Charge Control for high-current ion implant
06/01/1997
Charge control for high-current ion implant
Yuri Erokhin, Ronald N. Reece, Robert B. Simonton, Eaton Corporation, Beverly, Massachusetts
The electrostatic potential of high-current positive ion beams may reach hundreds of volts unless the space charges are compensated with electrons. Potential in excess of 10 V induced over sub-100-? gate oxides will deteriorate oxide integrity. Thus, all high-current implanters incorporate either secondary electron flood (SEF) or plasma electron flood (PEF) hardware to generate and inject electrons into the ion beam to reduce beam potential.
This paper discusses optimum conditions for wafer-charging control during ion implantation, including inherent limitations of beam neutralization through residual gas ionization. We compare PEF and SEF, and suggest a simple test that determines whether electrons with energies higher than 10-20 eV are generated and transported to wafers.
Experimental data show that device structures influence implant-induced oxide wearout. To minimize wearout of thin-gate oxides, the optimization of charge neutralization conditions should take into account the conductivity of the Si substrate and the well structure.
Ion implantation is critical for semiconductor manufacturing. Present generation high-current implanters are capable of delivering >20 mA to the wafer, with significant positive space charge. Beam electrostatic potential (not to be confused with ion energy) can reach up to a thousand volts, unless the beam`s positive space charge is neutralized by the injection of electrons from an external source.
Unneutralized or underneutralized ion beams are difficult to transport due to beam divergence induced by electrostatic repulsion between ions [1]. A divergent beam can strike beamline components, generating and trapping secondary electrons that temporarily neutralize part of the beam. These changes adversely impact implanter dosimetry, and cause both dose errors and reduced dose uniformity. It may also result in an increase in particles and wafer contamination.
Unneutralized, high-electrostatic-potential beams can charge insulating components on the wafer up to full beam potential [2]. This charging is enhanced by ion-induced emission of secondary electrons from the wafer. If the potential at the device level is not carefully controlled, either catastrophic oxide failure or a decrease in the voltage to breakdown (VBD) or charge to breakdown (QBD) may occur. The thinner gate oxides (<100 ?) associated with current-generation processes are considerably more sensitive to these failure modes than the relatively thick gate oxides employed in previous generations [3].
 |
Figure 1. Charge distribution within the ion beam in a high-current implanter.
Large interconnect areas result in large antenna-to-gate ratios, which enhance the collection and buildup of electric charge across gate oxides. Large antenna-to-gate ratios complicate charging control and require more precise beam neutralization [4]. For all of these reasons, the charging-control system must neutralize the positive beam with low-energy electrons.
Charging-control technology has to evolve constantly to keep pace with the demands of new device generations. This paper discusses the performance of SEF and PEF systems and the role of residual gas in beam neutralization. Design limitations of older SEFs and specific performance enhancements to optimize the generation of low-energy electrons are presented. For present device generations, an optimized SEF provides performance comparable or superior to a PEF, as shown by the electron energy distribution and experimental charge to breakdown (QBD) data obtained from specially fabricated, charge-sensitive structures.
Beam charge distribution
The net density (positive minus negative) and the distribution of charged particles determine the electrostatic beam potential. Once electrons and ions from all sources are accounted for, it is clear that the charge balance above the wafer depends on many processes that generate electrons and ions in the implanter beamline (Fig. 1).
Residual gas in the beamline can be ionized by the principal ion beam. Both ionization of residual gas and secondary electron emission from the wafers or disk generate and inject electrons into the beam. The other source of electrons is charge-control hardware (SEF or PEF).
