- Email: [email protected]
Characterization of a high-density plasma immersion ion implanter with scaleable ECR large-area plasma source
Surface & Coatings Technology 196 (2005) 172 – 179 www.elsevier.com/locate/surfcoat
Characterization of a high-density plasma immersion ion implanter with scaleable ECR large-area plasma source Yuri Glukhoya,1, Mahmud Rahmanb,*, Gotze Popova,1, Alexander Usenkoc,2, Hans J. Walitzkid,3 b
a American Advanced Ion Beam, Inc., 440 Arguello Blvd, #1, San Francisco, CA 94118, USA Department of Electrical Engineering, Electron Devices Laboratory, Santa Clara University, 500 El Camino Real, Santa Clara, CA 95053 0583, USA c Silicon Wafer Technologies, Inc., 240 King Blvd., Newark, NJ 07102, USA d Isonics Corporation, 12514 NE 95th Street C-110, Vancouver, WA 98682, USA
Available online 2 November 2004
Abstract A key issue for future high-performance CMOS techniques is the fabrication of 12 in. or larger silicon-on-insulator (SOI) wafers with thickness of the top silicon layer under 25 nm. The electron cyclotron resonance (ECR) plasma ion immersion implanters (ECR-PIII) are viable candidates to realize the above. We designed an ECR-PIII in which we integrated the new concept of a large-area plasma source. This new plasma source is an array of elementary ECR plasma sources created by an assembly of m linear microwave sources. Each linear microwave source has n radiating elements (nNm) and includes a system of permanent magnets that create magnetic induction for ECR to occur in the processing chamber. An array of 90 elementary ECR plasma sources generate highly homogeneous dense plasma with dimensions permitting processing of 12-in. wafers. Since hydrogen implantation requiring high dose often results in blistering of the wafer surface, a new Smart-Cutk-like technology has been suggested (US Patent 6,352,909) where the one-step hydrogen implantation is replaced by a two-step process. The designed ECR-PIII can offer substantial advantage in providing the high yield of protons used for the micro-bubble formation in the two-step process. D 2004 Elsevier B.V. All rights reserved. Keywords: Electron cyclotron resonance (ECR) plasmas; Plasma immersion ion implantation (PIII); Microwave; Hydrogen; Silicon; Silicon-on-insulator (SOI)
1. Introduction We have designed an electron cyclotron resonance (ECR) plasma ion immersion implanter (ECR-PIII) particularly suitable for fabrication of 12-in. bonded thin-layer silicon-on-insulator (SOI) wafers. As the semiconductor industry moves from 0.13 Am to b50 nm and beyond line width, thin-layer SOI is expected to be the technology of choice for high-performance CMOS systems. The ECR-
* Corresponding author. Tel.: +1 408 554 4175; fax: +1 408 554 5474. E-mail addresses: [email protected] (Y. Glukhoy)8 [email protected] (M. Rahman)8 [email protected] (G. Popov)8 http://www.si-sandwich.com, [email protected] (A. Usenko), [email protected] (H. Walitzki). 1 Tel./fax: +1 415 751 7666. 2 Tel.: +1 973 297 1410; fax: +1 973 297 1125. 3 Tel.: +1 360 885 9310x205; fax: +1 360 885 9311. 0257-8972/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2004.08.177
PIIIs are anticipated to replace the standard ion implanters for fabrication of 12 in. and larger thin-layer SOI wafers. The most advanced methods for manufacturing bonded SOI wafers are the layer transfer methods. Smart-Cutk [1] was the first layer transfer method described in the literature. It gives high-quality top layer in a final SOI wafer. However, the Smart-Cutk is not well suited for making SOI wafers with a top layer thinner than 100 nm. Recently, Usenko et al. [2] suggested an improved layer transfer method which allows manufacturing of SOI wafers with thinner top layers required for next-generation chips. The layer transfer method consists of the following steps: ! forming a fragile plane layer inside of a Si wafer called bdonor waferQ ! forming a silicon dioxide layer on top of another Si wafer, called bhandle waferQ
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
! bonding the top side of the donor wafer to the top side of the handle wafer, bsandwichingQ the silicon dioxide layer in between ! separation of the donor wafer along the fragile plane. In this way, a bonded SOI wafer consists of a handle wafer with buried oxide layer and the top layer that has been separated from the donor wafer. The separation techniques depend on the how the fragile plane layer is formed. In the Smart-Cutk method, the fragile layer is formed by hydrogen ion implantation at high ion dose such as 51016/cm2 and at an ion dose rate less than 1015/cm2 s. The depth to which the hydrogen ions penetrate into silicon determines the inherent thickness of the transferred layer. For typical ion implantation, energies between 30 and 100 keV, the corresponding depth is 0.3–1 Am. Implantation at low energy raises several problems. When a high-energy ion reaches the target Si, it dissipates its energy mostly because of interactions with the electronic subsystem of the Si crystal. The energy of the penetrating ion gradually decreases as its energy is transferred mostly to the target’s electrons. When the ion energy drops to i10 keV, the ion produces mostly atomic displacements. Ion implantation at low energy, e.g., 10 keV and less, produces effective atomic displacements also at the surface of the target resulting in the formation of a damaged layer that is not buried under the target’s surface. This damaged surface has an increased roughness and cannot be bounded to another (handle) wafer [3]. The process for ultra-thin SOI fabrication proposed by Usenko et al. consists of the following steps (also see Usenko, US Patent 6,352,909 B1 [4]): ! forming a buried trap layer for hydrogen in the donor wafer substrate ! nucleation of hydrogen platelets in the above buried trap layer by hydrogenation at a temperature less than 250 8C ! growing hydrogen platelets in the buried trap layer by hydrogenation at temperatures in the range of 250–400 8C, thus forming a buried fragile layer in the donor wafer ! forming a silicon dioxide layer on the top surface of the handle wafer ! activating the top surfaces of donor and handle wafers ! forming a temporary wafer assembly by making initial contact of the activated surfaces of the donor and handle wafers ! separating the above wafer assembly into two wafers by cleaving the assembly at the buried fragile layer. The buried trap layer for hydrogen is created by ion implantation. The ions used for such implantation could be either those that do not have electrical activity in the Si crystal, e.g., oxygen ions, or the same ions as the substrate, i.e., Si, or noble gas ions that do not chemically interact with silicon. Hydrogen is delivered to the buried trap layer from a
173
hydrogen plasma, and this defect layer serves as an infinite capacity trap for hydrogen. The ECR-PIII can be used for manufacturing SOI wafers by the Smart-Cutk method as well as thin-layer SOI wafers by the Usenko’s process. For the Usenko process, ECR-PIII can be used for both noble gas ion implantation and hydrogenation at ion energies of approximately 20 keV and at the temperatures mentioned above. The ECR-PIII consists of the following systems: ! processing chamber, which includes an assembly of m linear microwave sources used to create a highly homogeneous dense plasma ! high-vacuum system ! gas delivery system ! wafer transfer and positioning system ! microwave generators and HV power supplies for the generators ! pulsed HV generator for application of short negative HV pulses at the implantation target ! a computer controller and PLC that ensure the proper function of all subsystems.
