- Email: [email protected]
High-voltage implantation facility at GM Research
NOMB
Nuclear Instruments and Methods in Physics Research B74 (1993) 13-17 North-Holland
High-voltage
implantation
Beam Interactions with Materials&Atoms
facility at GM Research
Gerard W. Malaczynski, Alaa A. Ehnoursi, Aboud H. Hamdi and Xiaohong Qiu Electrical
and Electronics
Engineering Department,
General Motors Research,
Warren, MI 48090-9055,
USA
This paper describes the effort undertaken at General Motors Research to build a plasma immersion ion implantation (PIII) system in order to study the potential of pulsed implantation as a means of modifying the tribological properties of automotive parts and to investigate the feasibility of producing surface modified components. The system has been designed for high dose ion implantation of irregularly shaped objects. It consists of a stainless steel vacuum chamber evacuated by a cryopump system, standard plasma diagnostic tools, a ring-cusp hot filament plasma source and a custom designed 150 kV, lo-20 t.~spulser. Details of the system are discussed together with justification for our choice of the specific components and the adopted techniques.
1. Introduction
Plasma immersion ion implantation (PI111 is a relatively new, non-line-of-sight technique for surface modification of materials. The method was originally developed by John Conrad at the University of Wisconsin at Madison [l] with the intention of surface modification for metallurgical applications. In the PI11 process, the workpiece is placed inside a vacuum chamber in which a plasma is generated. A series of negative high-voltage pulses are applied to this workpiece which results in the motion of ions toward and electrons away from the material to be treated. In effect, high velocity ion strike the biased target and penetrate its surface to a depth defined by the ion’s kinetic energy at impact and collisions with the lattice of the solid material. Charge separation caused by the strong applied electric field leads to a strong disturbance of the plasma’s electric neutrality. Eventually, the ion matrix sheath created by the initial movement of electrons away from the biased workpiece is consumed and the negative bias has to be terminated to recover the plasma. Therefore, to achieve continuous operation of the plasma immersion implanter, a high-voltage short duration pulser is required. Efficient plasma recovery, the desired ion density, ion species selection, penetration depth and implant density control for irregularly shaped objects, together with the necessity for extensive system control and diagnostic needed at this developmental stage (which must precede any practical application), makes the task of building a PI11 system a serious scientific and engineering challenge. This paper describes the effort undertaken at General Motors Research to build a PI11 system in order to 0168-583X/93/$06.00
0 1993 -
study this method’s potential in modifying the tribological properties of automotive parts and to investigate the feasibility of producing surface modified components. To formulate the requirements for a reasonably effective experimental system which would allow us to investigate various implants in a variety of materials, a series of initial experiments were performed with an ad hoc system composed of standard vacuum and high voltage components. The first implants were performed with a high voltage pulser operating at 35 kV [2]. The construction of this low voltage implant system rendered the necessary experience to formulate our list of desirable features, limitations, and the necessary conditions for efficient plasma implantation. In this initial learning process, our close cooperation and consultations with numerous members of the University of Wisconsin team was essential. This resulted in a successful “proof of principle” demonstration, which culminated in a nitrogen implant of a metallic foil, verified with Auger electron spectroscopy analysis [2]. In general, PI11 is a multidisciplinary project requiring knowledge of high-voltage engineering, vacuum technology, plasma physics, material science, and tribology. Since the intention of this paper is to describe our PI11 implanter, the discussion of material science and tribology will be limited and provided only when it is needed to justify a particular choice in the construction of the system. With so many complexities in the PI11 approach compared with the well-understood process occurring in line-of-sight implantation techniques, it is worthwhile to stress the uniqueness of the plasma process which goes beyond being able to achieve a uniform treatment of contoured targets. Specifically, by employ-
Elsevier Science Publishers B.V. All rights reserved
II. IMI’I.,ANTER SYSTEMS
14
G. W. Malaczynski et al. / High-voltage implantation facility at GM
ing a pulsing mode of operation, the implantation is performed without causing an electrostatic charge buildup on non-conductive targets since electrons, during the plasma recovery period, return to the target and neutralize any charge built up on the workpiece by the positive ion bombardment. This feature is obviously attractive for semiconductor applications, but also may be important in the modification of tribological behavior of materials other than metals, such as polymers, which are used more and more each year in the automotive industry. In what follows, we attempt to describe the key components of our PI11 system. In addition, some justification for our choice of components or techniques is given in the hope that a brief discussion of the principle features of the experiment will highlight problems which one will encounter when experimenting with PIII.
