WO2006133132A2 - Combinations of plasma production devices and method and rf driver circuits with adjustable duty cycle - Google Patents

Abstract

A method and apparatus capable of combining one or more RF driver circuit and an orthogonal antenna assembly/configuration with conventional RF generators or other similar units are disclosed as part of a method and system for generating high density plasma for different power and coverage levels. The disclosed low-output impedance RF driver circuits eliminate or greatly reduce the need for a matching circuit for interfacing with the inherent impedance variations associated with plasmas. Also disclosed is the choice for capacitance or an inductance value to provide tuning for the RF plasma source. There is also provided a method for rapidly switching the plasma between two or more power levels at a frequencies of about tens of Hz to as high as hundreds of KHz.

Description

COMBINATIONS OF PLASMA PRODUCTION DEVICES AND METHOD AND RF DRIVER CIRCUITS WITH ADJUSTABLE DUTY CYCLE

REFERENCE TO RELATED APPLICATIONS

This application claims priority to United States Provisional Patent Application Nos. 60/687,560 filed on June 3, 2005, 60/715,779 filed on September 8, 2005 and 60/715,933 filed on September 9, 2005 under 35 U. S. C. §119 and this application is also a continuation-in-part of the United States Patent Application No. 10/874,096 filed on June 21 , 2004, which is a continuation-in-part of the United States Patent Application No. 10/419,052 filed on April 17, 2003, which is a continuation-in-part of the United States Patent Application No. 10/268,053 filed on October 9, 2002, which claims the benefit under 35 U. S. C. §119 of priority to United States Provisional Patent Application No. 60/328,249 filed on October 9, 2001 , all of which are incorporated herein in their entirety by reference.

FIELD OF THE INVENTION

The present invention relates generally to the design and implementation of a plasma generation system. More particularly, it relates to radio frequency amplifiers, antennas and effective circuit connections for interfacing combinations of the amplifiers and antennas for generating plasma.

BACKGROUND OF THE INVENTION

Plasma is generally considered to be a fourth state of matter, the others being solid, liquid and gas states. In the plasma state the elementary constituents of a substance are substantially in an ionized form rendering them useful for many applications due to, inter alia, their enhanced reactivity, energy, and suitability for the formation of directed beams.

Plasma generators are routinely used in the manufacture of electronic components, integrated circuits, and medical equipment, and in the operation of a variety of goods and machines. For example, plasma is extensively used (i) to deposit layers of a desired substance, for instance, following a chemical reaction or sputtering from a source, (ii) to etch material with high precision, (iii) to sterilize objects by the free radicals present in the plasma or induced by the plasma, and (iv) to modify surface properties of materials.

Plasmas are also used in approximately one third of the steps involved in manufacturing semiconductors. Plasma generators are routinely used in the manufacture of electronic components, integrated circuits, and medical equipment, and in the operation of a variety of goods and machines. For example, plasma is extensively used to deposit layers of a desired substance, for instance following a chemical reaction or sputtering from a source, to etch material with high precision, and to sterilize objects by the free radicals in the plasma or induced by the plasma or to modify surface properties of materials.

Plasma processing has been used for many years in a variety of industries, including electronics, aerospace, automotive, and biomedical applications, among others. Plasma techniques are used both in deposition and removal of material on many different types of substrates, from plastic sheeting to flat screen displays to silicon wafers. In the case of flat screen displays, the commercial market supports increasingly larger formats.

Plasma generators based on radio frequency ("RF") power supplies are often used in experimental and industrial settings since they provide a ready plasma source, and are often portable and easy to relocate. Such plasma is generated by coupling the RF radiation to a gas, typically at reduced pressure (and density), causing the gas to ionize. In any RF plasma production system, the plasma represents a variable load at the antenna terminals as the process conditions changes. Among other process control factors, changes in working gas and pressure affect the amount of loading seen at the antenna terminals. In addition, the amplitude of the RF drive waveform itself affects the plasma temperature and density, which in turn also affects the antenna loading. Thus the antenna/plasma combination represents a non-constant and nonlinear load for the RF power source to drive.

As the various industries develop their processes, they look to economies of scale to increase the yield of a given fabrication process. Examples of this include tools used in semiconductor processing, where originally wafers were 100 mm in diameter. Tools have evolved through stages where the typical wafer size was 150 mm, 200 mm, 300 mm, and manufacturers are now looking forward to 450 mm in the not-too-distant future. The scale-up has typically been accomplished by enlarging a reactor's single plasma source to encompass larger and larger substrates.

These applications have relied on increasingly larger scale plasma sources. Plasmas in these sources are created by coupling radio frequency (RF) electromagnetic energy into a chamber through an insulator. This RF energy serves to ionize a working gas or gases within the chamber, producing the useful products and processes for the target reactions. The RF energy is typically derived from an RF generator, with an output impedance of 50 ohms, coupled through a variable matching network to an antenna that produces and interacts closely with the plasma. The inventor of this patent application filed several patent applications (collectively "the Priority Patent Applications"), to which priority is claimed by this application. The priority patent applications disclose an improved technique of driving a plasma reactor without the need of a variable matching network or a match box.

To achieve larger sources, existing geometries the obvious solution has been to scale them up in a self-similar fashion. However, simply increasing the linear dimensions of the source is not sufficient to guarantee that uniformity will be maintained over the larger substrate even if the source being scaled up does provide sufficiently uniformity over a given substrate size. Various physical processes govern the RF production and loss of plasma, and these processes have characteristic lengths that do not scale usefully with increasing linear dimensions of the RF antennas.

This has been the case with e.g. Applied Materials' 300 mm etch tool, which required addition of a second independently controllable antenna in a concentric arrangement with the first. Applied Materials accomplished this by re-engineering the variable matching network. That is, a single RF generator provides the RF energy at 50 ohms; this energy is coupled through a specialized variable matching network having two outputs, to drive two separate concentric antennas. The sum and differences of the currents in the two antennas can then be manipulated to control the process and its uniformity over the full extent of a 300 mm wafer.

Applied Materials' technique is scalable to provide larger substrates with three, or four antenna elements, with each of these antenna elements driven by separate outputs from an increasingly sophisticated matching network. The matching networks will need to be specially redesigned in each case. Furthermore, this technique is not free of problems. For instance, each antenna, couples energy to other antennas, and the extent of coupling depends on plasma conditions. Thus, the variable capacitors in the matching network need to be readjusted in a non-obvious way to accomplish the target process uniformity. The combined adjustment must present an approximately 50 ohm, mostly resistive load to the RF power generator in order for the generator to operate correctly. Loads that are too reactive reflect power back to the RF generator, and typically require the generator to reduce its output power in order to keep this reflected power below some specified safe operating limit.

Thus, adding variable elements to the matching network represents an undesirable increase in complexity, not just in the design but also in the controllability of the process. For perspective, one antenna element requires two variable capacitors in the matching network, and much work has gone into developing computer-implementations of control algorithms that navigate in two-dimensional space (e.g. x = C1 tuning setpoint, y = C2 tuning setpoint) to reach a target operating point that is a nonlinear function of x and y. Adding a third capacitor makes this space 3-dimensional, with a necessary increase in complexity of the computer control algorithm. Higher numbers of outputs from a single variable matching network represent ever-increasing complexity.

Chen, (see e.g. Chen, F. F. et al, Plasma Sources Science and Technology, 10 (2001) 236-249.) discloses a fixed geometry for arranging multiple small helicon sources, each circle represents a small helicon source with a winding forming the antenna around an insulating ampoule. The , ampoules are mounted on a cover plate for the process reactor chamber. All of the multiple antennas are connected in series, so that all the individual small sources are driven from the same RF generator and variable matching network combination.

Fixed geometry techniques like those disclosed by Chen allow optimization of the antenna layout for a given process. However, this antenna layout is typically sub-optimal or even unusable for another chemistry or process. On the other hand, coupling several antennas, using dedicated separate RF generator/variable matching network combinations, is not only undesirably complicated (two separate 2-dimensional search spaces must be navigated for proper tuning and process control) but also prohibitively expensive for many applications

A typical RF source has a 50 ohm output impedance, and requires a load that presents a matching 50 ohm impedance in order for the RF source to couple to the load most efficiently. Because of the often unpredictable changes in the plasma self inductance, effective resistance, and mutual inductance to the antenna, provision for impedance matching is made by retuning some circuit elements and possibly the plasma to obtain satisfactory energy transfer from the RF source to the generated plasma. To achieve this, an adjustable impedance matching network, or "matching box" is typically used to compensate for the variation in load impedance due to changes in plasma conditions.