2. The scalable plasma source Microwaves can generate and sustain plasma at low pressures. In the absence of a magnetic field, the plasma density, n, is limited by the frequency of the microwave, x, according to the following relationship: n V nc ¼ x2 eo m=e2 ; where n c is the critical plasma density, e o is the dielectric constant in vacuum, and m and e are the mass and the charge of an electron, respectively. When a steady magnetic field B is now applied to the microwave plasma, an electron cyclotron resonance (ECR) occurs. This is a resonance between the applied microwave frequency x and the cyclotron frequency x c=eB/m. At ECR conditions, the electrons rotate along the magnetic field lines in phase with the right-hand circularly polarized microwaves and obtain sufficient energy to ionize the ambient gas. The injection of microwave along the magnetic field allows it to propagate in a dense plasma and create plasma densities Nn c. A standard ECR plasma source includes a cylindrical chamber. The microwave power is applied to the chamber via a low-loss dielectric window positioned perpendicular to the chamber axis. Magnetic field coils generate nonuniform axial magnetic field B(z) within the chamber, and its magnetic induction can achieve the ECR condition in one or more points on the axis. In such a configuration, it is almost impossible to obtain a homogeneous plasma with dimensions that permit processing of 300-mm wafers. Our approach to this problem is based on the new concept of
174
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
using an array of elementary ECR plasma sources to realize a large area homogeneous plasma. 2.1. Geometry of the array of elementary plasma sources An array of elementary ECR plasma sources is the key element of the ECR-PIII proposed in our work. This is a fully scalable plasma source that is constructed using a rectangular array of n m elementary ECR plasma sources, where nNm. This array of elementary ECR plasma sources is created by an assembly of m linear microwave sources. Each linear microwave source in our design is powered by its own microwave generator at 2.45 GHz and includes n equally spaced elementary microwave sources. The details of the design are discussed further below. Fig. 1 shows an ECR plasma source array for 200-mm wafers consisting of 36 (4 on the short side and 9 on the long side of the rectangular array) elementary ECR plasma sources The distance between two adjacent elementary plasma sources on the short side of the rectangular array is D=3 in., whereas that on the long side is d=k/4 (=3.04 cm). The distances D and d depend on the frequency of the microwave used and the power of the microwave generators. The microwave power distribution and the position of the wafer are so chosen as to ensure homogeneity of the plasma better than 98% over the entire wafer surface. In order to ensure the homogeneity of the plasma in case of 300-mm wafers, we have also designed an ECR plasma source array consisting of 90 elementary ECR sources. 2.2. Linear microwave source Each linear microwave source consists of a microwave coaxial line with uniformly spaced slots cut at 1208 in the outer electrode of the line. A microwave generator supplies power to the slotted microwave line whence the slots emit the microwave power. The distance between two adjacent slots is ~k/4=3.04 cm for a frequency, f=2.45 GHz. Two
Fig. 1. An array of 36 elementary ECR plasma sources produced by 4 linear microwave sources. Each linear microwave source has 9 radiating elements.
permanent ring magnets are installed around each slot in the NS position. The inner diameter of the ring magnet is equal to the outer diameter of the outer line electrode. The magnetic induction, B r corresponding to the ECR condition (B r=0.0875 T at f=2.45 GHz) is achieved in the processing chamber at about 3 cm away from the axis of the coaxial line. A ceramic tube with an outer radius of 19 mm covers the slotted coaxial line and the permanent magnets. The ceramic tube also serves as a low-loss window for applying microwave in the low-pressure processing chamber. Fig. 2a shows the longitudinal cross section of a section of a microwave source with only four slots. The core radius, r=4.25 mm, and the radius of the outer line electrode, R=9.85 mm. The distance between two adjacent slots is 3.04 cm, the slot thickness is 3.5 mm, and the gap between the two permanent magnets installed around each slot is 6.5 mm. Fig. 2b shows a section in the center of the slot perpendicular to the axis of the coaxial line. 2.3. Simulation of a linear microwave source In order to ascertain the feasibility of a scalable array of elementary ECR plasma sources, a computer simulation of a linear microwave source with dimensions given in Fig. 2a and b was performed under the following assumptions: ! A linear microwave source consists of n radiating elements (slots) equally spaced at a distance d apart. ! Each radiating element has the same amplitude of E field. There is a phase shift d between the E fields of two adjacent radiating elements. ! An efficient linear microwave source transmits most of the microwave power to the plasma. ! A system of two ring magnets installed on the two sides of each radiating element creates a magnetic induction, B r=0.0875 T in the processing chamber at about 3 cm away from the axis of the slotted coaxial line. ! The plasma is considered as an absorbing layer with a thickness of 3 mm and an outer radius of 30 mm (see Fig. 2b). The superficial resistance of the absorbing layer varies between 10 and 50 V/cm2. Although these assumptions are inadequate to take into account the complicated interaction between the plasma and the microwave, nonetheless, they are good enough to demonstrate the preliminary feasibility of such a linear microwave source at f=2.45 GHz that can generate ECR plasmas with density up to 51012/cm3. Fig. 2c shows the resonance characteristics of the linear microwave source with dimensions as in Fig. 2a and b. The resonance occurs at f r=2.42 GHz. In the presence of a real plasma, the resonance is expected to be shifted slightly to higher frequency, f=2.45 GHz, which is the most used microwave frequency. Fig. 3a shows the distribution of the E field vectors in a section perpendicular to the line axis and passes through the
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
175
Fig. 2. (a) The longitudinal sectional view of a linear microwave source having four radiating elements (slots). The permanent ring magnets positioned on two sides of each slot are also shown. 1–core of the coaxial line, radius=4.25 mm; 2–slotted outer line electrode, radius=9.85 mm; 3–slot, cut at 1208 on the outer line electrode; 4–permanent ring magnet; 5–ceramic tube, outer radius=19 mm. (b) Cross section of a linear microwave source shown in (a) through the center of one slot. The plasma is represented as an absorbing layer. (c) S 11 as a function of the microwave frequency calculated for a linear microwave source shown in (a).
slot center. It is clearly seen how the E power passes through the ceramic tube and is absorbed in the absorbing layer. Fig. 3b shows the superposition of the magnetic fields of permanent magnets around three slots at positions 3a, 3b and 3c. At distances z3 cm, the magnetic induction lines are parallel to the coaxial line axis. Therefore, the elementary ECR plasma sources created by each linear microwave source will have nearly the same plasma density. Fig. 3c shows the radiated microwave power in a plane that passes through the coaxial line axis and the centers of the three slots at positions 3a, 3b and 3c. It can be seen here that the radiating elements (slots) emit microwave power.