SIDEWALL-ANOOE
MAGNET RING
%REEN-ANODE
MAGNET RING
ENDWALL-ANODE MAGNET RING
CATM
SIOEWALLENOWALL
ANODE
Fig. 1. Ring-cusp hot filament chamber (after ref. [9]). 2. Plasma source Any known ion source can be used for PIII. However, to achieve controllable results, plasma equilibrium conditions are preferred. A source area with a high concentration of primary electrons should be arranged to allow diffusion of thermodynamically balanced plasma into a workable volume. Since the implantation dose per pulse depends on plasma density, a non-electrode discharge seems to be preferable. However, unless commercially available, these sources are more complex and require more experience in their design in comparison to a filament discharge. Another issue is the variety of different ion species which are generated when a molecular gas such as nitrogen is used. It is not clear how well one can control the generation of undesired species when a molecular gas is involved. This is an obvious disadvantage of the PI11 method since some processes may require a monoenergetic dopant species. Moreover, the lack of mass selecting electrostatic optics which are common in the beam type processes is especially attractive when estimating the costs of building a PI11 system for industrial applications. However, this, makes the PI11 approach barely viable for semiconductor manufacturing #l. Having in mind the complications with the theoretical interpretation of PI11 experiments when more than one sort of ion is involved, we believe that the source selection and design might be crucial for the experi-
#r A niche for implementation of a PI11 process was found and it is being intensely explored by researchers at University of California at Berkeley (see eg., [3,4,51).
ments. Theoretical calculations [6] indicate that a nitrogen plasma would consist predominantly of Nl ions with about 10% N+ in experiments run with an rf source (plasma produced in a borosilicate glass vacuum chamber by researchers from Lucas Heights Research Labs, Australia [7]). This does not provide the plasma environment which would allow for the “clean” experiments needed at this stage of understanding the implantation mechanisms in PIII. An investigation conducted at Hughes Research Laboratories [S] indicates that fairly good control over the generation of various ion species can be achieved in a hot filament source if attention is paid to the dissociation and ionization processes. The entrapment of primary electrons in the confined filament area and thermodynamically well balanced plasma diffusion into the work area provides a better chance for desired ion species selection. To investigate this option, provisions were made to initially utilize a ring-cusp hot filament discharge chamber (fig. 1) as the plasma source for our PI11 experiments. The high level of performance characteristic of the ring-cusp ion source is the result of the magnetic field distribution used to confine the discharge-chamber plasma (see ref. [9]). That is to say, the ring-cusp design is known to effectively confine a high level of primary electron population [9] leading to increased performance in ion production. Having a well-defined ion-production region and knowing that ion flow within the discharge chamber can be shaped by an electric field, one can attempt to extract the desired ion species from the source. This increases chances for selection of the desired ion species.
15
G. W. Malaczynski et al. / High-voltage implantation facility at GM 3. Main plasma
chamber
Losses due to ionized gas relaxation on the vacuum chamber walls decrease with volume increase of the chamber. Also, any unit which would be scaled up for mass production app!ications would be substantially bigger than the lab experimental vacuum vessel and would require different types of access (loading) ports, controllers, etc. Our design is a compromise between a small unit designed to conduct well controlled experiments and larger one suitable for mass production applications. The main chamber is a one cubic meter stainless steel cube with standard size ports for a vacuum system consisting of cryo and roughing pumps located below the cube base. The cryo system is used since our early experiments indicated the presence of contamination of the processed target with carbon which was traced to the diffusion pump oil (fig. 2). An access port is provided in the form of a hinged front door for the ease of loading. A number of standard conflat ports located on every side of the cube are included to accommodate plasma and implantation process diagnostic tools. There is a provision for attaching a plasma source discharge chamber at each side of the cube, except the bottom, in order to verify the effectiveness of a multi-source system and the coexistence of various plasma sources. In the future we also plan to run hybrid experiments with other processes such as vapor deposition, etc. To avoid excessive charge recombination at the chamber’s wall a full linecusp magnet plasma confinement system was used. The permanent magnets are located outside the vacuum chamber to avoid plasma contamination and to provide flexibility in experimenting with various geometrical arrangements of the permanent magnets. The magnets’ column spacing was optimized by accounting for the number of cusps [lo] and the increased distance due to
100
16 -
‘\
.:\,
‘\ 0
o “c, ‘., 0 100
100
a00
400
-\
\ \_.,__,* so0
400
Depth tnml Fig. 2. Auger electron spectroscopy analysis of an aluminum sample implanted with nitrogen ions at 35 kV. Shown are atomic percentages of aluminum, nitrogen, oxygen and carbon.
