Conventional beamline implantation of decaborane*
06/01/2001
*Based on "Decaborane Ion Source Demonstration"
by Michael C. Vella, Michael Reilly, Bob Brown. Original paper presented at IIT2000.
Alex Perel, principal scientist, Axcelis Technologies
overview
Low-energy boron implantation is a critical process for semiconductor doping today [1]. Decaborane has been shown to be an attractive alternative to monatomic boron for shallow junction formation. Motivated by the requirement for these ultra-shallow junctions, researchers have demonstrated the conventional application of decaborane beams.
In 1997, Kyoto University researchers reported decaborane implants without mass analysis, followed by successful activation [2, 3]. More recently, work performed at the New Jersey Institute of Technology has quantified transient enhanced diffusion of boron with very low-current, analyzed decaborane ion beams [4]. Lucent Technologies has implanted scanned wafers and characterized devices [5].
 |
Figure 1. Decaborane ion spectrum, B10Hx+ and B9Hx+ at 50kV. The effects of arc parameters and vaporizer temperature are shown. Note that the second high point at mass 118 in the 0.95A, 60V data was taken after the rest of the spectrum was scanned, implying that the vaporizer temperature had drifted higher.
A commercially viable delivery system is available [6]. In principle, this could be deployed on any modern high-current implanter in a manner that is interchangeable with standard hardware and processes. Commercialization has been hampered by the lack of a high-current ion source, however. Early attempts to ionize decaborane in Bernas sources were informally reported as failures, due to dissociation of decaborane and heavy boron residues.
In the work presented here, a modified Bernas source was mounted on an NV-10/80 implanter. Using standard extraction and beamline components, 2.3mA of boron nucleon current was produced in the form of B10Hx+ at 50kV. The beam was routinely sustained for a shift (6-8 hr). Due to under-dense plasma, beam current scaled linearly with extraction voltage. Commercial refinement of the delivery system and ion source is underway. With modern extraction optics, commercially attractive boron current below 1keV/nucleon is expected.
Space charge molecular beam scaling
High-current, low-energy beam extraction and transport is limited by basic physics. For example, consider the implications of the Child-Langmuir limits for extraction of space-charge-dominated beams [7], where beam current density, J (A/cm2), is proportional to the square root of charge to mass ratio, (Z/A)1/2, and the three-halves power of the extraction potential, V3/2, but inversely with the square of the gap, d2 (cm2).
The numerical coefficient is model-dependent, but the parametric scaling contains the essential physics. The strong scaling with potential severely limits low-energy beam current. In the past, tool vendors have used adjustable extraction gaps to maximize extraction current down to a few kV. Gap reduction is limited by high-voltage sparking, however.
Below 10kV, reduced potential begins to overwhelm the practical limits of a reduced gap. Some compensation can be gained by increasing the extraction area, but for space-charge-limited beam transport, this translates into a larger, more expensive beamline and analyzing magnet. Without innovation, standard implanter architecture will become an increasingly expensive method for delivering high-current dopant below 1kV.
A simple example illustrates the challenge and theoretical advantage of decaborane.
Consider a hypothetical ion source/extraction system capable of 10mA of 11B+ at 10kV extraction. For convenience, assume that the extraction area is 10cm2. Using the scaling of J = 5.44 x 10-8 (Z/A)1/2V3/2/d2, the maximum 11B+ performance of this system with fixed gap would be 0.3mA at 1kV. Decaborane benefits from the double advantage of 10kV extraction, and 10x dopant nucleons/charge. The exact nucleon energy depends on the ion mass spectrum, as discussed later. So, the expected performance is illustrated in the table by assuming one-tenth the energy for the boron nucleons. With a fixed gap, the decaborane current at 10kV would be 3mA, which corresponds to 30mA of boron nucleon current.