Fig. 3d shows the radiated microwave power in the same plane as in Fig. 3a. The radiated microwave power in this figure conform to the distribution of the E field vectors shown in Fig. 3a. The above results of simulation have demonstrated clearly the feasibility of the design of a linear microwave source capable to create n elementary ECR plasma sources. Therefore, an assembly of m linear microwave sources would create an array of n m elementary ECR plasma sources. The minimum distance between two adjacent linear microwave sources is determined by the physical dimension of the slotted coaxial line and the safe vacuum seals of the ceramic tubes in the processing
176
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
Fig. 3. (a) Distribution of the E field vectors in the cross section of the slot in a linear microwave source with dimensions shown in Fig. 2a and b. Labels 1, 2, 3, 4, and 5 refer to same positions in all four figures (a–d). 1–inner electrode (core) of the slotted transmission line; 2–outer electrode of the slotted transmission line; 3–slot; 4–ceramic tube; 5–absorbing layer. (b) Magnetic induction lines as a result of superposition of magnetic fields created by the permanent magnets around three slots at positions 3a, 3b, and 3c. (c) Distribution of radiated microwave power in a plane that passes through the axis of the slotted coaxial line and the centers of slots 3a, 3b, and 3c. (d) Distribution of radiated microwave power from a single slot in a plane perpendicular to the axis of the linear microwave source.
chamber. In our design ( f=2.45 GHz), this distance is about 2.5 in. The superposition of n m elementary ECR plasma sources will create homogeneous plasma at some distance from the surface where the ECR condition occurs. Our estimate for this distance for 156=90 elementary ECR plasma sources with D=3 in. and d=3.04 cm for processing of 300-mm wafers is 2 in. away from the ECR surface. The 300-mm wafer will be processed at this distance in order to achieve plasma homogeneity better than 98%. Certainly, at this distance, the plasma density, n p will be lower, say, n p~0.6n r, where n r is the plasma density at the ECR surface. In our proposed design, we can introduce up to 30 kW microwave power for processing of 300-mm
wafers. This power is 15 times higher than the microwave power used in the standard cylindrical ECR plasma sources at f=2.45 GHz. Therefore, the n p we can achieve in our design is indeed much higher than the plasma density obtained in a standard ECR plasma source. 2.4. Design of a linear microwave source and an assembly of linear sources Fig. 4 shows a linear microwave source with 15 radiating elements (slots). The ceramic tube (1) is opened partially to show the design. The slots (3) of this microwave source are cut at 608 in the outer electrode (2). The thickness of the slot is 6.5 mm, and the gap between the two permanent magnets around the slot is also 6.5 mm. The spacers (5) and (6) fix
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
177
Fig. 4. A linear microwave source with 15 radiating elements (slots). 1–ceramic tube 2–outer electrode of the slotted transmission line; 3–permanent ring magnet; 4–slot; 5–half-ring spacer; 6–ring spacer.
the positions of the permanent magnets. The distance between two adjacent slots is ~k/4. It is imperative to use air-cooling to ensure working temperature below the Curie point of the permanent magnets. Fig. 5a and b shows the schematics of the montage of seven linear microwave sources in the processing chamber. In order to simplify the drawing, each linear microwave source is drawn with only seven radiating elements. The ceramic tubes that also serve as windows for application of
the microwave power in the chamber are mounted in the process chamber by means of vacuum-tight compression flanges. A gas delivery system ensures a homogeneous distribution of the process gas in the chamber. The load-lock system transfers and positions the wafer on the electrostatic chuck. The wafer is surrounded by a highly homogeneous plasma. Fig. 6 shows an assembly of six linear microwave sources. Each linear microwave source (5) has 15 radiating
Fig. 5. Montage of linear microwave sources in the processing chamber. (a) longitudinal sectional view (b) cross-sectional view.
178
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
The EEDF depends on the parameter E/N [V cm2], where N [cm3] is the density on the neutral atoms or molecules in the discharge. In ECR discharges E/NN1014 V cm2. At these conditions, the EEDF is non-Maxwellian, and the anisotropic part becomes important (availability of a magnetic field), and the calculation of the EEDF using the two-term expansion of the Boltzmann equation seems to be no more valid. The mean electron energy in the ECR discharge in the pressure interval of 0.2–10 mTorr is between 8 and 15 eV according to some measurements in the Ruhr-University, Bochum, Germany [6]. 3.1. Electron-impact ionization cross section for hydrogen atom Fig. 6. An assembly of six linear microwave sources. Each linear source has 15 radiating elements. 1–5 kW, 2.45 GHz water-cooled magnetron; 2– launcher; 3–three-stub tuners. Although three-stub tuners are used in this prototype, in the commercial product, computer-controlled step-motor driven tuners will be used; 4–transition from waveguide to coaxial line; 5– linear microwave source; 6–tuning piston. At this position, a fan for cooling the slotted transmission line and the permanent magnet is mounted.
elements (also see Fig. 4) and is powered by a 5-kW watercooled magnetron (1) via a three-stub tuner (3) and a transition from waveguide to coaxial line (4). This assembly has a total of 90 radiating elements. The distance between the axes of two adjacent linear microwave sources is 3 in. A wafer will be immersed in the chamber at distance ~3.2 in. from the plane formed by the axes of the linear microwave sources. Fig. 7 shows a perspective view of an open ECR-PIII. The electromagnetic, electronic, and pneumatic elements and the two roughing pumps are not shown in order to present the main assemblies clearly. The process chamber is pumped by two turbomolecular pumps (7). Two computercontrolled gate valves function in accordance with the loadlock system (3), the protection grid moving system (4), and the electrostatic chuck (8).
The electron-impact ionization cross section for hydrogen atom has its maximum, r max=0.61016 cm2 at an energy of 55 eV [7]. The electron-impact ionization cross section for hydrogen molecule ion (H2) varies similarly with energy with a maximum, r max=0.171016 cm2 at an energy of 100 eV [8]. The electron-impact ionization cross section for hydrogen atom is about four times larger than that for hydrogen molecule, but it must be taken into account that hydrogen atoms are obtained by the dissociation of a hydrogen molecule. The electron-impact excitation of the H2 leads to its dissociation.