the permanent magnets’ location outside of the chamber. To achieve a high magnetic field strength, the magnet array consists of Magnequench 3R permanent magnets whose exceptional high field strength at the pole faces compensate for the magnets’ location outside the chamber. The system is designed to maintain an ion density on the order of 10’ ions/cm3 at a pressure of lop4 Torr. A target table, electrically insulated from the ground so as to withstand the high voltage pulses, is supported by structural insulators which are able to carry a substantial workpiece load. A 150 kV feedthrough provides the electrical connection to the target table. The target temperature, which might rise due to the heating effect of intense ion bombardment, is controlled by coolant circulation from a multi-kilowatt heat exchanger through this custom designed high-voltage feedthrough. Our early experiments indicated a tendency to develop a surface flashover along the surface of the HV feedthrough between the target table and the grounded vacuum chamber floor. Plasma in the feedthrough area was responsible for this flashover initiation. To prevent plasma penetration, the distance between the table and the side walls of the chamber was minimized. In addition, the feedthrough area lacks magnetic confinement to ensure high recombination of ions in this region which should further prevent plasma diffusion into this area.
4. High-voltage modulator system The high voltage pulse applied to the target must comply with certain requirements in order to achieve the desired implantation efficiency. As square a pulse as possible is desired to simplify any theoretical analysis and to achieve the maximum acceleration of the ions. However, such a requirement is not realistic and is really not necessary in the case of metallurgical applications. In general, the kinetic energy of an ion at impact with the target is defined by gas pressure (path between atomic collisions), the amount of acquired charge in the ionization process, and the electric field strength. In practice however, the above relationship becomes much more complicated due to the complexity of transient effects in the electron-ion separation under the stress of the high voltage pulse. A high-voltage negative pulse applied to a processed workpiece results in the creation of an electrically non-neutral region formed between the plasma and the target. The resulting electric field accelerates the plasma ions toward the workpiece surface. The charge separation process typically can be divided into two stages (see, e.g. ref. [ll]). First, on a time scale which is short in comparison with the ion motion, the electrons are repelled from the target resulting in an II. IMPLANTER SYSTEMS
16
G. W. Malaczynski et al. / High-voltage implantation facility at GM
ion matrix sheath [12,13]. Next, the ions are accelerated by the electric field toward the target and, at the same time, the ion matrix sheath moves away in the other direction. Consequently, the kinetic energy of the ions at impact, during at least the initial portion of the negative pulse rise time, is strongly degenerated. Ions which are too close to the target do not achieve a high velocity regardless of the initial acceleration. On the other hand, regardless of the sheath’s instantaneous position, maximum ionic speed at any electric field is a function of the mean free path between collisions, and, thus, a function of the gas pressure. Therefore, a square pulse rise is not an efficient way to secure the required kinetic energy in the implantation process and a multifactorial analysis is required for process optimization. This should include the ion matrix sheath dynamics, ion collisions, etc. On the other hand, assuming ion attraction toward the target without particle collision during the core of the applied high voltage pulse, implies that a flat pulse would be highly desirable to ensure a fixed kinetic energy at impact. This latter assumption obviously has to be verified by the calculation of ion sheath movement for constant gas pressure (free path between collisions). The width of the applied negative pulse is again defined by the sheath motion, since eventually the non-neutral region is terminated at the chamber wall or, in case of multiple targets, the sheaths would start to overlap each other and the plasma would be totally consumed. If the pulse width is properly selected, its falling tail is not important for the implantation process since most of the available ions are consumed Grnplanted)
during the core time of the applied negative voltage. However, if some ions are still available at the trailing pulse edge, their energy at impact could be lower. In the limit, detrimental sputtering may occur which would wipe out a portion of the already implanted material. At this experimental stage, the best approach seems to be to compromise, allowing for a reasonably short pulse fall time thereby avoiding the excessive costs one would face if a more demanding requirement were involved. The character of the electric load, comprised of the target immersed in the plasma, seems to be the key feature which determines the price of the high-voltage pulser. The maximum capacitance of the load represented by the target (at the very instant the pulse is applied) can be calculated knowing the target area under the assumption that the plasma is at ground potential. The Debye length defines the effective capacitor spacing and can be easily calculated if the plasma parameters are given. The minimum capacitance represents the load at the termination of the negative pulse. In the limit, its value is defined by the position of the target in the grounded vacuum vessel. Theoretical calculation of the resistive portion of the load would be.a very difficult task, but fortunately a fairly good estimation can be obtained from known experimental data. Numerous experiments performed in our lab and at the University of Wisconsin provided us with current traces taken by a Rogowski coil on the high voltage line leading to the target. This information combined with assumed total target area (for known plasma conditions the current density rather than the total current indicates the system load characteristic)
Fig. 3. Overall block diagram of the modulator system.