The matching box typically contains two independent tunable components, one that adjusts the series impedance and the other that adjusts the shunt impedance. These components must be adjusted in tandem with each other in order to achieve the optimum power transfer to the plasma. Not surprisingly, accurate tuning of these components is often a difficult process. Typically, retuning requires manual/mechanical operations/actuators to adjust one or more component values and generally sophisticated feedback circuitry for the rather limited degree of automation possible.

Application of a sufficiently large electric field to a gas separates electrons from the positively charged nuclei within the gas atoms, thus ionizing the gas and forming the electrically conductive fluid-like substance known as plasma. Coupling radio frequency electric and magnetic fields to the gas generates, via an antenna, induces currents within this ionized gas. This, in turn, causes the gas to further ionize, and thereby increasing its electrical conductivity, which then increases the efficiency with which the antenna fields couple to the charged particles within the gas. This leads to an increase in the induced currents, resulting in the electrical breakdown and substantial ionization of the gas by various mechanisms. The effectiveness of the RF coupling is dependent upon the particular RF fields and/or waves that are used. Some types of waves that are suitable for the efficient production of large volumes of plasma are described next. Whistler waves are right-hand-circularly-polarized electromagnetic waves (sometimes referred to as R-waves) that can propagate in an infinite plasma that is immersed in a static magnetic field B₀. If these waves are generated in a finite plasma, such as a cylinder, the existence of boundary conditions - i.e. the fact that the system is not infinite - cause a left-hand- circularly-polarized mode (L-wave) to exist simultaneously, together with an electrostatic contribution to the total wave field. These "bounded Whistler" are known as Helicon waves. See Boswell, R.W., Plasma Phys. 26, 1147 (1981). Their interesting and useful qualities include: (1) production and sustenance of a relatively high-density plasma with an efficiency greater than that of other RF plasma production techniques, (2) plasma densities of up to Np ~ 10¹⁴ particles per cubic centimeter in relatively small devices with only a few kW of RF input power, (3) stable and relatively quiescent plasmas in most cases, (4) high degree of plasma uniformity, and (5) plasma production over a wide pressure range, from a fraction of a mTorr to many tens of mTorr. Significant plasma enhancement associated with helicon mode excitation is observed at relatively low B₀-fields, which are easily and economically produced using inexpensive components.

Significant plasma density (N_p) enhancement and uniformity may be

achieved by excitation of a low-field m = +1 helicon R-wave in a relatively compact chamber with B₀ < 150 G. This may be achieved, for instance, through the use of an antenna whose field pattern resembles, and thus couples to, one or more helicon modes that occupy the same volume as the antenna field. The appropriate set of combined conditions include the applied magnetic field B₀, RF frequency (F_RF), ), the density N_p itself, and physical

dimensions.

Some antenna designs for coupling RF power to a plasma are disclosed by United States Patent Nos. 4,792,732, 6,264,812 and 6,304,036. However, these designs are relatively complex often requiring custom components that increase the cost of system acquisition and maintenance.

RF power sources typically receive an external RF signal as input or include an RF signal generating circuit. In many processing applications, this RF signal is at a frequency of 13.56 MHz, although this invention is not limited to operation at this frequency. This signal is amplified by a power output stage and then coupled via an antenna to a gas/ plasma in a plasma generator for the production of plasma.

Amplifiers are conventionally divided into various classes based on their performance characteristics such as efficiency, linearity, amplification, impedance, and the like, and intended applications. In power amplification, an important concern is the amount of power wasted as heat, since heat sinks must be provided to dissipate the heat and, in turn, increase the size of devices using an inefficient amplifier. A classification of interest is the output impedance presented by an amplifier since it sets inherent limitations on the power wasted by an amplifier.

Typical RF amplifiers are designed to present a standard output impedance of 50 Ohms. Since, the voltage across and current through the output terminals of such an amplifier are both non-zero, their product provides an estimate of the power dissipated by the amplifier. This product can be reduced by introducing a phase difference between the voltage and the current across the output terminals of the amplifier in analogy with the power dissipated in a switch. In contrast to conventional amplifiers, a switch presents two states: it is either ON, corresponding to a short circuit, i.e., low impedance, or OFF, corresponding to an open circuit, i.e., infinite (or at least a vary large) impedance. In switched mode amplifiers, the amplifier element acts as a switch under the control of the signal to be amplified. By suitably shaping the signals, for instance with a matching load network, it is possible to introduce a phase difference between the current and the voltage such that they are out of phase to minimize the power dissipation in the switch element. In other words, if the current is high, the voltage is low or even zero and vice versa. United States Patents Nos. 3,919,656 and 5,187,580 disclose various voltage/current relationships for reducing or even minimizing the power dissipated in a switched mode amplifier.

United States Patent No. 5,747,935 discloses switched mode RF amplifiers and matching load networks in which the impedance presented at the desired frequency is high while harmonics of the fundamental are short circuited to better stabilize the RF power source in view of plasma impedance variations. These matching networks add to the complexity for operation with a switched mode power supply rather than eliminate the dynamic matching network. Such a matching load network is also not very frequency agile since it depends on strong selection for a narrow frequency band about the fundamental. United States Patent No. 6,432,260 discloses use of switched elements in matching impedance networks to ensure that the dynamic complex impedance of the plasma is seen as a near resistive value, effectively neutralizing the reactive components of plasma impedance. This allows a power source to only respond to resistive changes in the plasma since it is only such changes that are seen by the power source. The dynamic plasma resistance controls the power delivered to the plasma.

When plasma impedance is a small fraction of the impedance seen by the RF source, variations in plasma impedance are a smaller factor. Thus, it is possible to drive a plasma with an RF power supply without an intervening dynamic matching network if the dynamic plasma impedance is a small fraction of the total impedance seen by the source or if a significant fraction of the power supplied to the plasma is via a relatively directly driven plasma. Overwhelming the plasma inductance results in compromising efficiency to some extent. As a result, a matching network is required when the dynamic plasma impedance is a significant fraction of the total impedance seen by the RF power source.

United States Patent Nos. 6,150,628, 6,388,226, 6,486,431 , and 6,552,296 disclose constant current switching mode RF power supply containing an inductive element in series with the plasma load. The plasma is primarily driven as the secondary of a iron- or ferrite-core transformer, the primary of which is driven by the RF power supply. In such a configuration, dynamic impedance matching network is disclosed to be not required. The current through the plasma is maintained at about the value of the initial inductor current to adjust the power based on the size of the load. Also disclosed in these patents are various methods for igniting a plasma that include high voltage pulses, ultra-violet light and capacitative coupling, which also serve to restrict variations in plasma impedance by sidestepping the large impedance variations encountered upon plasma ignition.

There are other known designs that use the plasma as a secondary in a transformer like design in which the secondary and the primary are relatively weakly coupled via a shared core. R. J. Taylor invented a plasma production technique for cleaning the inside of a toroidal vacuum chamber using a process plasma, and had built such a device in 1973. The circuit as its transformer primary used the air-core Ohmic Heating (OH) winding of a tokamak, and a matching network consisting of fixed C1 and C2. Similar designs operating on other tokamaks, some having iron-core transformers, are known. These designs typically operated in the frequency range 10-50 kHz.

In designs similar to those of R. J. Taylor, the changes in the plasma impedance do not significantly affect the loading of the driver because the

parameter Mi_p is the mutual inductance between the

primary inductance Li and the plasma inductance L_p, is quite small. Consequently, variation in the inductive load seen at the terminals of the transformer primary is smaller. In contrast, when the plasma is substantially

directly driven, e.g., via current-straps, where _{^ wnere}j_n M_ant_

plasma is the mutual inductance between the antenna inductance L_ant and the plasma inductance L_pι_asma, is not small, and as a result changes in the plasma impedance represent relatively large changes in the load impedance seen by the RF source. This variation typically requires the use of variable matching network to provide a reasonable match with a 50 Ohm impedance of the RF source for delivering power.

When plasma is driven directly, i.e., without a core for substantially coupling a plasma secondary to a primary winding connected to the RF source, changes in plasma impedance are significant at the leads of the antenna or at the primary winding of a coupling transformer. This configuration has been coupled to a plasma or plasma/antenna combination via a dynamic matching network to continually adjust in response to the changing plasma impedance.

The problems faced in efficient plasma generator design include the need for a low maintenance and easily configured antenna, the elimination of expensive dynamic matching networks for directly coupling the RF power source to the non-linear dynamic impedance presented by a plasma, combining multiple plasma sources to provide desired power and coverage, and the need for RF power sources that can be efficiently modulated and are frequency agile.

SUMMARY OF THE INVENTION

It is therefore an object of the present disclosure to provide an improved antenna design for efficiently coupling RF sources to a plasma. It is yet another object of the present disclosure to provide systems for generating a plasma with the aid of an RF power source without requiring the use of a matching network to couple the RF power source to the plasma. It is yet another objective of the present disclosure to describe apparatus and methods for combining a plurality of plasma sources to generate plasma sources with a desired power level and coverage.