3. ECR discharge in hydrogen The rate of hydrogenation depends on the type of the discharge, the power absorbed by the plasma, and the gas pressure. The chemical nonequilibrium of the hydrogen plasma can be described by taking into account the nine species H2, H, H(n=2), H(n=3), H+, H2+, H3+, H, and e, where H(n=2), H(n=3) are excited species and e is the electron [5]. The ECR discharges have very high degrees of ionization, d IV0.06 and work at low pressures (0.1–10 mTorr). In the given pressure range for the ECR discharges, the gas temperature, T g, is less than 600 K for absorbed power density up to 10 W/cm3. The rates of the chemical reactions, which involve electron-heavy particle collisions, depend on the electron energy distribution function (EEDF).
Fig. 7. A perspective view of an open ECR-PIII. The electromagnetic, electronic, and pneumatic elements and the roughing pumps are not shown. 1–water-cooled processing chamber; 2–inlets for homogeneous distribution of process gases in the processing chamber; 3–wafer load-lock system; 4–protection grid moving system; 5–high-voltage feed through for applying short HV negative pulses to the protection grid-wafer assembly without damaging the sensitive ceramic layer of the electrostatic chuck; 6–turbomolecular pump. Two units placed symmetrically; 7–gate valve. Two units placed symmetrically; 8–electrostatic chuck; 9–fan for cooling the slotted coaxial line and permanent magnets mounted in the ceramic tube of the linear microwave source. Each microwave source has its own cooling fan.
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
179
3.2. ECR discharge with ion flux ratio H+/H2+N1
4. Conclusion
It is possible to find a regime for the H2 ECR discharge where the relative ion flux ratio H+/H2+N1. Cielaszyk et al. [9] have measured the relative ion flux ratio H+/H2+ vs. pressure by mass spectrometer peak heights. As the pressure in the ECR discharges is decreased from 2 to 0.25 mTorr, an increase in the flux of H+ to the substrate is observed. In addition, the energies of both ionic species reaching the substrate increase with decreasing pressure. At 0.25 mTorr, 75% of the implanted species should be H+ ions. It may be mentioned here that with appropriate microwave power and gas pressure ECR plasma can generate up to 100% H+ ions [10].
An ECR plasma immersion ion implanter (ECR-PIII) for fabrication of thin layer bonded SOI wafers is proposed. The key element of the ECR-PIII is the scalable array of elementary ECR plasma sources created by an assembly of m linear microwave sources. Such a large-area plasma source generates highly homogeneous dense plasma with dimensions permitting processing of 300-mm wafers. The results of computer simulation of the distribution of the electric field, the magnetic induction lines, and the radiated microwave power of a linear microwave source have shown the feasibility of the scalable area of elementary ECR plasma sources. The ECR-PIII ensures high H+ ion implantation dose of the order of 1012/cm2 in the one-step process, e.g., the Smart-Cutk method. The ECR-PIII can also be used for fabrication of thin-layer SOI wafers by the two-step process, i.e., low-dose implantation of noble gas ions for formation of buried trap layer for hydrogen and hydrogenation with H ions at energy ~20 keV and temperatures in the range of 200–400 8C.
3.3. Estimation of implantation dose of H+ ions in ECR-PIII For an estimate of the dose of H+ ions in ECR-PIII, we have used the models developed by Scheuer et al. [11], Lieberman[12], and Briehl and Urbassek [13]. In an ECRPIII, short negative pulses are applied to the target which is immersed in the plasma. The electrons rapidly respond to the electric field produced by the negative pulse, and an ion matrix sheath (IMS) is created. The extent of the IMS at the pulse front, S 0, is given by S0 ¼ ½2eo V =ðe ni Þ1=2 ; where e o=8.851012 F/m, e=1.61019 C, V is the pulse amplitude, and n i is the ion volume density. The extent of the ion matrix, S t , for time t is:
Acknowledgments The authors thank Dr. A. Krasnikh (SLAC, Stanford University) for specific simulations of electric field, magnetic induction lines, and radiated power in a linear microwave source and for helpful discussions.
St ¼ S0 ½ð2=3Þxpi t þ 11=3 where x pi is the ion plasma frequency. The estimate was made for a H2 pressure of 0.25 mTorr. At this pressure, the density of the molecules, N=0.8851013/cm3. For the degree of ionization of 0.06, the ion density, n i=51011/ cm3 . It is to be noted that 75% of the ions at 0.25 mTorr are H+ ions, and the rest (25%) are H2+ ions. For this estimate, we assume that the IMS starts from the plasma bulk. Now, S 0=0.332 cm for V=5104 V and S t =S 0 [(2/ 3)(x H++x H2+)t+1]1/3=6.53 cm at t=10 As. The average H+ ion current per each square centimeter area of the target surface for one pulse having a duration, t=10 As, and an amplitude of 5104 V is J av=0.75en iS t /t=39.2 mA. If all H+ ions that reach the target have energy greater than the threshold for implantation, we obtain a dose per pulse, D=2.451012/cm2. However, the total ion current to the target is always greater than the implanted current. An estimate for the dose per pulse, D, with the assumption that the IMS starts from the steady-state (SS) sheath, where the average ion density, n*=0.6n i, gives the result D=9.31011/cm2.
References [1] M. Bruel, US Patent 5,374,564 (1992). [2] A. Usenko, et al., Proc. IEEE/SEMI Conf. and Workshop, Boston, MA, Institute of Electrical and Electronics Engineers (IEEE), Piscataway, NJ, 2002, pp. 6. [3] C. Quan, B. Terreault, in: Proc. Mat. Research Soc. Symposium, 585, 2000, pp. 177. [4] A. Usenko US Patent 6,352,909 B1 (2002). [5] K. Hassouni, et al., J. Appl. Phys. 86 (1999) 134. [6] K. Wiesemann, private communication. [7] http://physics.nist.gov/cgi-bin/Ionization/graph-new.pl?element=H.0. [8] http://physics.nist.gov/cgi-bin/Ionization/graph-new.pl?element= H2%2B.0. [9] E.S. Cielaszyk, et al., Appl. Phys. Lett. 67 (1995) 3099. [10] J. Pelletier, in: A. Anders (Ed.), Handbook of Plasma Immersion Ion Implantation and Deposition, vol. 7, Wiley, New York, 2000. [11] J.T. Scheuer, M. Shamin, J.R. Conrad, J. Appl. Phys. 67 (1990) 1241. [12] M.A. Lieberman, J. Appl. Phys. 66 (1989) 2926. [13] B. Briehl, H. Urbassek, Surf. Coat. Technol. 156 (2002) 131.