G. W Malaczynski et al. / High-voltage implantation facility at GM
led us to the load specification needed for the power supply design. The above considerations resulted in the following specification for the high voltage pulser adopted for the PIII experiment: fixed negative polarity and variable output pulse amplitude up to 150 kV, pulse repetition rate 10 to 175 Hz, pulse width, 10 to 20 us changed in 1 us steps, pulse rise time, better than 6 us, pulse fall time, better than 12 ps, ripple on the pulse flattop less than 5% load resistance, not less than 20 kfi (8 h operation at 175 Hz), output voltage monitored by a resistive voltage divider, output current monitored by a Pearson Coil. The m~ulator (fig. 31, built at Maxwell Laboratories, is made from a series of pulse forming networks and a thyratron, which is used to switch the forming network into a pulse transformer having a step-up ratio of 1: 10. Lowering the pulse repetition rate allows for higher pulse current (total input power remains constant) thus, if necessary the system can be used for large area target implantation, or in other words the system can be used with a load resistance lower than 20 k0 at frequencies lower than 175 Hz.
5. ~xpe~men~1
effort
There is a surprisingly poor theoretical fo~dation for the tribological benefits gained from the implantation of metals. Also, not too much is known concerning the effects of ion surface modifications on wear resistance. To build our own knowledge and systematic data base, a pin-on-disk friction tester and high frequency wear tester with reciprocal motion (Cameron Plint) are being used to verify the tribological changes
17
of implanted metals. Implant profiles, doses and changes in treated surfaces are analyzed with Scanning Auger Microscopy, Rutherford Backscattering Spectroscopy and Low Angle X-ray Diffraction techniques. Wear and friction tests are run without lubricant, with lubricant and under poor lubrication conditions. Poor lubrication is simulated by excessive Ioad which breaks the iubricating film or by applying sub-standard lubricants.
References [l] J.R. Conrad, J.L. Radtke, R.A. Dodd, F.J. Worzala and N.C. Tran, J. Appl. Phys. (1987) 4591. [2] AA. Elmoursi, G.W. Malaczynski and AH. Hamdi, Nucl. Instr. and Meth. B62 (1991) 293. [3] X.Y. Qian, N.W. Chermg, M.A. Lieberman, S.B. Felch, R. Brennan and M.I. Current, Appl. Phys. Lett. 59 (1991) 348. [41 CA. Pica, X.Y. Qian, E. Jones, M.A. Lieberman and N.W. Cheung, Mat. Res. Sot. Proc., Pittsburgh, PA, vol. 223 (1991) p. 115. 1.51N.W. Cheung, Mat. Res. Sot. Conference, Session K, #K7.4, San Francisco, CA, 1992. 161 I.J. Donnelly and E.K. Rose, Proc. 8th Topical Conf. on RF Power in Plasmas, Irvine, CA, 1989 (AIP, New York, 19891 p. 410. 171 G.A. Collins, R. Hutchings and J. Tendys, Mater. Sci. Eng. Al39 (1991) 171. ISI J.N. Matossian and J.R. Beattle, J. Propulsion and Power AIAA 5 (2) (1989) 188. [9] J.N. Matossian and J.R. Beattle, DGLR/~~/JSASS 20th Int. Electric Propulsion Co& 1988, GarmischPartenkirchen, Germany. 1101K.N. Leung, T.K. Samec and A. Lamm, Phys. Lett. 51A (8) 490. 1111M. Shamim, J.T. Scheuer and J.R. Conrad, J. Appl. Phys. 69 (5) (1991) 2904. 1121 G. Andrews and R.H. Varey, Phys. Fluids 14 (1971) 339. 1131J.R. Conrad, J. Appl. Phys. 62 (1987) 777.