This application discloses an alternative apparatus and method for driving multiple antenna elements and overcoming the above-described drawbacks. The modifications described herein extend the disclosure of the Prior Patent Applications to drive the individual current elements of a multiple- antenna plasma source. Briefly, each source element may be independently controlled; expenses due to the use of one or more variable matching networks are significantly reduced or even eliminated; control of RF current in each antenna element corresponds (e.g., proportional) to a control input and even if the control input may be affected by mutual inductances or plasma- driven coupling to other sources; and such an arrangement is scalable; reduces the need for increased output power from single RF supplies for increasingly large scale reactors - e.g., inexpensive 1 kilowatt RF sources could be individually connected to each antenna element, which makes more sense from an RF transistor usage standpoint because increasing the output power of a single RF generator usually entails adding and combining separate transistor output stages. Instead of dynamic impedance adjustment capable matching network, a reactive network couples the RF power source to the antenna-plasma combination. The reactive network is selected so that at at least a first plasma impedance value, a substantially resistive load is presented to the RF power source. Further at a second plasma impedance value, preferably, selected so as to significantly cover the expected dynamic plasma reactance range, the reactance seen by the RF power source is about the same as that of the RF power source itself. Thus, disclosed herein are a method of designing a reactive circuit to eliminate the need for a dynamic matching circuit between a plasma and a RF power source. Also disclosed is a reactive circuit suitable for a plasma generator operating at about 13.56 MHz. The described method is also applicable for designing reactive circuits for many frequencies other than 13.56 MHz.

In addition to the plasma impedance, other considerations may also be taken into account in the design of the reactive circuit. For instance, it may be designed so as to present a phase difference at the switched power supply, the RF power source, since this improves the efficiency of the power supply by reducing resistive losses at the switches. Such additional conditions may, in general require the values of three or more reactance elements to be determined for providing the desired behavior.

An illustrative plasma generator system in accordance with one embodiment of the present invention comprises at least one plasma source, the at least one plasma source having an antenna including a plurality of loops, each loop having a loop axis, the plurality of loops arranged about a common axis such that each loop axis is substantially orthogonal to the common axis; at least one radio frequency power source for driving the plurality of loops in quadrature and coupled to a plasma load driven in a circularly polarized mode, preferably a helicon mode, via the antenna; a static magnetic field substantially along the common axis; and a reactance coupling the switching amplifier to the antenna loops such that the reactance and the antenna loops without the plasma have a resonant frequency that is about equal to a specified frequency and dispensing with the requirement for a matching network. The reactance coupling the switching amplifier to the antenna loops is preferably provided at least in part by a capacitor.

The radio frequency power source preferably comprises at least one member from the group consisting of a substantially Class A amplifier, a substantially Class AB amplifier, a substantially Class B amplifier, a substantially Class C amplifier, a substantially Class D amplifier, a substantially Class E amplifier, and a substantially Class F amplifier. In one embodiment, these are connected to the primary of a transformer to reduce the drive impedance to a low value. Even more preferably the radio frequency power source includes a Class D amplifier in a push-pull configuration with a relatively low output impedance.

In a preferred embodiment, the radio frequency power source exhibits a low output impedance in comparison with the input impedance of an antenna. Often the low output impedance is significantly less than the standard impedance of 50 Ohm. The output impedance is preferably within a range selected from the set consisting of less than about 0.5 Ohms, less than about 2 Ohms, less than about 3 Ohms, less than about 5 Ohms, less than about 8 Ohms, less than about 10 Ohms, and less than about 20 Ohms. Preferably the output impedance is less than 5 Ohms, even more preferably the output impedance is between 0.5 to 2 Ohms, and most preferably the output impedance is less than 1 Ohm. Use of this low-impedance driver together with the disclosed circuit for connecting the driver to the current strap of an antenna eliminate the need for a match box, thus reducing circuit complexity and eliminating a source of failure in plasma processing systems.

A further advantage of the disclosed system is that the voltage applied to the antenna can be made quite large prior to plasma formation, thus increasing the ability to initiate the plasma in a variety of working conditions. Once the plasma is formed, the voltage reduces to a lower level to sustain the plasma, mitigating the harm resulting from possible high antenna voltages.

Depending upon the phasing between antenna elements and the value of B₀, the system can be run as a helicon source, or as a magnetized inductively coupled plasma (MICP) source, or as an ICP source at Bo = 0. Furthermore, it is observed to operate efficiently and robustly in pressure regimes (e.g., with P₀ approximately 100 mTorr) that are very difficult to access and/or make good use of using prior art plasma sources. The currents in the antenna elements appear to abruptly "lock" into a quadrature excitation mode when the conditions on neutral pressure Po, input power PRF, and externally applied axial magnetic field B₀, are right. When this occurs, the plasma appears to fill the chamber approximately uniformly, which is advantageous over other sources due to the ability to produce uniform processing conditions.

Additionally, the combination of antenna system plus RF generator can create and maintain a plasma under conditions where the plasma parameters vary over much larger ranges than have been reported for other sources (e.g. neutral pressure Po varied from 100 mTorr down to 5 mTorr, and then back up again to 100+ mTorr, in a cycle lasting approximately one minute), without the need for the adjustment of any matching network components.

Another advantage of the disclosed system is that the elimination of the matching network can result in an "instant-on" type of operation for the plasma source. This characteristic can be used to provide an additional control for the process being used. In particular, it is possible to modulate the amplitude of the RF power that is generating the plasma, between two (or more) levels such as 30% and 100%, or in a fully on-off manner (0% to 100%). This modulation can occur rapidly, e.g. at a frequency of several kilohertz, and can accomplish several purposes. For instance, the average RF power can be reduced with a consequent reduction in average plasma density. The "instant-on" operation can generate plasma with an average RF input power of as little as 5 W to a volume of 50 liters.

In addition, modulation can be used to control the spatial distribution of the working gas within the reaction chamber: The distribution of the working gas is modified by the plasma, which often contributes to the non-uniformity of fluxes of the active chemicals or radicals. By modulating the duty cycle of plasma production, the flow characteristics of the neutral gas during the plasma off time (or reduced-power-level time) can be adjusted to control the uniformity of the process by way of the duty cycle. Since the plasma initiation time is usually within 10-20 microseconds of the application of the RF, the duty cycle may be controlled at frequencies as high as tens or hundreds of kHz. In another embodiment, the increase in complexity from adding an antenna to an existing single antenna plasma source configuration is reduced by using the RF generator disclosed herein. In this case, a prior art control system may maintain its two-dimensional control for the first antenna, while the additional antenna is controlled by a direct drive RF source described herein, which is able to directly set the level of the current in the antenna to a specified value. The current in this second antenna is programmably controlled to depend on the current in the first.

In another embodiment, the power delivered by the second antenna is programmably controlled by the power delivered by the first antenna. A preferred embodiment of this system uses an additional set of windings either on the plasma chamber (as a third antenna) or situated away from the chamber, having an equal and opposite mutual inductance to the first antenna configuration, so that current driven in the second antenna is substantially decoupled from current driven in the first. This embodiment enables scale-up of a known prior art system with minimal changes to the control system for the first antenna, but enables additional control of process uniformity. Another embodiment has more than one additional antenna, with each controlled by direct drive RF generators as described herein. Another embodiment has multiple antenna elements driven by a prior art system, together with one or more antennas controlled by one or more direct drive RF generators of this invention.

These and other features of the invention are described next with the help of the following illustrative figures. BRIEF DESCRIPTION OF THE FIGURES

The following illustrative figures are provided to better explain the various embodiments of the invention without intending for the figures to limit the scope of the claims.

FIGURE 1 illustrates a plasma source chamber with two sets of antenna elements;

FIGURE 2 illustrates a tunable circuit with an RF power source coupled to an antenna;

FIGURE 3 illustrates a second tunable circuit with an RF power source coupled to an antenna;

FIGURE 4 illustrates a third tunable circuit with an RF power source coupled to an antenna;

FIGURE 5 illustrates a circuit with an RF power amplifier coupled to an antenna current strap;

FIGURE 6 illustrates a second circuit with an RF power amplifier coupled to an antenna current strap;

FIGURE 7 illustrates a third circuit with an RF power amplifier coupled to an antenna current strap;

FIGURE 8 illustrates a simplified model of the RF power amplifier, antenna current strap, and plasma;

FIGURE 9 illustrates a lumped circuit equivalent of the model depicted in FIGURE 8; FIGURE 10 illustrates the frequency response of a plasma source without a plasma;

FIGURE 11 illustrates the frequency response of a plasma source with a plasma present;

FIGURE 12 illustrates a feedback arrangement for controlling a plasma source.