Residual gas ionization is a very effective source of low-energy electrons to neutralize the ion beam [5]. Although the total charge created during ionization of residual gas is zero (electrons minus cations), the electrons are trapped by the positive beam potential, while positively charged residual gas ions are repelled from the beam. Simply bleeding in gas up to a beamline pressure in the range of 10-5 torr can bring the beam potential down from a few hundreds to a few tens of volts. Unfortunately, if the pressure is increased beyond the 10-4 torr range, the residual ionized gas molecules are no longer repelled by the reduced beam potential, and the net effective charge (from disassociated electrons) fails to increase.
 |
Figure 2. The potential difference (DV) reflects the average energy of electrons generated by the charging control flood. For a given fixed electron energy, one can tune the flood to produce one of three possible charging responses: overflooding, underflooding, and balanced (cross-) flooding conditions.
Nevertheless, beam gas bleed is a very important tool for charging control. It injects low-energy electrons (eV range) directly into the beam at a rate proportional to local ion beam current density. Further beam neutralization, down to a potential of a few volts, can be provided by either SEF or PEF [1]. PEF inherently incorporates gas bleed (typically Ar or Xe) and cannot function without it. Latest generation SEF may also be used with gas bleed.
Distribution of negative and positive charges across the beam is not uniform. Positive charges dominate close to the beam center, while electrons outnumber them at the beam perimeter (Fig. 2) [6, 7]. This happens because of ambipolar diffusion and because SEF or PEF electrons are introduced at the beam perimeter. Separation between positive and negative charges creates a field that slows down electron diffusion and prevents them from leaving the beam. SEF or PEF electrons injected at the beam perimeter that are not immediately thermalized by the beam plasma are trapped by the positive beam potential. These electrons oscillate around the beam center, and when they approach it, their potential energy decreases, but their kinetic energy reaches a maximum.
 |
Figure 3. Possible mechanisms for transport of high-energy electrons in an old SEF design.
The magnitude of the beam-potential swing (U) across the diameter increases with the electron energy. Thus, low-electron energy is extremely important for beam neutralization. Conditions when positive and negative swings of the beam potential are equal are referred to as "cross-over" conditions (Fig. 3). When positive swing is greater, it is referred to as "underflooding" with electrons; when negative swing is greater, it is termed "overflooding."
Different charge-sensitive structures on implanted wafers respond differently to positive and negative voltage stresses. Therefore, depending on the device structure, either underflooding or overflooding can achieve the lowest gate oxide wearout.
Experimental procedure
An Eaton NV-GSD/200E high-current ion implanter was the test vehicle for experiments. Implant processes were selected consistent with mainstream process technology. All implants were 20-80 keV As+ ion beams with currents ranging from 10-20 mA, which is common for source/drain and poly-Si doping. Doses ranged from 1 ? 1015 to 1 ? 1017 ion/cm2.
The implanter is equipped with active charge monitors on the implant disk. These capacitive devices, similar to Langmuir probes, measure the positive and negative beam potentials during process. The charge monitors help characterize the effects of SEF and PEF setting variations, as well as other system parameters.
These experiments used capacitors fabricated by depositing poly-Si over 90-? gate-oxide test structures. The poly-Si areas act as antennas or charge collectors. The structures had antenna-to-gate ratios ranging from 10:1 to 100,000:1. The largest ratio antennae were the focus of this work because of their extreme sensitivity to charging conditions.
The charge sensitive structures were tested using semiconductor parametric test apparatus under constant current conditions. The product of the current forced through the oxide (through Fowler-Nordheim tunneling) and the time before breakdown determined implantation-induced oxide wearout. Unimplanted structures had QBDs of 10 (?10%) Coulomb/cm2.
Secondary electron flood
In SEF, primary electrons with energy of a few hundred eV are directed to a target located close to the ion-beam path [2]. When typical 300 eV primary electrons strike a target, a cloud of low-energy secondary electrons is generated. The energy distribution of secondary electrons peaks at about 10 eV with virtually no electrons above 20 eV. Secondary electrons are trapped by the ion beam, reducing its electrostatic potential. The number of secondary electrons and the extent of beam neutralization are controlled by varying the current of the primary electrons.
