaning: Methodologies for Process and Chemistry Optimization,” 8th International Symposium on Ultra Clean Processing of Semiconductor Surfaces, p. 52, 2006.
- SRIM: The Stopping Range of Ions in Matter, http://www.srim.org/.
- L.A. Marquis, et al., “Characterization of Octadecaborane Implantation into Si Using Molecular Dynamics,” Phys. Rev. B 74, 201201, 2006.
- A. Agarwal, “Ultra-Shallow Junction Formation Using Conventional Ion Implantation and Rapid Thermal Annealing,” International Conference on Ion Implantation Technology, edited by H. Ryssel et al., p. 293, 2004.
- M.A. Harris et al., “Dose Retention Effects in Atomic Boron and ClusterBoron (B18H22) Implant Processes,” 16th International Conference on Ion Implantation Technology, edited by K. J. Kirby et al., AIP Conf. Proc. 866, p.155, 2006.
Acknowledgment
ClusterBoron is a trademark of SemEquip.
Mark A. Harris received his bachelors and masters in chemical engineering from the U. of Idaho and is process technology manager for the Ion Implant Division at Axcelis Technologies, 108 Cherry Hill Drive, Beverly, MA 01915 United States; e-mail [email protected].
Chuong Huynh received his bachelors in materials science and engineering from the U. of Pennsylvania and is a process engineer at Axcelis Technologies.