FIGURE 13 illustrates a reactive network for coupling an RF power source to a plasma; and

FIGURE 14 shows an illustrative embodiment of the invention with elements selected in a reactive network to eliminate the need for a dynamic matching network.

FIGURE 15A-D show some possible illustrative arrangements of combined plasma sources.

FIGURE 16 shows direct drive RF generators mounted on or around a hemispherical insulating dome, a non-planar arrangement.

FIGURE 17 shows two illustrative DC supplies coupled by a switch to provide high output voltage at low currents and low voltages and high currents without requiring use of DC power supplies that are rated significantly in excess of the required specifications for many different applications.

FIGURE 18 shows illustrative DC supplies coupled by switches to provide three (or a multiple thereof) output voltage levels corresponding to output currents such that relatively constant output power is produced at any given output voltage and output current combination. FIGURE 19 shows illustrative DC supplies coupled by switches in a recursive arrangement to provide four levels of output voltage levels corresponding to output currents such that relatively constant output power is produced at any given output voltage and output current combination.

FIGURE 20 shows illustrative DC supplies coupled by switches in a recursive arrangement to provide an even number of output voltage levels corresponding to output currents such that relatively constant output power is produced at any given output voltage and output current combination.

FIGURE 21A-21B shows illustrative phase relationships between the output current and output voltage.

FIGURE 22 shows an illustrative arrangement to control the phase relationship between output current and output voltages ro provide a desired phase difference.

DETAILED DESCRIPTION OF THE INVENTION

Turning to the figures, FIGURE 1 illustrates a plasma source chamber with two sets of antenna elements. The antenna design includes two orthogonal single- or multi-turn loop elements 105, 110, 115, and 120, arranged about a common axis. The antenna elements 105, 110, 115, and 120 are each driven by RF power sources, A 125 or B 130 as shown. Each antenna loop may be coupled to the same RF power source with a phase splitter, or to distinct RF power sources, to drive the antenna elements in quadrature. Preferably the loops in the antenna are constructed from eight (8) gauge teflon coated wire although bare copper wire or other conductors may also be used.

FIGURE 1 shows two orthogonal sets of two-element Helmholtz-coil- like loop antennas, with loop elements 105 and 115 in one set and loop elements 110 and 120 in the second set. The loop elements are wrapped azimuthally around an insulating cylinder 135 such that the magnetic fields that are produced when a current is passed through them are approximately transverse to the axis of the cylinder. The opposing elements of each set are connected in series, in a Helmholtz configuration. The wires interconnecting opposing loop elements are preferably arranged such that adjacent segments carry currents flowing in opposite directions in order to enhance cancellation of stray fields associated with them, although this is not necessary to the device operation. The antennas are energized such that the currents in both orthogonal branches are nearly equal and phased 90 degrees apart to produce an approximation to a rotating transverse magnetic field. In the example case of a helicon mode plasma, a static axial B₀-field 140 is produced, for instance, by a simple electromagnet. This field runs along the axis of the cylinder. The direction of this static field is such that the rotating transverse field mimics that of the m = +1 helicon wave. In practice, the amplitude and direction of the current producing the external field may be adjusted to modulate the performance of the plasma generator. The overall amplitude of the necessary field is typically in the range 10-100 Gauss for the parameters discussed here, but for different size sources alternative ranges may be employed. Once the static field optimum amplitude and direction are chosen, they typically need no further adjustment.

In combination, the static field and the RF field of the antenna elements produce the m = +1 helicon mode in the plasma inside the insulating cylinder, which sustains the plasma discharge. It should be noted that it is also possible to vary, and thus de-tune the static magnetic field, or to not apply the field at all, so that the helicon mode is not directly excited. This operation produces a plasma as well, but typically not as efficiently as the helicon mode. Of course, the static field may then be applied to improve the operation of the plasma source/generator.

It should also be noted that it is possible to achieve the same overall conditions of FIGURE 1 using for instance multi-turn loop antennas instead of single loop, and/or a squat bell jar. Although not a requirement, it is preferable for the Bell jar to fit within the antenna frame with no more than a 1/2" gap.

One example plasma source setup is as follows: A quartz bell jar has approximately 12" inside diameter (such as a standard K. J. Lesker 12 x 12), consisting of a straight-cylindrical section approximately 15 cm tall with a 6" radius hemispherical top. The jar rests atop a vacuum chamber approximately 12" i.d. x 8" tall (not part of the plasma source). The antennas consist of two sets of opposing, close-packed, approximately rectangular, two- turn continuous loop antenna elements that surround the bell jar, with approximately 1/8" to 1/2" spacing between the antennas and the bell jar at every point. The turns within each element are connected in series, and the two elements within each set are also connected in series, such that their fields are additive. The self-inductance of each set is approximately 10 microHenries in this example, and the mutual inductance between the two sets is less than 1 microHenry. Vertical and horizontal antenna loop sections approximately 25cm and 20 cm long, respectively, consist of 8-guage Teflon- coated wire. In alternative embodiments single turns of rigid copper conductors may be employed in place of one or two turns of Teflon-coated wire. The particular embodiments described herein for producing a transverse rotating field should not be interpreted to limit the scope of the claimed invention in the absence of express indications.

A conventional RF power source and dynamic matching scheme, see FIGURES 2 to 4, may be used to excite the antenna currents in the antenna described above. Moreover, the circuits of FIGURES 2 to 4 are compatible with many of the disclosed methods. Some of these methods include steps such as providing a low output impedance to an RF power source; and adjusting a reactance coupling the RF power source to the antenna such that the resonance frequency in the absence of a plasma is the desired RF frequency. A low output impedance can be understood by reference to the quality factor ("Q") for the circuit with and without the plasma. The "Q" with no plasma present should be five to ten-fold or even higher than in the presence of the plasma. Notably, unlike known circuits, such a combination of the RF power source and antenna will not need to be readjusted in the presence of plasma by changing the reactance in response to changes in the plasma impedance.

In FIGURE 2 the RF source 200 may be a commercial 2 MHz, 0-1 kW generator, connected to the quadrature/hybrid circuit at port "A" 125 illustrated in FIGURE 1 via 50 ohm coax. The "+45 degree" and "- 45 degree" legs of the quadrature/hybrid circuit are connected to individual L-type capacitative matching networks composed of adjustable capacitors 205, 210, 215, and 220 as shown. The reactance of capacitors 225 is about 100 ohms each at the operating frequency, and the reactance of either side of the transformer 230 is about 100 ohms with the other side open. As shown in FIGURE 2, a single RF source 200 may be used, together with a passive power splitter (the quadrature/hybrid circuit) and four adjustable tuning elements 205, 210, 215, and 220 to match to the two separate antenna inductances 235 and 240.

Another embodiment, illustrated in FIGURE 3, employs two separate RF power sources 305 and 310, and thus entirely separates the two antenna power circuits connected to inductances 335 and 340 via tunable capacitors 315, 320, 325, and 330 respectively. Such an arrangement is advantageous in that each RF source can be operated at full power, thus doubling the amount of input power as compared to that of a single RF source, and the phasing and amplitude ratio may be adjusted between the antennas. Typically, sources 305 and 310 are operated at roughly the same amplitude and at 90 degrees out of phase, although the amplitude and/or phase difference might be varied in order to change the nature of the excited mode. For example, by operating them at different amplitudes, an elliptically polarized plasma helicon mode rather than a strictly circularly polarized mode could be sustained.

A third embodiment, illustrated in FIGURE 4, places a passive resonant circuit, comprising inductor/antenna inductance 405 and adjustable capacitor 410 on one leg, and drives the other leg with an RF source 400 with a dynamic matching circuit having tunable capacitors 415 and 420 connected to antenna inductance 425. This arrangement tends to excite the same sort of elliptical helicon mode in the plasma, with the passive side operating approximately 90 degrees out of phase with the driven side, thus providing many advantages but with only a single RF source and dynamic matching network.

The working gas in this example setup is Argon, with pressure ranging from 10 mTorr to over 100 mTorr. A static axial field is manually settable to 0 - 150G and is produced by a coil situated outside the bell jar/antenna assembly, with a radius of about 9".