H. IMPLANTER SYSTEMS
Nuclear Instruments and Methods in Physics Research B74 (1993) 13-17 North-Holland
High-voltage
implantation
Beam Interactions with Materials&Atoms
facility at GM Research
Gerard W. Malaczynski, Alaa A. Ehnoursi, Aboud H. Hamdi and Xiaohong Qiu Electrical
and Electronics
Engineering Department,
General Motors Research,
Warren, MI 48090-9055,
USA
This paper describes the effort undertaken at General Motors Research to build a plasma immersion ion implantation (PIII) system in order to study the potential of pulsed implantation as a means of modifying the tribological properties of automotive parts and to investigate the feasibility of producing surface modified components. The system has been designed for high dose ion implantation of irregularly shaped objects. It consists of a stainless steel vacuum chamber evacuated by a cryopump system, standard plasma diagnostic tools, a ring-cusp hot filament plasma source and a custom designed 150 kV, lo-20 t.~spulser. Details of the system are discussed together with justification for our choice of the specific components and the adopted techniques.
1. Introduction
Plasma immersion ion implantation (PI111 is a relatively new, non-line-of-sight technique for surface modification of materials. The method was originally developed by John Conrad at the University of Wisconsin at Madison [l] with the intention of surface modification for metallurgical applications. In the PI11 process, the workpiece is placed inside a vacuum chamber in which a plasma is generated. A series of negative high-voltage pulses are applied to this workpiece which results in the motion of ions toward and electrons away from the material to be treated. In effect, high velocity ion strike the biased target and penetrate its surface to a depth defined by the ion’s kinetic energy at impact and collisions with the lattice of the solid material. Charge separation caused by the strong applied electric field leads to a strong disturbance of the plasma’s electric neutrality. Eventually, the ion matrix sheath created by the initial movement of electrons away from the biased workpiece is consumed and the negative bias has to be terminated to recover the plasma. Therefore, to achieve continuous operation of the plasma immersion implanter, a high-voltage short duration pulser is required. Efficient plasma recovery, the desired ion density, ion species selection, penetration depth and implant density control for irregularly shaped objects, together with the necessity for extensive system control and diagnostic needed at this developmental stage (which must precede any practical application), makes the task of building a PI11 system a serious scientific and engineering challenge. This paper describes the effort undertaken at General Motors Research to build a PI11 system in order to 0168-583X/93/$06.00
0 1993 -
study this method’s potential in modifying the tribological properties of automotive parts and to investigate the feasibility of producing surface modified components. To formulate the requirements for a reasonably effective experimental system which would allow us to investigate various implants in a variety of materials, a series of initial experiments were performed with an ad hoc system composed of standard vacuum and high voltage components. The first implants were performed with a high voltage pulser operating at 35 kV [2]. The construction of this low voltage implant system rendered the necessary experience to formulate our list of desirable features, limitations, and the necessary conditions for efficient plasma implantation. In this initial learning process, our close cooperation and consultations with numerous members of the University of Wisconsin team was essential. This resulted in a successful “proof of principle” demonstration, which culminated in a nitrogen implant of a metallic foil, verified with Auger electron spectroscopy analysis [2]. In general, PI11 is a multidisciplinary project requiring knowledge of high-voltage engineering, vacuum technology, plasma physics, material science, and tribology. Since the intention of this paper is to describe our PI11 implanter, the discussion of material science and tribology will be limited and provided only when it is needed to justify a particular choice in the construction of the system. With so many complexities in the PI11 approach compared with the well-understood process occurring in line-of-sight implantation techniques, it is worthwhile to stress the uniqueness of the plasma process which goes beyond being able to achieve a uniform treatment of contoured targets. Specifically, by employ-
Elsevier Science Publishers B.V. All rights reserved
II. IMI’I.,ANTER SYSTEMS
14
G. W. Malaczynski et al. / High-voltage implantation facility at GM
ing a pulsing mode of operation, the implantation is performed without causing an electrostatic charge buildup on non-conductive targets since electrons, during the plasma recovery period, return to the target and neutralize any charge built up on the workpiece by the positive ion bombardment. This feature is obviously attractive for semiconductor applications, but also may be important in the modification of tribological behavior of materials other than metals, such as polymers, which are used more and more each year in the automotive industry. In what follows, we attempt to describe the key components of our PI11 system. In addition, some justification for our choice of components or techniques is given in the hope that a brief discussion of the principle features of the experiment will highlight problems which one will encounter when experimenting with PIII.