 |
For a given dopant energy, both nucleon current and extraction potential are proportional to the number of nucleons in a molecule. Thus, J = 5.44 x 10-8 (Z/A)1/2V3/2/d2 implies surprisingly high leverage for molecular beams. For example, let the molecular mass be expressed as, Amole = Nd Ad + NndAnd, where Nd denotes the number of dopant nuclei/molecule, and "nd" denotes the average for nondopant species. If the desired dopant energy is Ed and a charge state of one is assumed, the extraction potential must be Vmole = Amole Ed /Ad. Taking the ratio of molecular dopant nucleon current to corresponding dopant ion current, the theoretical extraction-limited dopant nucleon current gain for molecules is Gmolecule = Nd (Nd + Nnd And /Ad). Note that gain in this equation is independent of extraction potential. For decaborane, Nboron = 10, Nnd = 14, and mnd/mboron = 1/11. So, the space-charge-limited current gain is Gdecaborane = 113. This remarkable result illustrates that no existing ion source comes close to the theoretical extraction limits for decaborane or other heavy molecular ions.
Alpha installation
A proprietary, modified Bernas ion source was installed on a 1980s vintage NV-10/80 implanter. The decaborane vaporizer was mounted in the gas box in place of a bottle. Modest heating was required for the delivery system and gas line on the atmospheric side to maintain temperatures above 20°C for decaborane evaporation and transport. The standard utility capabilities of the gas box were sufficient for the modest heating requirements of decaborane. On the vacuum side, the gas line was below 300°C, so most of the decaborane reached the ion source intact.
The extraction system and beamline were standard. Before decaborane testing began, the tool was qualified using a standard Bernas source. At the conclusion of testing, the standard Bernas was reinstalled, and the tool was qualified for production implant. The analyzed beam current was measured with the Faraday. The transfer ratio of the total B10Hx+ current was <1%, probably because the extraction optics were poorly matched to decaborane. The target wheel was not functional during our tests, but it has since been made operational for future decaborane implants.
The decaborane and delivery system were provided by ATMI [6]. The only difficulty with the delivery system was thermal lag, which limited temperature control to 1° or 2°. The nominal mass of decaborane is 124, but neither the boron nor the hydrogen constituents were isotopically pure, which smeared the spectrum. Microamps of beam current at masses above 124 were seen under some operating conditions.
Health and safety issues
Decaborane is a toxic solid, with low vapor pressure at room temperatures and a distinctive, unpleasant odor. Toxicity is lower than many other process materials. So, decaborane vapor has the advantageous characteristic of being detectable by its unpleasant odor before dangerous levels are reached. Cleaning and handling precautions are required, however, because decaborane can react explosively with oxidizing agents routinely found in fabs.
 |
Figure 2. Total B10Hx+ current scaled linearly with extraction potential at pv32C and pv38C. Solid symbols represent data points; open symbols represent V3/2 scaling.
Decaborane is readily soluble in alcohol. This makes cleanup easy, but excellent splash protection is required, because skin absorption is also hazardous. Since decaborane dissolves latex rubber, butyl nitrile gloves are required. A full face shield breathing apparatus is also recommended.
Refractory temperature ion source components reduce decaborane into its constituent elements. Elemental boron is difficult to clean, but not specifically hazardous. This experimental source was removed after approximately 20 hr of operation. No detectable vapor pressure was found in the source enclosure, even with a hydride monitor. The most obvious scenario for hazardous tool contamination would be an event where the decaborane delivery system (vaporizer and gas line) is energized with the arc off, with significant contamination of the source and enclosure. Thus, appropriate maintenance procedures are required.
Results
This project demonstrated feasibility of a high-current decaborane ion source suitable for commercial processing. To focus on the source, the extraction potential was usually set at 50kV, which is comfortable for the extraction optics of the NV-10/80. The dominant part of the spectrum was B10Hx+, which appeared to blend with the B9Hx+ peaks, as illustrated in Fig. 1. The total current in the B10Hx+ peaks was more than 2.3mA. The most important control parameter was the vaporizer temperature, which is indicated by the process value set point, e.g., pv38C means 38°C. Beam current was stable over a standard working shift (6-8 hr).
Decaborane extraction scaling is illustrated by plotting the total B10Hx+ current for 10-50kV. This summed molecular current scaled linearly with the extraction voltage, rather than following the usual V3/2, as illustrated in Fig. 2. Data with two temperature set points are shown, pv38C for the upper data and pv32C for the lower data set. Curves scaled by V3/2 from the maximum 50kV point are shown for comparison. The roll-off from 40-50kV in the pv32C data set is probably due to the vapor pressure being too low. The pv38C current scales linearly with extraction potential over a range of a factor of five.