Plasma operation at a pressure of approximately 75 mTorr exhibits at least three distinct modes. First, a bright mode in which the plasma is concentrated near the edge of the bell jar is observed for B₀ < Bcπti_cai when PRF is less than or approximately 200W. Here, B₀ is the axial magnetic field while Bcriticai is a critical value for the axial field for exciting a plasma using a helicon mode. Similarly, power levels PRF and Pt_hres_hoi_d denote the RF power supplied to the antenna and a threshold power described below. In this mode, the RF antenna currents tend to not be in quadrature, instead being as much as 180 degrees out of phase. Second, a dull-glow-discharge-like mode, with uniform density/glow at higher power but with approximately 1-2cm thick dark space along the wall of the bell jar at lower powers, is observed for B₀ > B_criticai but PRF < Pthres_hoid- In this case the RF currents are in robust quadrature, appearing to abruptly lock at approximately 90 degrees phase shift shortly after the plasma is formed. Third, at higher PRF > Pthreshoid and with B₀ > B_cri_ti_cai, a bright plasma is formed that appears to be more evenly radially distributed than that of mode (1), and the antenna currents again tend to lock into quadrature phasing. The third regime represents an efficient mode of operation, and can be achieved at a neutral gas pressure that has proven to be very difficult to access for known plasma sources, although each of these regimes may have application in plasma processing.

In an aspect, the conventional RF power source and tunable matching network described in FIGURES 2 to 4may be eliminated in favor of a streamlined power circuit.

In a preferred embodiment, an RF power circuit drives the antenna current strap directly, using an arrangement such as that shown in FIGURE 5. The RF amplifier illustrated in FIGURE 5 is preferably one of the many types of RF amplifiers having a low output impedance (i.e. a push-pull output stage) that are known in the field. Transistors 505 and 510 are driven in a push-pull arrangement by appropriate circuitry 500, as is known to one of ordinary skill in the art. In this arrangement typically one transistor is conducting at any time, typically with a duty cycle of or less than 50%. The output of the the transistors is combined to generate the complete signal. Preferably, the power semiconductors, e.g., transistors 505 and 510, in the output stage are operated in switching mode. In the FIGURES 5-7 these are depicted as FETs, but they can also be, for example, bipolar transistors, IGBTs, vacuum tubes, or any other suitable amplifying device. An example of switching mode operations is provided by Class D amplifier operation. In this mode alternate output devices are rapidly switched on and off on opposite half-cycles of the RF waveform. Ideally since the output devices are either completely ON with zero voltage drop, or completely OFF with no current flow there should be no power dissipation. Consequently class D operation is ideally capable of 100% efficiency. However, this estimate assumes zero ON- impedance switches with infinitely fast switching times. Actual implementations typically exhibit efficiencies approaching 90%.

Preferably, the RF driver is coupled directly to the antenna current strap 520 through a fixed or variable reactance 515, preferably a capacitor. This coupling reactance value is preferably such that the resonant frequency of the circuit with the coupling reactance and the antenna, with no plasma present, is approximately equal to the RF operating frequency.

An alternative arrangement of the output stage of this circuit, illustrated in FIGURE 6 (A), includes a transformer 620 following or incorporated into the push-pull stage, with driver 600 and transistors 605 and 610, to provide electrical isolation. Transformer 620 may optionally be configured to transform the output impedance of the push-pull stage, if too high, to a low impedance. Capacitor 615 is arranged to be in resonance at the desired drive frequency with the inductive circuit formed by transformer 620 and antenna current strap 625. A similar embodiment is shown in FIGURE 6(B), where capacitor 615 is used for DC elimination, and capacitor 630 is resonant in the series circuit formed by leakage inductance of transformer 620 and inductance of the current strap 625.

FIGURE 7 illustrates yet another RF power and antenna current strap configuration. A center-tapped inductor 725 incorporated in the DC power feed is connected to the output stage having push-pull driver 700 and transistors 705 and 710. Isolation is provided by transformer 720. Again, only one or the other transistor is conducting at any time, typically with a duty cycle of less than 50%. The circuits of FIGURES 5-7 are provided as illustrative examples only. Any well-known push-pull stage or other configurations providing a low output impedance may be used in their place.

The RF power source may also be used with any helicon antenna, such as either a symmetric (Nagoya Type III or variation thereof, e.g., Boswell-type paddle-shaped antenna) or asymmetric (e.g., right-hand helical, twisted-Nagoya-lll antenna) antenna configuration, or any other non-helicon inductively coupled configuration.

The RF power source may be amplitude modulated with a variable duty cycle to provide times of reduced or zero plasma density interspersed with times of higher plasma density. This modulation of the plasma density can be used to affect the flow dynamics and uniformity of the working gas, and consequently the uniformity of the process. A more spatially uniform distribution comprising plasma may therefore be generated by a plasma generator system by a suitable choice of a modulation scheme. In general, a plasma generator system may use radio frequency power sources based on operation as a substantially Class A amplifier, a substantially Class AB amplifier, a substantially Class B amplifier, a substantially Class C amplifier, a substantially Class D amplifier, a substantially Class E amplifier, or a substantially Class F amplifier or any sub- combination thereof. Such power sources in further combination with the antennas for exciting helicon mode are suitable for generating high density plasmas. Moreover, for non-switching amplifiers, such as those shown in FIGURES 2-4, an intermediate stage to transform the RF source impedance to a low output impedance may be employed to approximate the efficient operation of the switching amplifier based embodiments described herein.

In inductively coupled plasma sources, the antenna current strap is located in proximity to the region where plasma is formed, usually outside of an insulating vessel. From a circuit point of view, the antenna element forms the primary of a non-ideal transformer, with the plasma being the secondary. An equivalent circuit is shown in FIGURE 8, in which inductor 810 represents a lumped-element representation of the current strap and any inductance in the wiring, including any inductance added by e.g., the driver's output transformer present in some embodiments. Components in the box labeled P represent the plasma: inductor 820 is the plasma self inductance, and impedance 815 represents the plasma dissipation, modeled as an effective resistance. M represents the mutual inductance between the antenna and plasma. Transistor driver 800 is represented as a square-wave voltage source. The capacitance 805 is adjusted at the time the system is installed to make the resonant frequency of the circuit approximately match the desired operating frequency. In an alternate embodiment with a fixed capacitor, the RF frequency may be adjusted to achieve the same effect.

For illustrating the operation of the system, the overall system may be modeled as shown in FIGURE 9. In FIGURE 9 all inductors have been lumped into inductance 905, all capacitors into capacitance 910, and all dissipating elements into resistor 915, and the amplifier should ideally operate as an RF voltage source (i.e., having zero output impedance).

With no plasma present, R is small since there is little dissipation, and the circuit of FIGURE 9 exhibits a narrow resonant response to changes in frequency, as shown in FIGURE 10. This provides one of the advantages of the circuit's operation: it is possible to drive the voltage on the antenna to a high value with relatively little power input, thus facilitating the initial breakdown of the gas in the reaction chamber. Once the plasma forms, the damping in the system considerably broadens the resonant peak, as shown in FIGURE 11 , reducing the Q of the overall circuit. Although the center frequency of the resonance may shift with plasma conditions, that shift is negligible compared to the width of the resonant response when the plasma load is present. Therefore, when operating with a plasma load the circuit is relatively insensitive to variations in operating conditions, and requires no retuning. This is illustrated in FIGURE 11, where the overall system resonance has shifted its frequency slightly, although the Q is sufficiently reduced that the operation of the system remains efficient. With the reduced Q of the circuit, the voltage applied to the plasma self-adjusts to be considerably reduced over the no-plasma case. In some embodiments, it may be somewhat advantageous to actually detune the operating frequency of the RF drive slightly from the exact no-plasma resonance to one side or the other, depending on the shift of the resonant frequency when the plasma forms.

The level of power input to the plasma may be controlled by a variety of techniques, such as adjusting the DC supply level on the RF output stage. In one embodiment, the supply voltage may be in response to sensed variations in plasma loading to maintain a relatively constant power into the plasma source. As illustrated in FIGURE 12, the sensing of plasma loading for adjustments by DC supply regulator 1230 may be achieved, for example, by monitoring the voltage from the DC supply 1215 by voltage sensor 1200 and the DC current into the RF/Plasma system by current sensor 1205, and using their product together with a previously measured approximation to the amplifier efficiency in module 1210 to estimate the net power into the plasma 1225 from RF Amplifier 1220. Efficiency multiplier for gain module 1235 can be measured for different output levels, for instance by monitoring heat loads at various points of the system, and stored digitally, so that variations in efficiency with output level are accounted for. Alternatively, the RF voltage and current can be measured, and their in-phase product evaluated to estimate the real power being dissipated in the plasma.

The sensing of plasma may also extend to sensing spatial uniformity by either direct sensing or indirect sensing by way of variations in the voltage or current. Changing the duty cycle in response to such variations can then control the spatial distribution of plasma. In addition, modulating the duty cycle can further allow control over the average input power to improve the efficiency of plasma generation. The feedback arrangement of FIGURE 12 can also allow switching between two or more power levels as described previously.