Characterization of a high-density plasma immersion ion implanter with scaleable ECR large-area plasma source Yuri Glukhoya,1, Mahmud Rahmanb,*, Gotze Popova,1, Alexander Usenkoc,2, Hans J. Walitzkid,3 b
a American Advanced Ion Beam, Inc., 440 Arguello Blvd, #1, San Francisco, CA 94118, USA Department of Electrical Engineering, Electron Devices Laboratory, Santa Clara University, 500 El Camino Real, Santa Clara, CA 95053 0583, USA c Silicon Wafer Technologies, Inc., 240 King Blvd., Newark, NJ 07102, USA d Isonics Corporation, 12514 NE 95th Street C-110, Vancouver, WA 98682, USA
Available online 2 November 2004
Abstract A key issue for future high-performance CMOS techniques is the fabrication of 12 in. or larger silicon-on-insulator (SOI) wafers with thickness of the top silicon layer under 25 nm. The electron cyclotron resonance (ECR) plasma ion immersion implanters (ECR-PIII) are viable candidates to realize the above. We designed an ECR-PIII in which we integrated the new concept of a large-area plasma source. This new plasma source is an array of elementary ECR plasma sources created by an assembly of m linear microwave sources. Each linear microwave source has n radiating elements (nNm) and includes a system of permanent magnets that create magnetic induction for ECR to occur in the processing chamber. An array of 90 elementary ECR plasma sources generate highly homogeneous dense plasma with dimensions permitting processing of 12-in. wafers. Since hydrogen implantation requiring high dose often results in blistering of the wafer surface, a new Smart-Cutk-like technology has been suggested (US Patent 6,352,909) where the one-step hydrogen implantation is replaced by a two-step process. The designed ECR-PIII can offer substantial advantage in providing the high yield of protons used for the micro-bubble formation in the two-step process. D 2004 Elsevier B.V. All rights reserved. Keywords: Electron cyclotron resonance (ECR) plasmas; Plasma immersion ion implantation (PIII); Microwave; Hydrogen; Silicon; Silicon-on-insulator (SOI)
1. Introduction We have designed an electron cyclotron resonance (ECR) plasma ion immersion implanter (ECR-PIII) particularly suitable for fabrication of 12-in. bonded thin-layer silicon-on-insulator (SOI) wafers. As the semiconductor industry moves from 0.13 Am to b50 nm and beyond line width, thin-layer SOI is expected to be the technology of choice for high-performance CMOS systems. The ECR-
* Corresponding author. Tel.: +1 408 554 4175; fax: +1 408 554 5474. E-mail addresses: [email protected] (Y. Glukhoy)8 [email protected] (M. Rahman)8 [email protected] (G. Popov)8 http://www.si-sandwich.com, [email protected] (A. Usenko), [email protected] (H. Walitzki). 1 Tel./fax: +1 415 751 7666. 2 Tel.: +1 973 297 1410; fax: +1 973 297 1125. 3 Tel.: +1 360 885 9310x205; fax: +1 360 885 9311. 0257-8972/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.surfcoat.2004.08.177
PIIIs are anticipated to replace the standard ion implanters for fabrication of 12 in. and larger thin-layer SOI wafers. The most advanced methods for manufacturing bonded SOI wafers are the layer transfer methods. Smart-Cutk [1] was the first layer transfer method described in the literature. It gives high-quality top layer in a final SOI wafer. However, the Smart-Cutk is not well suited for making SOI wafers with a top layer thinner than 100 nm. Recently, Usenko et al. [2] suggested an improved layer transfer method which allows manufacturing of SOI wafers with thinner top layers required for next-generation chips. The layer transfer method consists of the following steps: ! forming a fragile plane layer inside of a Si wafer called bdonor waferQ ! forming a silicon dioxide layer on top of another Si wafer, called bhandle waferQ
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
! bonding the top side of the donor wafer to the top side of the handle wafer, bsandwichingQ the silicon dioxide layer in between ! separation of the donor wafer along the fragile plane. In this way, a bonded SOI wafer consists of a handle wafer with buried oxide layer and the top layer that has been separated from the donor wafer. The separation techniques depend on the how the fragile plane layer is formed. In the Smart-Cutk method, the fragile layer is formed by hydrogen ion implantation at high ion dose such as 51016/cm2 and at an ion dose rate less than 1015/cm2 s. The depth to which the hydrogen ions penetrate into silicon determines the inherent thickness of the transferred layer. For typical ion implantation, energies between 30 and 100 keV, the corresponding depth is 0.3–1 Am. Implantation at low energy raises several problems. When a high-energy ion reaches the target Si, it dissipates its energy mostly because of interactions with the electronic subsystem of the Si crystal. The energy of the penetrating ion gradually decreases as its energy is transferred mostly to the target’s electrons. When the ion energy drops to i10 keV, the ion produces mostly atomic displacements. Ion implantation at low energy, e.g., 10 keV and less, produces effective atomic displacements also at the surface of the target resulting in the formation of a damaged layer that is not buried under the target’s surface. This damaged surface has an increased roughness and cannot be bounded to another (handle) wafer [3]. The process for ultra-thin SOI fabrication proposed by Usenko et al. consists of the following steps (also see Usenko, US Patent 6,352,909 B1 [4]): ! forming a buried trap layer for hydrogen in the donor wafer substrate ! nucleation of hydrogen platelets in the above buried trap layer by hydrogenation at a temperature less than 250 8C ! growing hydrogen platelets in the buried trap layer by hydrogenation at temperatures in the range of 250–400 8C, thus forming a buried fragile layer in the donor wafer ! forming a silicon dioxide layer on the top surface of the handle wafer ! activating the top surfaces of donor and handle wafers ! forming a temporary wafer assembly by making initial contact of the activated surfaces of the donor and handle wafers ! separating the above wafer assembly into two wafers by cleaving the assembly at the buried fragile layer. The buried trap layer for hydrogen is created by ion implantation. The ions used for such implantation could be either those that do not have electrical activity in the Si crystal, e.g., oxygen ions, or the same ions as the substrate, i.e., Si, or noble gas ions that do not chemically interact with silicon. Hydrogen is delivered to the buried trap layer from a
173
hydrogen plasma, and this defect layer serves as an infinite capacity trap for hydrogen. The ECR-PIII can be used for manufacturing SOI wafers by the Smart-Cutk method as well as thin-layer SOI wafers by the Usenko’s process. For the Usenko process, ECR-PIII can be used for both noble gas ion implantation and hydrogenation at ion energies of approximately 20 keV and at the temperatures mentioned above. The ECR-PIII consists of the following systems: ! processing chamber, which includes an assembly of m linear microwave sources used to create a highly homogeneous dense plasma ! high-vacuum system ! gas delivery system ! wafer transfer and positioning system ! microwave generators and HV power supplies for the generators ! pulsed HV generator for application of short negative HV pulses at the implantation target ! a computer controller and PLC that ensure the proper function of all subsystems.