For optimum SEF performance, it is crucial that no high-energy primary electrons are transported toward the wafer, or overflooding may damage gate oxides. If the ion beam is neutralized with only low-energy secondary electrons, then no structure on the wafer will acquire a negative potential higher than the maximum electron energy.
As negative wafer potential increases, fewer electrons are able to reach the wafer. Therefore, if only low-energy electrons are transported to the wafer, then SEF operates in a self-regulating mode. However, if some high-energy primary electrons reach the wafer, then not only can excessive potential develop, but SEF extension tube aging can degrade process control.
 |
Figure 4. Cut-off voltage for electron current with old and new SEF designs. Electrons striking the disk are measured as positive current. The negative current represents the loss of electrons from the disk (most likely due to secondary electron emission enhanced by elevated disk potentials). Beam off, e-shower gas off, -2.5 kV bias on.
The following experiment demonstrates high-energy electron transport in an early SEF design. Gate-antenna charge-sensitive structures implanted at different settings of the SEF primary current produced QBD and beam potential swing (measured by the charge monitor) data. Oxide wearout reached a minimum and then increased rapidly with the SEF primary current. This response suggests that overflooding due to high-energy electrons is the failure mechanism at elevated levels of SEF primary current.
Analysis of SEF design [8] identified several factors that may contribute to the transport of high-energy electrons to the wafer. The filament emitting primary electrons was usually located in close proximity to a negatively biased aperture (-2500 V) used to stabilize beam-transport conditions. The electric field of the biased aperture could deflect some primary (300 eV) electrons toward the wafer surface. The emission angle of the electron gun may have also allowed primary electrons to strike an extension tube located near the wafer. The inner surface of the extension tube gets coated with sputtered photoresist from the wafers, and reaches a relatively high negative floating potential if struck by primary electrons. The insulating layer then generates additional secondary electrons (Fig. 3) that would be accelerated by the floating potential of the resist surface.
The newly designed SEF almost completely suppresses the transport of high-energy electrons toward the wafer. Moving the biased aperture further away from the filament eliminates stray electric fields. A redesigned biased aperture uses a magnetic electron reflector to decouple ion-beam neutralization in the beamline and near-wafer regions. Decoupling allows for the reduction of the aperture bias voltage to around 1000 V. Optimization of the filament-extraction aperture and filament height focuses the primary electron beam on the target area with minimal dispersion. Finally, an extension tube with a serrated inner surface prevents the insulating coating from being deposited continuously.
The energy distribution of electrons generated by the SEF was qualitatively estimated by measuring implant disk current vs. negative disk bias. The SEF primary current was 400 mA during this test and the ion beam was off. Without the beam, low-energy electrons generated by the SEF are not transported to the disk but high-energy electrons, if generated and directed toward the disk, will be able to reach it. The improved SEF design suppresses the transport of high-energy electrons by more than an order of magnitude compared to the old SEF design (Fig. 4).
Plasma electron flood
PEF operation is based on the extraction of electrons from a discharge (usually Ar or Xe) maintained in an arc chamber (Fig. 5). Essentially, the PEF operates similarly to the implanter ion source, but is biased to extract electrons instead of positive ions. PEF also operates with zero to only a few volts for extraction, rather than the kV range typical for ion sources. The positive potential of the beam extracts adequate electron current from the PEF arc chamber (located close to the beam path).
A quasineutral plasma, consisting of positively charged ions and electrons, leaks from the arc chamber to the beamline. This plasma is often called a "plasma bridge." The positive beam potential (and the extraction voltage, if used) extracts electrons through the plasma bridge. The plasma bridge`s quasineutrality overcomes space-charge limitations on the magnitude of the extracted electron current, and allows effective injection into the beam of 5-6 eV electrons (in most cases below 10 eV). Low electron energy is the main advantage of the PEF compared to SEF; no high-energy electrons are generated within the PEF system.
 |
Figure. 5. Schematic of the PEF.