"Low" impedance, as used herein, means that the series resonant circuit shown in FIGURE 9 has a "Q" that should be five to ten-fold or even higher with no plasma present than with plasma present. That is, the amplifier output impedance should be sufficiently small that the energy dissipated in a half-cycle of output is much less than that stored in the reactive components.

This condition is mathematically defined as Z_out « J— , where L and C are

V ^ the lumped values shown in FIGURE 9. The RF amplifier will approach operation as a voltage source when this condition holds.

A low resistance, e.g., for the output impedance of the RF source, generally refers to a resistance of less than about 10 ohm, preferably less than about 6 Ohms, more preferably less than about 4 Ohms, and most preferably less than about 1 Ohm.

However, not all embodiments of the invention require that the elements in the reactive circuit coupling the RF power source to the antenna/plasma be selected based on the resonant frequency of the circuit without a plasma being present. Indeed several, alternative conditions are possible that allow a suitable specification of the reactive circuit such that there is no need for a dynamic matching circuit while efficient coupling is possible with the dynamic impedance of a plasma.

While presenting a variable impedance, it is possible to describe the plasma reactance as being expected to be confined between a high and a low limit. Thus, a high expected plasma reactance component and a low expected plasma reactance may be specified. For instance, such a specification may reflect a one-σ distance away from the expected mean value. Many other similar specifications are possible to indicate the likelihood of the plasma impedance actually falling outside the specified limits. Indeed, instead of a high expected plasma reactance, it is possible to specify a value that is not symmetrically placed relative to the low expected plasma reactance. Moreover, while a particular plasma impedance may fail to conform to a normal distribution, a collection of several plasmas is likely to collectively present a normal distribution for the combined impedance.

Similarly, a collection of several RF power sources connected together is likely to exhibit a normal distribution, both with respect to frequency and time. Then a suitable choice of a reactance network may actually ensure that the variation in plasma reactance is well matched to the variation in the RF power sources by matching them at two values of the expected plasma reactance.

With such a specification of the plasma reactance and a knowledge or estimate of the lowest or likely low plasma resistance, a value at which the variation of the plasma impedance is likely to be the greatest, it is possible to arrive at a method of specifying the components in the reactive circuit.

For example, in the illustrative circuit of FIGURE 13, from the publication "3kW and 5 kW Half-bridge Class-D RF Generators at 13.56MHz with 89% Efficiency and and Limited Frequency Agility", Directed Energy Inc. © 2002, document number 9300-0008 Rev. 1 ," retrieved on June 10, 2004 from the web address http://www.ixysrf.com/pdf/switch_mode/appnotes/3ap_3_5kw13_56mhz_gen. pdf, which is incorporated herein by reference, with the specification that the impedance at the RF jack is 50 Ohms, C_a=C₁H-C₂, and Cb=C₃ ⁺C₄, the series

impedance is Zi= +sL_x = — - , and shunt impedance sC_a sC_a

. The impedance seen at the input is Zi+Z₂. With L1

given as 2.1 μH, this complex value may be adjusted with suitable components to be 14 +/12.6 Ohms by adjusting C₃ to be about 81.6 pF and C_b at about 376 pF.

In a capacitively coupled system, e.g. for use at 13.56 MHz to provide an RF bias for a substrate in a semiconductor processing chamber, an illustrative plasma antenna combination may, for example, present a resistance R_p of about 1 to 4 Ohms and a reactance X_p of about -8 to -25 Ohms. Thus, hooking the circuit of Figure 13 to such an antenna/plasma combination is difficult in general. With the large imaginary component of the impedance that it sees, the transistor switching circuit will safely operate with a supply voltage that is a fraction of the desired peak supply voltage of about 700 to 800 V, e.g., at about 250 V (more likely 200). The peak output voltage is given by V_suppiy/2 X |H|, where |H| is the magnitude of the transfer function of the system, and in the 250 V case will range from about 28 to 83 V for the various plasma conditions.

when operating at a given frequency, a total impedance may be adjusted by adding an inductor (having a positive reactance) or a capacitor (having a negative reactance) in series with the impedance. As an example, if there is a stray inductance L due e.g. to leads and the like, the total impedance may be adjusted to a level at or near zero for a given operating frequency by adding a capacitor in series, with the capacitance adjusted so

that Z_to_t - ZL+ZC =iωL « 0. Similarly, in driver circuits using output iωC devices with significant output capacitance, such as transistors or mosfets, dissipation due to the output capacitance (e.g. C_oss on some specification sheets) may be reduced by providing a slightly inductive load. This is because of the charge stored in the capacitance: a properly tuned inductive load discharges the capacitance without having to dissipate this charge.

FIGURE 14 shows an illustrative general reactive circuit 1400 suitable for coupling radiofrequency power source 1405 to a capacitively driven plasma or an antenna-plasma combination. Although, this circuit relates to a capacitively coupled driver, e.g., for the RF biasing of a substrate in a semiconductor processing plasma, but the principle for determining the values of the components applies to an inductively coupled system as well. The illustrative general reactive circuit 1400 may be tuned either using the capacitors or inductors or both. For instance, the reactance of capacitors 1415 and 1425 may be chosen to be approximately the same as the minimum plasma reactive component, at about 500 pF each. Inductors 1410 and 1420 are then tuned to satisfy two conditions: a) at the largest magnitude of plasma reactance, i.e., a high expected plasma reactance limit, the imaginary part of the overall load seen by the transistor output stage is small, and b) at the smallest magnitude of plasma reactance, i.e., a low expected plasma reactance limit, the imaginary part of the load seen by the output stage is adjusted to optimize operation of the radio frequency power source, e.g., +12 Ohms as in the circuit described in the above Directed Energy reference.

The impedance seen by a transistor driver stage is given by Zi_oad = Zi4io + Z-1415 + (Z-I420 + Zi4₂5)||Z_p. Here Z1410 represents the impedance of inductor 1410 in FIGURE 14 and the like while Z_p represents one value of expected plasma reactance. That is, the driver sees capacitor 1410 in series with inductor 1415 and in series with the parallel combination of plasma impedance 1440 and capacitor 1425 + inductor 1420 series combination.

For a radio frequency power source, which operates best when it drives load with a with an output reactance of +12 Ohms, case "a" corresponds to lm(Z|_Oad) of about 0 Ohms at a plasma reactance, X_p, of about -25 Ohms. Case "b," then corresponds to lm(Z|_Oad) being about 12 Ohms at Xp of about -8 Ohms. These conditions result in a pair of equations that may be solved with R_p set at a low value, say about 1 Ohm, since this level of plasma resistance results in large variations in the load seen by the RF driver. These two equations can be solved for the unknown value of inductances 1415 and 1420. Under the described conditions, in this exemplary embodiment, values of inductance 1420 is about 345 nH and inductance 1415 is about 185 nH resulting in lm(Z|_Oad) of about 0 Ohms for condition "a" and lm(Z|_Oad) of about 11.9 Ohms for condition "b"' respectively. More sophisticated calculations preferably take into account stray inductances, coil inductances and the like along with other non-ideal effects.

Alternative choices may be elected for the value for capacitors 1410 and 1425, e.g., by electing smaller values to improve the tolerance if subtracting two comparable numbers is resulting in large errors. Additionally, instead of fixing the values for capacitors 1410 and 1425, and adjusting the values for inductors 1415 and 1420, it is also possible to fix the value for inductors 1415 and 1420 and adjust the value of capacitors 1410 and 1425. It will be recognized further that the total impedance is the important quantity for any series or parallel combination of reactive components, and that specific values of L and C or specific geometries can be used in the above circuit. As an example, a series combination of an inductor L and a capacitor C can have a reactance of about 5.9 ohms when L=345 nH and C=500 pF, or when L= 620.5 nH and C=250 pF. These values can be adjusted to satisfy other constraints in the system, such as the need to have a high (or low) impedance at a 2^nd harmonic.

Alternative output transistor stages may be operated at different impedances in the reactive load, including a slightly capacitive load. Then, the condition lm(Z|_Oad) is about 0 Ohms may be specified at some midpoint value rather than for the low or high expected plasma reactance limit. Thus, at this specified magnitude of plasma reactance, i.e., a specified plasma reactance limit, the imaginary part of the overall load seen by the transistor output stage is small. Further, the specified plasma reactance may be a value outside the range of expected operation. However, such a specification may result in higher output current. In addition, adding a resistive path in parallel with capacitor 1425 improves the performance of the reactive circuit. Thus, the reactive circuit may include resistive elements as well.