2. The scalable plasma source Microwaves can generate and sustain plasma at low pressures. In the absence of a magnetic field, the plasma density, n, is limited by the frequency of the microwave, x, according to the following relationship: n V nc ¼ x2 eo m=e2 ; where n c is the critical plasma density, e o is the dielectric constant in vacuum, and m and e are the mass and the charge of an electron, respectively. When a steady magnetic field B is now applied to the microwave plasma, an electron cyclotron resonance (ECR) occurs. This is a resonance between the applied microwave frequency x and the cyclotron frequency x c=eB/m. At ECR conditions, the electrons rotate along the magnetic field lines in phase with the right-hand circularly polarized microwaves and obtain sufficient energy to ionize the ambient gas. The injection of microwave along the magnetic field allows it to propagate in a dense plasma and create plasma densities Nn c. A standard ECR plasma source includes a cylindrical chamber. The microwave power is applied to the chamber via a low-loss dielectric window positioned perpendicular to the chamber axis. Magnetic field coils generate nonuniform axial magnetic field B(z) within the chamber, and its magnetic induction can achieve the ECR condition in one or more points on the axis. In such a configuration, it is almost impossible to obtain a homogeneous plasma with dimensions that permit processing of 300-mm wafers. Our approach to this problem is based on the new concept of
174
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
using an array of elementary ECR plasma sources to realize a large area homogeneous plasma. 2.1. Geometry of the array of elementary plasma sources An array of elementary ECR plasma sources is the key element of the ECR-PIII proposed in our work. This is a fully scalable plasma source that is constructed using a rectangular array of n m elementary ECR plasma sources, where nNm. This array of elementary ECR plasma sources is created by an assembly of m linear microwave sources. Each linear microwave source in our design is powered by its own microwave generator at 2.45 GHz and includes n equally spaced elementary microwave sources. The details of the design are discussed further below. Fig. 1 shows an ECR plasma source array for 200-mm wafers consisting of 36 (4 on the short side and 9 on the long side of the rectangular array) elementary ECR plasma sources The distance between two adjacent elementary plasma sources on the short side of the rectangular array is D=3 in., whereas that on the long side is d=k/4 (=3.04 cm). The distances D and d depend on the frequency of the microwave used and the power of the microwave generators. The microwave power distribution and the position of the wafer are so chosen as to ensure homogeneity of the plasma better than 98% over the entire wafer surface. In order to ensure the homogeneity of the plasma in case of 300-mm wafers, we have also designed an ECR plasma source array consisting of 90 elementary ECR sources. 2.2. Linear microwave source Each linear microwave source consists of a microwave coaxial line with uniformly spaced slots cut at 1208 in the outer electrode of the line. A microwave generator supplies power to the slotted microwave line whence the slots emit the microwave power. The distance between two adjacent slots is ~k/4=3.04 cm for a frequency, f=2.45 GHz. Two
Fig. 1. An array of 36 elementary ECR plasma sources produced by 4 linear microwave sources. Each linear microwave source has 9 radiating elements.
permanent ring magnets are installed around each slot in the NS position. The inner diameter of the ring magnet is equal to the outer diameter of the outer line electrode. The magnetic induction, B r corresponding to the ECR condition (B r=0.0875 T at f=2.45 GHz) is achieved in the processing chamber at about 3 cm away from the axis of the coaxial line. A ceramic tube with an outer radius of 19 mm covers the slotted coaxial line and the permanent magnets. The ceramic tube also serves as a low-loss window for applying microwave in the low-pressure processing chamber. Fig. 2a shows the longitudinal cross section of a section of a microwave source with only four slots. The core radius, r=4.25 mm, and the radius of the outer line electrode, R=9.85 mm. The distance between two adjacent slots is 3.04 cm, the slot thickness is 3.5 mm, and the gap between the two permanent magnets installed around each slot is 6.5 mm. Fig. 2b shows a section in the center of the slot perpendicular to the axis of the coaxial line. 2.3. Simulation of a linear microwave source In order to ascertain the feasibility of a scalable array of elementary ECR plasma sources, a computer simulation of a linear microwave source with dimensions given in Fig. 2a and b was performed under the following assumptions: ! A linear microwave source consists of n radiating elements (slots) equally spaced at a distance d apart. ! Each radiating element has the same amplitude of E field. There is a phase shift d between the E fields of two adjacent radiating elements. ! An efficient linear microwave source transmits most of the microwave power to the plasma. ! A system of two ring magnets installed on the two sides of each radiating element creates a magnetic induction, B r=0.0875 T in the processing chamber at about 3 cm away from the axis of the slotted coaxial line. ! The plasma is considered as an absorbing layer with a thickness of 3 mm and an outer radius of 30 mm (see Fig. 2b). The superficial resistance of the absorbing layer varies between 10 and 50 V/cm2. Although these assumptions are inadequate to take into account the complicated interaction between the plasma and the microwave, nonetheless, they are good enough to demonstrate the preliminary feasibility of such a linear microwave source at f=2.45 GHz that can generate ECR plasmas with density up to 51012/cm3. Fig. 2c shows the resonance characteristics of the linear microwave source with dimensions as in Fig. 2a and b. The resonance occurs at f r=2.42 GHz. In the presence of a real plasma, the resonance is expected to be shifted slightly to higher frequency, f=2.45 GHz, which is the most used microwave frequency. Fig. 3a shows the distribution of the E field vectors in a section perpendicular to the line axis and passes through the
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
175
Fig. 2. (a) The longitudinal sectional view of a linear microwave source having four radiating elements (slots). The permanent ring magnets positioned on two sides of each slot are also shown. 1–core of the coaxial line, radius=4.25 mm; 2–slotted outer line electrode, radius=9.85 mm; 3–slot, cut at 1208 on the outer line electrode; 4–permanent ring magnet; 5–ceramic tube, outer radius=19 mm. (b) Cross section of a linear microwave source shown in (a) through the center of one slot. The plasma is represented as an absorbing layer. (c) S 11 as a function of the microwave frequency calculated for a linear microwave source shown in (a).
slot center. It is clearly seen how the E power passes through the ceramic tube and is absorbed in the absorbing layer. Fig. 3b shows the superposition of the magnetic fields of permanent magnets around three slots at positions 3a, 3b and 3c. At distances z3 cm, the magnetic induction lines are parallel to the coaxial line axis. Therefore, the elementary ECR plasma sources created by each linear microwave source will have nearly the same plasma density. Fig. 3c shows the radiated microwave power in a plane that passes through the coaxial line axis and the centers of the three slots at positions 3a, 3b and 3c. It can be seen here that the radiating elements (slots) emit microwave power.