SIDEWALL-ANOOE
MAGNET RING
%REEN-ANODE
MAGNET RING
ENDWALL-ANODE MAGNET RING
CATM
SIOEWALLENOWALL
ANODE
Fig. 1. Ring-cusp hot filament chamber (after ref. [9]). 2. Plasma source Any known ion source can be used for PIII. However, to achieve controllable results, plasma equilibrium conditions are preferred. A source area with a high concentration of primary electrons should be arranged to allow diffusion of thermodynamically balanced plasma into a workable volume. Since the implantation dose per pulse depends on plasma density, a non-electrode discharge seems to be preferable. However, unless commercially available, these sources are more complex and require more experience in their design in comparison to a filament discharge. Another issue is the variety of different ion species which are generated when a molecular gas such as nitrogen is used. It is not clear how well one can control the generation of undesired species when a molecular gas is involved. This is an obvious disadvantage of the PI11 method since some processes may require a monoenergetic dopant species. Moreover, the lack of mass selecting electrostatic optics which are common in the beam type processes is especially attractive when estimating the costs of building a PI11 system for industrial applications. However, this, makes the PI11 approach barely viable for semiconductor manufacturing #l. Having in mind the complications with the theoretical interpretation of PI11 experiments when more than one sort of ion is involved, we believe that the source selection and design might be crucial for the experi-
#r A niche for implementation of a PI11 process was found and it is being intensely explored by researchers at University of California at Berkeley (see eg., [3,4,51).
ments. Theoretical calculations [6] indicate that a nitrogen plasma would consist predominantly of Nl ions with about 10% N+ in experiments run with an rf source (plasma produced in a borosilicate glass vacuum chamber by researchers from Lucas Heights Research Labs, Australia [7]). This does not provide the plasma environment which would allow for the “clean” experiments needed at this stage of understanding the implantation mechanisms in PIII. An investigation conducted at Hughes Research Laboratories [S] indicates that fairly good control over the generation of various ion species can be achieved in a hot filament source if attention is paid to the dissociation and ionization processes. The entrapment of primary electrons in the confined filament area and thermodynamically well balanced plasma diffusion into the work area provides a better chance for desired ion species selection. To investigate this option, provisions were made to initially utilize a ring-cusp hot filament discharge chamber (fig. 1) as the plasma source for our PI11 experiments. The high level of performance characteristic of the ring-cusp ion source is the result of the magnetic field distribution used to confine the discharge-chamber plasma (see ref. [9]). That is to say, the ring-cusp design is known to effectively confine a high level of primary electron population [9] leading to increased performance in ion production. Having a well-defined ion-production region and knowing that ion flow within the discharge chamber can be shaped by an electric field, one can attempt to extract the desired ion species from the source. This increases chances for selection of the desired ion species.
15
G. W. Malaczynski et al. / High-voltage implantation facility at GM 3. Main plasma
chamber
Losses due to ionized gas relaxation on the vacuum chamber walls decrease with volume increase of the chamber. Also, any unit which would be scaled up for mass production app!ications would be substantially bigger than the lab experimental vacuum vessel and would require different types of access (loading) ports, controllers, etc. Our design is a compromise between a small unit designed to conduct well controlled experiments and larger one suitable for mass production applications. The main chamber is a one cubic meter stainless steel cube with standard size ports for a vacuum system consisting of cryo and roughing pumps located below the cube base. The cryo system is used since our early experiments indicated the presence of contamination of the processed target with carbon which was traced to the diffusion pump oil (fig. 2). An access port is provided in the form of a hinged front door for the ease of loading. A number of standard conflat ports located on every side of the cube are included to accommodate plasma and implantation process diagnostic tools. There is a provision for attaching a plasma source discharge chamber at each side of the cube, except the bottom, in order to verify the effectiveness of a multi-source system and the coexistence of various plasma sources. In the future we also plan to run hybrid experiments with other processes such as vapor deposition, etc. To avoid excessive charge recombination at the chamber’s wall a full linecusp magnet plasma confinement system was used. The permanent magnets are located outside the vacuum chamber to avoid plasma contamination and to provide flexibility in experimenting with various geometrical arrangements of the permanent magnets. The magnets’ column spacing was optimized by accounting for the number of cusps [lo] and the increased distance due to
100
16 -
‘\
.:\,
‘\ 0
o “c, ‘., 0 100
100
a00
400
-\
\ \_.,__,* so0
400
Depth tnml Fig. 2. Auger electron spectroscopy analysis of an aluminum sample implanted with nitrogen ions at 35 kV. Shown are atomic percentages of aluminum, nitrogen, oxygen and carbon.