In another aspect, nonlinear resistive or reactive elements may be used for the purpose of reducing the impedance variation seen by the RF power source. In yet another aspect, the inductors 1415 and 1420 may be arranged to have a small amount of mutual inductance, which can be either positive or negative. A positive mutual inductance

β-9-, in the range

M ^■ , 1415,1420

^■< 0.02, may be used to reduce the sensitivity of the response

V 1415 A ^J14_' 20

transfer function H to changes in plasma reactance, while negative mutual inductance may increase the sensitivity.

These methods for tuning or setting up of a reactive network provide several advantages in addition to removing the need for a dynamically tuned matching circuit. For example, since the tuning at one plasma reactance in the range of reactance values expected for a plasma matches that for the operation of amplifier, it provides the transistors with the reactive impedance needed for efficiently operating at a high voltage. Further, although at the other end of the plasma range, the reactance seen by the output stage is small, the total load is also small, enabling operation at high current and low supply voltage resulting in the reactance presented to the transistors being less important. Moreover, this specification ensures that over a broad range of plasma reactance, a reasonable amount of power may be delivered from the RF source to the plasma. In another aspect, with this design enables use of a large number of output stages that may be combined, for instance, in parallel.

Often the specification for a RF power supply is an output voltage for application to the antenna terminals, with the RF input voltage level being adjusted to produce this desired output voltage according to what is necessary for varying plasma operating conditions. Examination of the transfer function H = Vpi_asma / V_in reveals that the system "voltage transfer function," or the ratio of output voltage to input voltage, H = V_pι_asma / V_1n = [

(Zi4io + Zi4i5)||Z_p ] / Zioad-

For the tuning as described, this transfer function has a resonant character, in that the magnitude of H is greater than one over a substantial, if not the entire, range of operation. |H| varies from approximately 75 at X_p of about -25 Ohms (case "a" above, with R_p of about 1 Ohm) down to approximately 1.5 at X_p = -8 Ohms (case "b" above, with R_p still at about 1 Ohm). For the higher plasma resistance, for instance, R_p of about 4 Ohms, |H| varies from approximately 21 to 1.6. Therefore, selecting a reactance network well suited for operation at the lowest expected plasma resistance ensures with high degree of certainty that the variation in plasma impedance would be smaller at a higher values of the plasma resistance.

It should be noted that although some of the discussion is in terms of the resonant frequencies for the reactive network, it is often desirable to drive the radio frequency power source at a frequency that deviates somewhat from the resonant frequency in the absence of a plasma in the direction of the frequency spread due to the presence of the plasma. This ensures stable and efficient operation over frequencies of interest.

The disclosed system and methods provide an advantage in being able to break down this gas and initiate the plasma by virtue of the fact that the high Q of the circuit with no plasma allows high voltages to be induced on the antenna element with relatively low power requirements. This no-plasma voltage can be controlled to give a programmed breakdown of the working gas; once the plasma forms, induced currents in the plasma serve to load the system and lower the high voltages for inducing the breakdown, and thus, avoid stressing the system.

The described circuit arrangements do not require a variable tuning element, such as a mechanically adjustable capacitor, since only fixed capacitance C is necessary. However, the various circuits can also be constructed using a variable capacitor that is adjusted, for example, for matching of the system resonance to the desired operating frequency, in a preferred embodiment, and is not needed for real-time impedance matching with the plasma operating point. Such matching is useful to counter the effects of mechanical vibration or aging that may cause the L-C resonant frequency to drift.

In one embodiment, the operating frequency is adjusted to compensate for small deviations from resonance, while mechanically tuning the capacitor compensates for large deviations. In an alternative embodiment, adjustments are made by tuning the capacitor. In the preferred (tuned) embodiment, this tuning is automated and takes place during periods when the source is offline. In another aspect, with tuning as part of the process control, for instance to provide small tweaks to the process conditions, the disclosed arrangement reduces the number of adjustable elements to as few as one in embodiments with adjustable tuning elements.

Example embodiments that combine several RF sources, at least some of which are designed in accordance with this disclosure to reduce the complexity in a matching network, if any, cover many geometries. These geometries include, without limiting the scope of the claimed invention, simply distributed, concentric, offset, and combination geometries. In FIGURES 15- 16, each depicted circle 1500 represents one or more turns of conductor, stacked vertically, or spaced radially in a flat spiral style, or simply bundled in some fashion. Although not all turns are labeled, a representative sample are illustratively identified with the numeral 1500 in FIGURES 15A-15D. FIGURE 16 depicts a non-planar surface with coils for generating a plasma in accordance with this disclosure.

A simply distributed geometry has several non-overlapping sources arranged in a geometry of choice. A concentric geometry has two or more concentric antenna elements with their axis substantially parallel and close together. The outer boundaries of the plasma sources may or may not be uniformly spaced apart at their respective perimeters. An offset geometry has two or more antenna elements with their respective axis substantially arranged to be parallel to each other, but displaced relative to each other. In a combined geometry, one or more of the above geometries are present. Preferably, the sources are cylindrical with circular cross-sections, but this is not intended to be a limitation on the scope of the invention. Further, the source combinations need not be planar. For instance, one or more direct drive RF generators could be mounted on or around a hemispherical dome.

When combining multiple sources, the current and voltage demands may require the use of a power supply that is rated for significantly more power than what is needed. When driving the RF generator system a specified amount DC power is provided to drive a widely varying load. In a given plasma process, the load may be optimally driven by 50 Amperes at an output voltage of 20 Volts, while in a second process the load may need 2.5 Amperes at an output voltage of 400 Volts. In either case the DC power output is 1 kilowatt. Providing power to both processes with a single conventional supply (at different times) requires a variable output 400 Volt supply that is capable of sourcing 50 Amperes, or nominally a 20 kilowatt supply.

A technique for avoiding the necessity of using a grossly over-rated power supply for such a widely varying load line is disclosed, which makes the design of plasma generators increasingly robust, economical and versatile. FIGURE 17 shows a suitable design for a reconfigurable multiple separate power supplies for providing either low voltage at high current, or higher voltage at lower current. In a preferred embodiment, the individual DC supplies are switching type power supplies, preferably controlled via pulse width modulation.

Each programmable DC supply 1705 is capable of putting out a variable DC voltage with a maximum voltage V₀ volts, and capable of supplying a maximum current I₀ amps. The switch represents a semiconductor switch such as an IGBT or MOSFET. If the switch is open, the two power supplies are connected through the diodes in parallel. As long as they are matched, according to known techniques for paralleling power supplies they will share the current, and the output from this circuit will be Vo, ignoring the diode voltage drop. See e.g., the article by Bob Mamano and Mark Jordan titled "Load Sharing with Paralleled Power Supplies" dated September, 1991.

A preferred a 2-way switched supply is schematically illustrated in FIGURE 17. The maximum current this arrangement can source is 2I₀. If the switch is closed, the two diodes that were formerly conducting become reverse biased, and the power supplies are forced to be connected in series. In this configuration, the voltage is 2V₀, and the maximum current is I₀. In either case, the maximum power output is the sum of the individual maximums, P = 2V₀I₀. By varying the output voltages Vo from the individual supplies it is possible to smoothly operate from 0 volts output to the maximum.

In a preferred embodiment, the individual power supplies are each connected to separate secondary windings of a single transformer in a switching supply configuration. The voltage in each secondary winding is rectified by the corresponding power supply. The primary winding is drive by known techniques for variable output switching supplies that are not reconfigurable. This arrangement, then, allows instead of a single output multiple largely similar outputs.

In a preferred embodiment, the default configuration is for the switch to be turned on. The desired program scheme is to source DC voltage and current at a constant power level. In this case, as the load impedance drops, more output current is drawn to maintain constant output power, e.g. l²RI_Oa_<_ is held constant, while V=IR is dropping. At some threshold of decreasing load impedance, the supplies in series are unable to provide sufficient current, below this threshold the switch is opened and the programmed output voltage is simultaneously doubled, so that the reconfigured supply can now supply higher current at the same voltage. If the load impedance subsequently rises, so that the constant power requirement results in higher voltage and less current, the switch is closed at the same threshold to adapt to the varying load. Another preferred embodiment has different thresholds for rising and falling load impedance, to provide hysteretic overlap in the controllable ranges.

In a preferred embodiment, the above system is built up recursively, wherein each of the power supplies shown as circles above in turn consists of a reconfigurable supply of the type described in this disclosure. In this way, a binary progression of multipliers on the output voltage is achieved. An implementation with one level of recursion has possible outputs of IxV₀, 2Λ/₀, 4^χV₀ at maximum currents of 4^χ|₀, 2*l₀, 1 *l₀, respectively. This system is illustrated in figure NNN. For IxV₀, no FETs are turned on; for 2^χV₀, Q2 and Q3 are turned on, and for 4^χV₀, all FETs are turned on. In this implementation the intermediate switch consisting of Q2+Q3 must carry twice the current of either Q1 or Q4, and is shown as 2 FETs of similar type; a similar requirement exists for the diodes. In another implementation Q2+Q3 are replaced by a single FET capable of carrying twice the current, and D1+D2 pair and D5+D6 pair are replaced by two single diodes capable of carrying twice the current. In a preferred embodiment, the switches are IGBTs. In a preferred embodiment, the switches are bipolar transistors, or any other semiconductor switch that can be turned on and off in some way.