Fig. 3d shows the radiated microwave power in the same plane as in Fig. 3a. The radiated microwave power in this figure conform to the distribution of the E field vectors shown in Fig. 3a. The above results of simulation have demonstrated clearly the feasibility of the design of a linear microwave source capable to create n elementary ECR plasma sources. Therefore, an assembly of m linear microwave sources would create an array of n m elementary ECR plasma sources. The minimum distance between two adjacent linear microwave sources is determined by the physical dimension of the slotted coaxial line and the safe vacuum seals of the ceramic tubes in the processing
176
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
Fig. 3. (a) Distribution of the E field vectors in the cross section of the slot in a linear microwave source with dimensions shown in Fig. 2a and b. Labels 1, 2, 3, 4, and 5 refer to same positions in all four figures (a–d). 1–inner electrode (core) of the slotted transmission line; 2–outer electrode of the slotted transmission line; 3–slot; 4–ceramic tube; 5–absorbing layer. (b) Magnetic induction lines as a result of superposition of magnetic fields created by the permanent magnets around three slots at positions 3a, 3b, and 3c. (c) Distribution of radiated microwave power in a plane that passes through the axis of the slotted coaxial line and the centers of slots 3a, 3b, and 3c. (d) Distribution of radiated microwave power from a single slot in a plane perpendicular to the axis of the linear microwave source.
chamber. In our design ( f=2.45 GHz), this distance is about 2.5 in. The superposition of n m elementary ECR plasma sources will create homogeneous plasma at some distance from the surface where the ECR condition occurs. Our estimate for this distance for 156=90 elementary ECR plasma sources with D=3 in. and d=3.04 cm for processing of 300-mm wafers is 2 in. away from the ECR surface. The 300-mm wafer will be processed at this distance in order to achieve plasma homogeneity better than 98%. Certainly, at this distance, the plasma density, n p will be lower, say, n p~0.6n r, where n r is the plasma density at the ECR surface. In our proposed design, we can introduce up to 30 kW microwave power for processing of 300-mm
wafers. This power is 15 times higher than the microwave power used in the standard cylindrical ECR plasma sources at f=2.45 GHz. Therefore, the n p we can achieve in our design is indeed much higher than the plasma density obtained in a standard ECR plasma source. 2.4. Design of a linear microwave source and an assembly of linear sources Fig. 4 shows a linear microwave source with 15 radiating elements (slots). The ceramic tube (1) is opened partially to show the design. The slots (3) of this microwave source are cut at 608 in the outer electrode (2). The thickness of the slot is 6.5 mm, and the gap between the two permanent magnets around the slot is also 6.5 mm. The spacers (5) and (6) fix
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
177
Fig. 4. A linear microwave source with 15 radiating elements (slots). 1–ceramic tube 2–outer electrode of the slotted transmission line; 3–permanent ring magnet; 4–slot; 5–half-ring spacer; 6–ring spacer.
the positions of the permanent magnets. The distance between two adjacent slots is ~k/4. It is imperative to use air-cooling to ensure working temperature below the Curie point of the permanent magnets. Fig. 5a and b shows the schematics of the montage of seven linear microwave sources in the processing chamber. In order to simplify the drawing, each linear microwave source is drawn with only seven radiating elements. The ceramic tubes that also serve as windows for application of
the microwave power in the chamber are mounted in the process chamber by means of vacuum-tight compression flanges. A gas delivery system ensures a homogeneous distribution of the process gas in the chamber. The load-lock system transfers and positions the wafer on the electrostatic chuck. The wafer is surrounded by a highly homogeneous plasma. Fig. 6 shows an assembly of six linear microwave sources. Each linear microwave source (5) has 15 radiating
Fig. 5. Montage of linear microwave sources in the processing chamber. (a) longitudinal sectional view (b) cross-sectional view.
178
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
The EEDF depends on the parameter E/N [V cm2], where N [cm3] is the density on the neutral atoms or molecules in the discharge. In ECR discharges E/NN1014 V cm2. At these conditions, the EEDF is non-Maxwellian, and the anisotropic part becomes important (availability of a magnetic field), and the calculation of the EEDF using the two-term expansion of the Boltzmann equation seems to be no more valid. The mean electron energy in the ECR discharge in the pressure interval of 0.2–10 mTorr is between 8 and 15 eV according to some measurements in the Ruhr-University, Bochum, Germany [6]. 3.1. Electron-impact ionization cross section for hydrogen atom Fig. 6. An assembly of six linear microwave sources. Each linear source has 15 radiating elements. 1–5 kW, 2.45 GHz water-cooled magnetron; 2– launcher; 3–three-stub tuners. Although three-stub tuners are used in this prototype, in the commercial product, computer-controlled step-motor driven tuners will be used; 4–transition from waveguide to coaxial line; 5– linear microwave source; 6–tuning piston. At this position, a fan for cooling the slotted transmission line and the permanent magnet is mounted.
elements (also see Fig. 4) and is powered by a 5-kW watercooled magnetron (1) via a three-stub tuner (3) and a transition from waveguide to coaxial line (4). This assembly has a total of 90 radiating elements. The distance between the axes of two adjacent linear microwave sources is 3 in. A wafer will be immersed in the chamber at distance ~3.2 in. from the plane formed by the axes of the linear microwave sources. Fig. 7 shows a perspective view of an open ECR-PIII. The electromagnetic, electronic, and pneumatic elements and the two roughing pumps are not shown in order to present the main assemblies clearly. The process chamber is pumped by two turbomolecular pumps (7). Two computercontrolled gate valves function in accordance with the loadlock system (3), the protection grid moving system (4), and the electrostatic chuck (8).
The electron-impact ionization cross section for hydrogen atom has its maximum, r max=0.61016 cm2 at an energy of 55 eV [7]. The electron-impact ionization cross section for hydrogen molecule ion (H2) varies similarly with energy with a maximum, r max=0.171016 cm2 at an energy of 100 eV [8]. The electron-impact ionization cross section for hydrogen atom is about four times larger than that for hydrogen molecule, but it must be taken into account that hydrogen atoms are obtained by the dissociation of a hydrogen molecule. The electron-impact excitation of the H2 leads to its dissociation.