the permanent magnets’ location outside of the chamber. To achieve a high magnetic field strength, the magnet array consists of Magnequench 3R permanent magnets whose exceptional high field strength at the pole faces compensate for the magnets’ location outside the chamber. The system is designed to maintain an ion density on the order of 10’ ions/cm3 at a pressure of lop4 Torr. A target table, electrically insulated from the ground so as to withstand the high voltage pulses, is supported by structural insulators which are able to carry a substantial workpiece load. A 150 kV feedthrough provides the electrical connection to the target table. The target temperature, which might rise due to the heating effect of intense ion bombardment, is controlled by coolant circulation from a multi-kilowatt heat exchanger through this custom designed high-voltage feedthrough. Our early experiments indicated a tendency to develop a surface flashover along the surface of the HV feedthrough between the target table and the grounded vacuum chamber floor. Plasma in the feedthrough area was responsible for this flashover initiation. To prevent plasma penetration, the distance between the table and the side walls of the chamber was minimized. In addition, the feedthrough area lacks magnetic confinement to ensure high recombination of ions in this region which should further prevent plasma diffusion into this area.
4. High-voltage modulator system The high voltage pulse applied to the target must comply with certain requirements in order to achieve the desired implantation efficiency. As square a pulse as possible is desired to simplify any theoretical analysis and to achieve the maximum acceleration of the ions. However, such a requirement is not realistic and is really not necessary in the case of metallurgical applications. In general, the kinetic energy of an ion at impact with the target is defined by gas pressure (path between atomic collisions), the amount of acquired charge in the ionization process, and the electric field strength. In practice however, the above relationship becomes much more complicated due to the complexity of transient effects in the electron-ion separation under the stress of the high voltage pulse. A high-voltage negative pulse applied to a processed workpiece results in the creation of an electrically non-neutral region formed between the plasma and the target. The resulting electric field accelerates the plasma ions toward the workpiece surface. The charge separation process typically can be divided into two stages (see, e.g. ref. [ll]). First, on a time scale which is short in comparison with the ion motion, the electrons are repelled from the target resulting in an II. IMPLANTER SYSTEMS
16
G. W. Malaczynski et al. / High-voltage implantation facility at GM
ion matrix sheath [12,13]. Next, the ions are accelerated by the electric field toward the target and, at the same time, the ion matrix sheath moves away in the other direction. Consequently, the kinetic energy of the ions at impact, during at least the initial portion of the negative pulse rise time, is strongly degenerated. Ions which are too close to the target do not achieve a high velocity regardless of the initial acceleration. On the other hand, regardless of the sheath’s instantaneous position, maximum ionic speed at any electric field is a function of the mean free path between collisions, and, thus, a function of the gas pressure. Therefore, a square pulse rise is not an efficient way to secure the required kinetic energy in the implantation process and a multifactorial analysis is required for process optimization. This should include the ion matrix sheath dynamics, ion collisions, etc. On the other hand, assuming ion attraction toward the target without particle collision during the core of the applied high voltage pulse, implies that a flat pulse would be highly desirable to ensure a fixed kinetic energy at impact. This latter assumption obviously has to be verified by the calculation of ion sheath movement for constant gas pressure (free path between collisions). The width of the applied negative pulse is again defined by the sheath motion, since eventually the non-neutral region is terminated at the chamber wall or, in case of multiple targets, the sheaths would start to overlap each other and the plasma would be totally consumed. If the pulse width is properly selected, its falling tail is not important for the implantation process since most of the available ions are consumed Grnplanted)
during the core time of the applied negative voltage. However, if some ions are still available at the trailing pulse edge, their energy at impact could be lower. In the limit, detrimental sputtering may occur which would wipe out a portion of the already implanted material. At this experimental stage, the best approach seems to be to compromise, allowing for a reasonably short pulse fall time thereby avoiding the excessive costs one would face if a more demanding requirement were involved. The character of the electric load, comprised of the target immersed in the plasma, seems to be the key feature which determines the price of the high-voltage pulser. The maximum capacitance of the load represented by the target (at the very instant the pulse is applied) can be calculated knowing the target area under the assumption that the plasma is at ground potential. The Debye length defines the effective capacitor spacing and can be easily calculated if the plasma parameters are given. The minimum capacitance represents the load at the termination of the negative pulse. In the limit, its value is defined by the position of the target in the grounded vacuum vessel. Theoretical calculation of the resistive portion of the load would be.a very difficult task, but fortunately a fairly good estimation can be obtained from known experimental data. Numerous experiments performed in our lab and at the University of Wisconsin provided us with current traces taken by a Rogowski coil on the high voltage line leading to the target. This information combined with assumed total target area (for known plasma conditions the current density rather than the total current indicates the system load characteristic)
Fig. 3. Overall block diagram of the modulator system.