With the increase of the beam current, the PEF extraction electron current also increases proportionally (Fig. 6). PEF extraction voltage was fixed for this experiment, so the positive beam potential had to rise in order to drive higher electron current from the PEF discharge chamber. This rise slightly accelerates extracted electrons, increasing the negative swing of beam potential as well.
Extraction electron current remains lower than the ion beam current. Additional neutralizing electrons come, most likely, from beam-induced ionization of PEF gas bled into the implanter beamline. Adjustment of the PEF extraction voltage (for underflooding, cross-over, or overflooding) fine tunes beam neutralization to minimize damage in devices with different sensitivities to positive and negative voltage stress.
 |
Figure 6. Dependence of the PEF electron extraction current and beam potential from the ion beam current for 80 keV As+ implantation. The PEF arc voltage and current were 28 V and 4 A; the extraction voltage was 5 V.
Influence of device design on implant damage
Beam-induced voltage can force electrons to tunnel through gate oxide. Electron tunneling, in turn, generates permanent charge traps within the oxide bandgap. These traps increase the leakage current through the gate oxide and ultimately manifest as gate oxide wearout. Nevertheless, beam characteristics constitute only one group of factors that affect implantation-induced gate-oxide wearout.
The type of silicon conductivity (p- or n-), the well structure below the oxide [9], and the secondary electron emission of materials (photoresist, polysilicon) surrounding the gate are significant factors [10] that determine gate-oxide sensitivity to positive and negative voltage stress. Beam neutralization for minimum gate-oxide wearout does not necessarily require that positive swing is equal to negative (cross-over). For some structures, the charging damage will be minimal under slight underflooding, while overflooding is optimal for others.
Positive voltage forms a depletion region in p-type Si and an accumulation region in n-type Si, with the opposite for negative voltage stress. In an accumulation region, the entire beam-induced potential drop is through the gate oxide that leads to increased wearout. In a depletion region, the beam-induced potential is divided between the depletion region and the gate oxide, with the voltage-drop distribution depending mostly on the ratio of gate oxide and depletion region capacitances.
Poly-Si gate structures tend to emit more secondary electrons compared to adjacent photoresist-coated areas (Fig. 7), which also influences wearout sensitivity to positive and negative charging. Higher emission of electrons accelerates buildup of positive charge and thus elevates positive voltage stress. Voltage induced at the gate stack may even exceed the beam potential, because the size of the gate structure is usually very small compared to the beam plasma Debye length (typically in the millimeter range).
 |
Figure 7. Increased beam-induced potentials on the areas with high coefficients of secondary electron emission.
Charge-sensitive structures with 90-? gate oxide and a 100,000:1 antenna-to-gate ratio were implanted with 80 keV As+ ions to a dose of 2 ? 1016 cm-2. Low-energy electrons, generated by the new SEF at various primary electron currents, neutralized the ion beam. The structures were more sensitive to positive than to negative voltage stress (Fig. 8).
We cannot conclude from this single experiment whether the difference in sensitivity was due to the formation of accumulation-depletion regions or due to variations in secondary electron emission. Nevertheless, these structures will clearly experience higher oxide wearout if implanted at cross-over conditions instead of the more optimal overflooding.
 |
Figure 8. The charge monitor response and QBD performance at different primary electron currents for a new SEF with suppressed transport of high-energy electrons toward the wafer. (The QBD for unimplanted structures was 10 C/cm2.) The implantation of 80 keV As+ ions was at 12 mA with an antenna:gate ratio of 100,000:1. QBD performance is robust over a very wide range of primary current settings.
The new SEF design prevents primary electrons from reaching wafers, and thus dramatically extends the SEF operating window. The QBD performance is robust over a wide range of primary currents with no indication of QBD degradation as a function of increasing primary current, a characteristic of the previous design. The QBD performance of the new SEF design is comparable to commercially available PEF systems. Performance is consistent over a wide range of conditions, including an extremely high-dose, 1 ? 1017 ion/cm2, 20 mA, 80 keV arsenic implant.