3. ECR discharge in hydrogen The rate of hydrogenation depends on the type of the discharge, the power absorbed by the plasma, and the gas pressure. The chemical nonequilibrium of the hydrogen plasma can be described by taking into account the nine species H2, H, H(n=2), H(n=3), H+, H2+, H3+, H, and e, where H(n=2), H(n=3) are excited species and e is the electron [5]. The ECR discharges have very high degrees of ionization, d IV0.06 and work at low pressures (0.1–10 mTorr). In the given pressure range for the ECR discharges, the gas temperature, T g, is less than 600 K for absorbed power density up to 10 W/cm3. The rates of the chemical reactions, which involve electron-heavy particle collisions, depend on the electron energy distribution function (EEDF).
Fig. 7. A perspective view of an open ECR-PIII. The electromagnetic, electronic, and pneumatic elements and the roughing pumps are not shown. 1–water-cooled processing chamber; 2–inlets for homogeneous distribution of process gases in the processing chamber; 3–wafer load-lock system; 4–protection grid moving system; 5–high-voltage feed through for applying short HV negative pulses to the protection grid-wafer assembly without damaging the sensitive ceramic layer of the electrostatic chuck; 6–turbomolecular pump. Two units placed symmetrically; 7–gate valve. Two units placed symmetrically; 8–electrostatic chuck; 9–fan for cooling the slotted coaxial line and permanent magnets mounted in the ceramic tube of the linear microwave source. Each microwave source has its own cooling fan.
Y. Glukhoy et al. / Surface & Coatings Technology 196 (2005) 172–179
179
3.2. ECR discharge with ion flux ratio H+/H2+N1
4. Conclusion
It is possible to find a regime for the H2 ECR discharge where the relative ion flux ratio H+/H2+N1. Cielaszyk et al. [9] have measured the relative ion flux ratio H+/H2+ vs. pressure by mass spectrometer peak heights. As the pressure in the ECR discharges is decreased from 2 to 0.25 mTorr, an increase in the flux of H+ to the substrate is observed. In addition, the energies of both ionic species reaching the substrate increase with decreasing pressure. At 0.25 mTorr, 75% of the implanted species should be H+ ions. It may be mentioned here that with appropriate microwave power and gas pressure ECR plasma can generate up to 100% H+ ions [10].
An ECR plasma immersion ion implanter (ECR-PIII) for fabrication of thin layer bonded SOI wafers is proposed. The key element of the ECR-PIII is the scalable array of elementary ECR plasma sources created by an assembly of m linear microwave sources. Such a large-area plasma source generates highly homogeneous dense plasma with dimensions permitting processing of 300-mm wafers. The results of computer simulation of the distribution of the electric field, the magnetic induction lines, and the radiated microwave power of a linear microwave source have shown the feasibility of the scalable area of elementary ECR plasma sources. The ECR-PIII ensures high H+ ion implantation dose of the order of 1012/cm2 in the one-step process, e.g., the Smart-Cutk method. The ECR-PIII can also be used for fabrication of thin-layer SOI wafers by the two-step process, i.e., low-dose implantation of noble gas ions for formation of buried trap layer for hydrogen and hydrogenation with H ions at energy ~20 keV and temperatures in the range of 200–400 8C.
3.3. Estimation of implantation dose of H+ ions in ECR-PIII For an estimate of the dose of H+ ions in ECR-PIII, we have used the models developed by Scheuer et al. [11], Lieberman[12], and Briehl and Urbassek [13]. In an ECRPIII, short negative pulses are applied to the target which is immersed in the plasma. The electrons rapidly respond to the electric field produced by the negative pulse, and an ion matrix sheath (IMS) is created. The extent of the IMS at the pulse front, S 0, is given by S0 ¼ ½2eo V =ðe ni Þ1=2 ; where e o=8.851012 F/m, e=1.61019 C, V is the pulse amplitude, and n i is the ion volume density. The extent of the ion matrix, S t , for time t is:
Acknowledgments The authors thank Dr. A. Krasnikh (SLAC, Stanford University) for specific simulations of electric field, magnetic induction lines, and radiated power in a linear microwave source and for helpful discussions.
St ¼ S0 ½ð2=3Þxpi t þ 11=3 where x pi is the ion plasma frequency. The estimate was made for a H2 pressure of 0.25 mTorr. At this pressure, the density of the molecules, N=0.8851013/cm3. For the degree of ionization of 0.06, the ion density, n i=51011/ cm3 . It is to be noted that 75% of the ions at 0.25 mTorr are H+ ions, and the rest (25%) are H2+ ions. For this estimate, we assume that the IMS starts from the plasma bulk. Now, S 0=0.332 cm for V=5104 V and S t =S 0 [(2/ 3)(x H++x H2+)t+1]1/3=6.53 cm at t=10 As. The average H+ ion current per each square centimeter area of the target surface for one pulse having a duration, t=10 As, and an amplitude of 5104 V is J av=0.75en iS t /t=39.2 mA. If all H+ ions that reach the target have energy greater than the threshold for implantation, we obtain a dose per pulse, D=2.451012/cm2. However, the total ion current to the target is always greater than the implanted current. An estimate for the dose per pulse, D, with the assumption that the IMS starts from the steady-state (SS) sheath, where the average ion density, n*=0.6n i, gives the result D=9.31011/cm2.
References [1] M. Bruel, US Patent 5,374,564 (1992). [2] A. Usenko, et al., Proc. IEEE/SEMI Conf. and Workshop, Boston, MA, Institute of Electrical and Electronics Engineers (IEEE), Piscataway, NJ, 2002, pp. 6. [3] C. Quan, B. Terreault, in: Proc. Mat. Research Soc. Symposium, 585, 2000, pp. 177. [4] A. Usenko US Patent 6,352,909 B1 (2002). [5] K. Hassouni, et al., J. Appl. Phys. 86 (1999) 134. [6] K. Wiesemann, private communication. [7] http://physics.nist.gov/cgi-bin/Ionization/graph-new.pl?element=H.0. [8] http://physics.nist.gov/cgi-bin/Ionization/graph-new.pl?element= H2%2B.0. [9] E.S. Cielaszyk, et al., Appl. Phys. Lett. 67 (1995) 3099. [10] J. Pelletier, in: A. Anders (Ed.), Handbook of Plasma Immersion Ion Implantation and Deposition, vol. 7, Wiley, New York, 2000. [11] J.T. Scheuer, M. Shamin, J.R. Conrad, J. Appl. Phys. 67 (1990) 1241. [12] M.A. Lieberman, J. Appl. Phys. 66 (1989) 2926. [13] B. Briehl, H. Urbassek, Surf. Coat. Technol. 156 (2002) 131.