G. W Malaczynski et al. / High-voltage implantation facility at GM
led us to the load specification needed for the power supply design. The above considerations resulted in the following specification for the high voltage pulser adopted for the PIII experiment: fixed negative polarity and variable output pulse amplitude up to 150 kV, pulse repetition rate 10 to 175 Hz, pulse width, 10 to 20 us changed in 1 us steps, pulse rise time, better than 6 us, pulse fall time, better than 12 ps, ripple on the pulse flattop less than 5% load resistance, not less than 20 kfi (8 h operation at 175 Hz), output voltage monitored by a resistive voltage divider, output current monitored by a Pearson Coil. The m~ulator (fig. 31, built at Maxwell Laboratories, is made from a series of pulse forming networks and a thyratron, which is used to switch the forming network into a pulse transformer having a step-up ratio of 1: 10. Lowering the pulse repetition rate allows for higher pulse current (total input power remains constant) thus, if necessary the system can be used for large area target implantation, or in other words the system can be used with a load resistance lower than 20 k0 at frequencies lower than 175 Hz.
5. ~xpe~men~1
effort
There is a surprisingly poor theoretical fo~dation for the tribological benefits gained from the implantation of metals. Also, not too much is known concerning the effects of ion surface modifications on wear resistance. To build our own knowledge and systematic data base, a pin-on-disk friction tester and high frequency wear tester with reciprocal motion (Cameron Plint) are being used to verify the tribological changes
17
of implanted metals. Implant profiles, doses and changes in treated surfaces are analyzed with Scanning Auger Microscopy, Rutherford Backscattering Spectroscopy and Low Angle X-ray Diffraction techniques. Wear and friction tests are run without lubricant, with lubricant and under poor lubrication conditions. Poor lubrication is simulated by excessive Ioad which breaks the iubricating film or by applying sub-standard lubricants.
References [l] J.R. Conrad, J.L. Radtke, R.A. Dodd, F.J. Worzala and N.C. Tran, J. Appl. Phys. (1987) 4591. [2] AA. Elmoursi, G.W. Malaczynski and AH. Hamdi, Nucl. Instr. and Meth. B62 (1991) 293. [3] X.Y. Qian, N.W. Chermg, M.A. Lieberman, S.B. Felch, R. Brennan and M.I. Current, Appl. Phys. Lett. 59 (1991) 348. [41 CA. Pica, X.Y. Qian, E. Jones, M.A. Lieberman and N.W. Cheung, Mat. Res. Sot. Proc., Pittsburgh, PA, vol. 223 (1991) p. 115. 1.51N.W. Cheung, Mat. Res. Sot. Conference, Session K, #K7.4, San Francisco, CA, 1992. 161 I.J. Donnelly and E.K. Rose, Proc. 8th Topical Conf. on RF Power in Plasmas, Irvine, CA, 1989 (AIP, New York, 19891 p. 410. 171 G.A. Collins, R. Hutchings and J. Tendys, Mater. Sci. Eng. Al39 (1991) 171. ISI J.N. Matossian and J.R. Beattle, J. Propulsion and Power AIAA 5 (2) (1989) 188. [9] J.N. Matossian and J.R. Beattle, DGLR/~~/JSASS 20th Int. Electric Propulsion Co& 1988, GarmischPartenkirchen, Germany. 1101K.N. Leung, T.K. Samec and A. Lamm, Phys. Lett. 51A (8) 490. 1111M. Shamim, J.T. Scheuer and J.R. Conrad, J. Appl. Phys. 69 (5) (1991) 2904. 1121 G. Andrews and R.H. Varey, Phys. Fluids 14 (1971) 339. 1131J.R. Conrad, J. Appl. Phys. 62 (1987) 777.
H. IMPLANTER SYSTEMS