Another embodiment, illustrated in FIGURE 18, has two levels of recursion, producing outputs at 1^χV₀, 2^χV₀, 4^χV₀, 8^χV₀₎ etc. at max currents of 8^χl₀, 4^χ|₀, 2^χlo, 1 *lo, respectively. In another embodiment, these individual elements can be recursively combined with a 2-way reconfigurabie arrangement described above to provide output voltages at 1 ^χV₀₎ 2^χV₀, 3*V_Ol 6^χV₀ at respective currents of 6^χ|₀, 3^χ|₀, 2xl₀, 1*l₀ as shown in FIGURE 19. A 4-way implementation of this is shown in FIGURE 20. In this implementation, each independent power supply comprises a transformer secondary, two diodes, two inductors and a capacitor connected in a known current double configuration. The primary is driven by a pulse-width modulation scheme. An output stage, (e.g. S1 , D3, D4, L1 , L5, C2) produces an output of V₀, at a maximum current I₀. In one implementation these transformer secondaries (S1-S4) are all wound on a single core.

S1-S4 are transformer secondary windings that have AC voltage on them, e.g. in the range 20 - 500 kHz. The diodes and inductors connected directly to each winding (e.g. D3, D4, L1 , L5) form a known "current doubler" output stage. The MOSFETS and remaining diodes (e.g. Q9, D9, D21) perform the functions described above. With none of the switches turned on this supply produces an output voltage of V₀, at a maximum current 4I₀. Turning on Q9 and Q13 gives 2V₀, max current 2I₀. Turning on all switches gives voltage 4V₀, at maximum current I₀.

A second implementation uses a full-wave rectified output from the various Si windings. The technique is not limited to multipliers of 2, or 3, but may be extended to any number or combination of multiple supplies, in any recursive combination to produce a wide range of output voltages from a given supply voltage.

In a preferred embodiment, the number of individual power supplies is not a power of two. A preferred embodiment comprising three power supplies (and two switches that are simultaneously actuated) each capable of supplying V₀ volts and supplying Io amps can provide outputs at 1 ^χV₀ and 3*V₀ at respectively 3^χl₀, 1*l₀ amps. This is shown in Figure 19. In another aspect, it is desirable to control and be able to specify the phase relationship between the current and voltage output by a power supply. This is a challenge when dealing with the dynamically changing plasma impedance. In order to achieve an easily controlled output power, an embodiment has the output current in a defined phase relationship to the output drive voltage. In other words, the output voltage and current waveforms should look approximately like those shown in FIGURE 21 A

Preferably, a phase locked loop is used to lock the phase of the driver output square wave voltage to the phase of the output in a given relationship. This relationship should maintain a constant phase shift of approximately n degrees, where n may be zero.

In an alternative preferred embodiment, the output load has an inductive component in order reduce switching losses in the transistors, and in this case the driver output current lags the output square wave. An example of these waveforms is illustrated in FIGURE 21 B.

In an alternative preferred embodiment, the phase relationship is dynamically adjustable. A block diagram of one embodiment of such a controller is illustrated in FIGURE 22. The phase detector should have an adjustable offset, such that a given phase relationship will be maintained between its two inputs.

A schematic diagram of a preferred embodiment is shown in FIGURE 22. In this implementation, a phase locked loop, for instance, SN74HC4046 from Texas Instruments, is used, with its "XOR" phase detector. This circuit diagram also illustrates an overcurrent protection scheme, wherein if an output current lever higher than a threshold value set by R18 is detected, the output driver is turned off. Even though the output drive signals for the gates of the FETs are zero, the output current continues to ring at approximately the same frequency it was driven at, flowing through the body diodes of the FETs. When it has decayed (rung down) to a level of about 80% of the overcurrent trip point, the gates begin firing again.

An advantage of the illustrated circuit is that even during the time when the gates are inhibited, the phase relationship is maintained between the gate control signals and the output current. This is achieved by taking the voltage reference input for the phase locked loop from an internal point in the circuit (in this case just before the output inhibit stage composed of U3C and U3D), rather than from the FET output stage. For the type of phase locked loop described herein, using the XOR phase detector, the phase offset between the two signals may be adjusted by changing the center frequency of the V_Co, using R2. The current reference for the phase locked loop is derived from a current transformer and resistor, and must be of the polarity that provides negative feedback for the phase controller setpoint. The current reference for the overcurrent detector is from a separate current transformer that has been full-wave rectified.

As one skilled in the art will appreciate, the disclosed invention is susceptible to many variations and alternative implementations without departing from its teachings or spirit. Such modifications are intended to be within the scope of the claims appended below. For instance, one may provide impedance matching for a low impedance with a transformer in combination with a conventional amplifier. Also, although the invention obviates the need for dynamic matching circuits, the use of some dynamic matching circuit in combination with the reactive circuits disclosed herein to reduce the otherwise stringent requirements placed on dynamic matching networks is intended to be included within the scope of the invention as indicated. Therefore, the claims must be read to cover such modifications and variations and their equivalents. Moreover, all references cited herein are incorporated by reference in their entirety for their disclosure and teachings.

Claims

1. A method for reducing the need for a dynamic matching circuit for directly driving a dynamic plasma impedance, the method comprising the steps of: providing a plurality of radio frequency power sources, each with a low output impedance; providing a reactive network comprising a first and second reactance between the radio frequency power source and the plasma, wherein the first reactance and the second reactance are selected such that at a first plasma reactance, a substantially resistive load is presented to a first radio frequency power source in the plurality of radio frequency power sources and at a second plasma reactance a specified reactance seen by the first radio frequency power source in the plurality of radio frequency power sources; controlling an average input power by modulating a duty cycle for operating the first radio frequency power source.; and wherein the plurality of radio frequency power sources are combined in one or more of simply distributed, concentric, offset, and combined geometries to provide a desired total power level and coverage.

2. The method of claim 1 , wherein values of the first plasma reactance and second plasma reactance span a substantially fraction of a range expected for the dynamic plasma reactance.

3. The method of claim 1 , wherein the first and second plasma reactance values correspond to a high expected plasma reactance limit and a low expected plasma reactance limit respectively.

4. The method of claim 1 further comprising the step of providing direct current power to the plurality of radio frequency power sources by reconfigurable multiple separate power supplies for providing either low voltage at high current, or higher voltage at lower current.

5. The method of claim 4, wherein the reconfigurable multiple separate power supplies provide output voltages at one or more sets from 1 ^χV₀, 2xV₀, 4^χV_0l 8^χV₀, and I xV₀ and 3* V₀, and UV₀, 2*V₀, 3^χ\Z₀, 6^χV₀.

6. The method of claim 4, wherein the reconfigurable multiple separate power supplies are arranged in a recursive configuration to provide adjustable voltages from a given supply voltage.

7. An adjustable plasma generator system comprising: a plurality of radio frequency power means; and at least one reactive circuit for interfacing the plurality of radio frequency power means to a plasma with a dynamic impedance; wherein, at a low expected plasma resistance limit, the at least one reactive circuit presents a small total reactance when a plasma reactance is at a high expected plasma reactance limit and presents a reactance that does not exceed a specified reactance; and wherein one or more of the plurality of radio frequency power means are combined in one or more of simply distributed, concentric, offset, and combined geometries to provide a desired total power level and coverage.

8. The plasma generator system of claim 7, wherein a direct current power to the plurality of radio frequency power means is supplied by reconfigurable multiple separate power supplies for providing either low voltage at high current, or higher voltage at lower current.

9. The system of claim 8, wherein at least one of the plurality of radio frequency power means comprises a push-pull circuit.

10. A radio frequency plasma generator comprising a reactive circuit, coupled to at least one radio frequency power supply for driving a plasma with a dynamic impedance such that the reactive circuit presents a small total reactance when the plasma reactance is at a first plasma reactance and presents a reactance that does not exceed a specified limit at a second plasma reactance.

11. The radio frequency generator of claim 10, wherein the at least one radio frequency power supply is combined with a plurality of radio frequency power supplies in one or more of simply distributed, concentric, offset, and combined geometries to provide a desired total power level and coverage.

12. The reactive circuit of claim 11 , wherein a direct current power to the plurality of radio frequency power supplies is supplied by reconfigurable multiple separate DC power supplies for providing either low voltage at high current, or higher voltage at lower current.