Significantly higher boron-equivalent decaborane current, at lower extraction potential, is required for commercialization. This will require opening up the extraction aperture to maximize total beam current, which reduces mass-resolving power. Preliminary indications [5] are that this is a process-compatible approach, provided cross-contamination with heavy metals such as antimony (122) and indium (118) is not an issue.
References
- International Technology Roadmap for Semiconductors, Semiconductor Industry Association, San Jose, California, 1999.
- D. Takeuchi, N. Shimada, J. Matsuo, I. Yamada, "Shallow Junction Formation by Polyatomic Cluster Ion Implantation," Proc. 11th Conf. Ion Implantation Technology, pp. 772-775, 1997.
- M.A. Foad et al., "Formation of Shallow Junctions Using Decaborane Molecular Ion Implantation Comparison with Molecular Simulation," Proc. 12th Intl. Conf. Ion Implantation Technology, pp. 106-109, 1999.
- M. Albano, V. Babaram, J.M. Poate, M. Sosnowski, "Low Energy Implantation of Boron with Decaborane Ions," Mat. Res. Soc. Symp. Proc., 2000.
- D. Jacobsen et al., "Decaborane, An Alternative Approach to Ultra Low Energy Ion Implantation," Proc. 13th Intl. Conf. Ion Implantation Technology, 2001.
- L. Wang, ATMI, unpublished.
- S. Humphries, Jr., Charged Particle Beams, John Wiley & Sons, pp. 195-201, 1990.
Michael C. Vella is an ion source and accelerator designer at the Lawrence Berkeley National Laboratory and has authored several patents on related technology. He designed the magnet optics for the Berkeley Gas Separator used to discover elements 116 and 118. He has also served on the ETAB of the SRC and worked as a consultant to Electro-Graph and other semiconductor tool vendors.
Michael Reilly is VP of engineering at Electro-Graph. He joined the company in 1984 and previously served as manager of engineering and product development. He has more than 15 years of experience in the design, engineering, and manufacture of semiconductor-type ion sources and PECVD- and PVD-processing hardware. Electro-Graph Inc., 2365 Camino Vida Roble, Carlsbad, CA 92009; ph 800/824-8922, [email protected].
Bob Brown is currently a senior marketing specialist for ion implant source feed materials with ATMI Inc. Previously, he served as an ion implant technologist for IBM until his retirement in 1997. He has authored and co-authored several papers and articles on ion implant subatmospheric source feed materials and related technology. Brown was the chairman of the Fifth International Conference on Ion Implantation Technology that was held in Vermont in 1984, and is presently organizing the 14th International Conference on Ion Implantation Technology that will be held in Taos, New Mexico, in 2002.
Michael C. Vella, Lawrence Berkeley National Laboratory, Berkeley, California; Michael Reilly, Electro-Graph Inc., Carlsbad, California; Bob Brown, ATMI Inc., Danbury, Connecticut
__________________________
Critical temperature control for decaborane implantation
Decaborane is a fragile molecule that is extremely susceptible to dissociation under conditions found in a standard ion implanter. Large temperature gradients, vaporizers that are poorly controlled near room temperature, ion source temperatures near 1000°C, and high-density plasmas all conspire to make the vaporization, ionization, and successful transport of high ion currents of decaborane a daunting technical challenge.
These problems can be addressed by a decaborane vaporizer/ion source that meticulously controls the temperature from the vaporizer to the ion source and ensures that the walls of the ion source do not exceed the decaborane fragmentation temperature of 300°C. The plasma density within the ion source should also be kept low to maintain a large fraction of ions with 10 boron atoms.
 |
This kind of vaporizer and ion source technology, such as that developed by Axcelis, can also be used for other solid source species. As volume production of 130nm devices approaches, the use of high-mass species such as antimony and indium will become more prevalent. These species require greater attention to source temperature control than do boron, phosphorous, and arsenic. This will drive the need for advanced vaporizers and ion sources with better-controlled operating temperature ranges. The achievement of high beam currents for decaborane is an important milestone in the development of broad-range temperature-controlled vaporizer/ion sources.
Alex Perel, principal scientist,
Axcelis Technologies