Conclusion
This work demonstrates optimization of a SEF design to allow only low-energy electrons to reach the wafer. Simple tests can estimate the energy distribution of electrons reaching implanted wafers. The results indicate a wide operating window for varied SEF primary currents. Performance improvements were quantified using QBD structures.
The new SEF design minimizes gate-oxide damage under extreme implantation conditions (dose, beam current, antenna:gate ratio). Neutralization results are comparable (or superior for some implant conditions) to commercial PEF designs. Actual device structures have different sensitivities to positive and negative voltage stress, so beam neutralization conditions must be optimized for each device.n
Acknowledgment
The authors would like to thank Peter Kellerman for his advice and contribution to the project, and Brian Freer and Jim Bernstein for their experimental contributions.
References
1. A.J.T. Holmes, "Beam Transport," Radiation Effects, Vol. 44, p. 47, 1979.
2. M. Mack, "Wafer Cooling and Charging in Ion Implantation," in Ion Implantation Technology, ed. J. Ziegler, North Holland Publ., p. 601, 1992.
3. H. Shin, K. Noguchi, C. Hu, "Modeling Oxide Thickness Dependence of Charging Damage by Plasma Processing," IEEE Electron Device Letters, Vol. 14, No. 11, p. 509, 1993.
4. H. Muto, H. Fujii, K. Nakanishi, S. Ikeda, IEEE Transactions on Electron Devices, Vol. 38, p. 1300, 1991.
5. A.J.T. Holmes, "Theoretical and Experimental Study of Space Charge in Intense Ion Beams," Physics Review A, Vol. 19, p. 389, 1979.
6. F. Sinclair, "Gate Oxides in High Current Implanters: How Do They Survive?" Nucl. Instr. and Meth., Vol. B55, p. 115, 1991.
7. M.C. Vella, W. Lukaszek, M.I. Current, N.H. Tripsas, "Plasma Model for Charging Damage," Nucl. Instr. and Meth., Vol. B96, p. 48, 1995.
8. R.N. Reece et al., "Optimizaion of the Secondary Electron Flood Design for the Production of Low Energy Electrons,"Proceedings of VII International Conference on Ion Implantation Technology, Austin, TX, June, 1996, in press.
9. C.P. Wu, F. Kolondra, "Wafer Charging Control in High Current Ion Implanters," J. Electrochem. Soc., Vol. 138, p. 3100, 1991.
10. K. Lai, K. Kumar, A. Chou, J.C. Lee, "Effect of Oxide Exposure, Photoresist and Dopant Activation on Plasma Damage Immunity of Ultrathin Oxides and Oxynitrides," IEDM, p.319, 1995.
YURI EROKHIN received his MS degree in quantum electronics from Moscow Physical Technical Institute, and his PhD degree in materials science from the Russian Academy of Sciences, where he worked on ion implantation into silicon. For more than 10 years, his research has focused on implantation-induced damage in Si and silicide formation on crystalline and preamorphized silicon. He has authored or coauthored more than 30 articles and a textbook chap
ter in the field of ion implantation. He joined Eaton in 1995, and is presently product applications manager. Eaton Corp., Semiconductor Equipment Division, 108 Cherry Hill Dr., Beverly, MA 01915; ph 508/921-9806, fax 508/927-3652.
RONALD N. REECE received his BS degree in electrical engineering technology from Northeastern University. He has worked in semiconductor process engineering for over 20 years, specializing in thin-film and ion implantation technology. He joined Eaton in 1990, where he is applications engineering manager.
ROBERT B. SIMONTON received his BS degree in materials science and engineering from MIT, where he specialized in electronic materials processing. For the past 19 years, he has held product engineering, advanced product development, and process application positions with ion implantation equipment manufacturers. He has authored or coauthored more than 40 articles and three textbook chapters on ion implantation. He joined Eaton Corp. in 1986, and is presently director of marketing.