TW202243548A - Frequency tuning for modulated plasma systems - Google Patents

Abstract

Plasma processing and power supply systems and methods are disclosed. The plasma processing system comprises a high-frequency generator configured to deliver power to a plasma chamber and a low-frequency generator configured to deliver power to the plasma chamber. A filter is coupled between the plasma chamber and the high-frequency generator, and the filter suppresses mixing products of high frequencies produced by the high-frequency generator and low frequencies produced by the low-frequency generator. The plasma processing system also comprises means for frequency tuning the high-frequency generator using a probe signal that is concurrently applied with the power applied to the plasma chamber at the primary frequency.

Description

Frequency Tuning for Modulated Plasma Systems

Embodiments disclosed herein relate generally to plasma processing systems, and more specifically, to plasma processing systems having modulated plasmas.

The corresponding priority of this patent application is the continuation-in-part of patent application No. 16/934,257, entitled "Apparatus and System for Modulated Plasma Systems," filed on July 21, 2021. This continuation-in-part Continuation-in-Part of Patent Application No. 16/230,923, entitled "Plasma Delivery System for Modulated Plasma Systems," filed on December 21, 2018 and published as U.S. Patent No. 10,720,305 on July 21, 2020 , and all of the above applications are assigned to the assignee of the present invention and are hereby expressly incorporated herein by reference.

Although plasma processing systems for etching and deposition have been used for decades, upgrades in processing technology and equipment technology continue to produce more and more complex systems. These increasingly complex systems create more difficult interactions between the multiple generators driving the same plasma system.

An aspect may be characterized as a power generation system comprising: a high frequency generator configured to apply power to a plasma chamber at a primary frequency; and a filter configured to suppress The products are mixed to limit the variation presented to a time-varying load reflection coefficient of the high frequency generator. The power generation system also includes a frequency tuning subsystem configured to: apply a probe signal comprising one or more probe frequencies while the high frequency generator is applying power at the primary frequency; and respond to an indication The primary frequency of the high frequency generator is adjusted to improve the one or more detection frequencies measured for performance.

Another aspect may be characterized as a method for automatic frequency tuning of a power generation system comprising applying, by a high frequency generator, a primary power signal to a plasma load at a primary frequency, and at one or A detection signal is applied to the plasma load at a plurality of detection frequencies. Mixing products are suppressed by a filter to reduce variations in a time-varying load reflection coefficient present to the high frequency generator, and the primary frequency is adjusted based on a measure of performance in response to the probe signal.

Yet another aspect can be characterized as a plasma processing system comprising: a plasma chamber; a high frequency generator configured to apply power to a plasma chamber at a primary frequency; and a a low frequency generator for applying power to the plasma chamber at a low frequency. A filter in the system is configured to suppress the mixing product of the dominant frequency and the low frequency to limit changes in a time-varying load reflection coefficient presented to the high frequency generator. And the system includes means for frequency tuning the high frequency generator using a probe signal applied simultaneously with the power applied to the plasma chamber at the primary frequency.

The interaction between generators driving the same plasma, where one of the generators modulates the load seen by the other, becomes increasingly intractable as power levels increase; therefore, the need for New and improved methods and systems to address this problem.

The word "exemplary" is used herein to mean "serving as an example, instance, or illustration." Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.

Referring to FIG. 1 , there is shown a block diagram depicting an exemplary environment in which embodiments may be implemented. As shown, the plasma load of plasma chamber 100 is coupled to high frequency generator 102 via filter 104 and matching network 106 (also referred to as matching 106 ). In addition, the low frequency generator 108 is also coupled to the plasma load via the matching 110 . Matching 106 may be combined with matching 110 in many applications. Optional broadband measurement components 114, 116, 118 and 120 and optional delay element 112 are also shown. The optional delay element 112 may use a length of coaxial cable or a fixed or variable RLCM (i.e., circuit containing resistors, inductors, capacitors, and coupled inductors) circuit or contain discrete circuit elements (i.e., , transmission line circuit) circuit to achieve. Also shown are optional connections 122 and 124 which allow one of the optional broadband measurement systems 116, 120 to interface with the other provided that the optional delay element 112 is properly characterized. Feature.

Although high frequency generator 102 and low frequency generator 108 may each operate within a range of frequencies, in general, high frequency generator 102 operates at a higher frequency than low frequency generator 108 . In many embodiments, the high frequency generator 102 may be a generator that delivers RF power to the plasma load in the plasma chamber 100 in the frequency range of 10 MHz to 200 MHz, and the low frequency generator 108 may be, for example, at 100 kHz to 2 MHz range. Accordingly, an exemplary frequency ratio of the frequency of the low frequency generator 108 to the frequency of the high frequency generator 102 is between 0.0005 and 0.2. In many embodiments, for example, the frequency ratio of the frequency of the low frequency generator 108 to the frequency of the high frequency generator 102 is less than 0.05, and in some embodiments the ratio of the frequency of the low frequency generator 108 to the high frequency generator 102 The frequency ratio is less than 0.01. For example, the ratio can be 1:150 or about 0.0067.

Depending on the application, the high frequency generator 102 can be used to ignite and maintain the plasma load in the plasma chamber 100, and the low frequency generator 108 can be used to apply a periodic voltage function to the substrate support of the plasma chamber 100 to achieve Desired distribution of ion energy at the surface of the substrate in the plasma chamber 100 .

With regard to power levels, the low frequency generator 108 may apply a relatively large amount of power (eg, in the range of 10 kW to 30 kW) to the plasma load of the plasma chamber 100 . A large amount of power modulation applied to the plasma at low frequencies modifies the plasma impedance presented to the high frequency generator 102 .

Applicants have found that in previous systems with generators that modulated the plasma load (eg, low frequency generator 108 ), power was not measured at a sufficient number of mixing products produced by the system. And failure to do so is a matter of generating errors of about 100% or more in power measurements. The typical approach taken in the past (when low frequency power disturbs the plasma) is to filter out only the mixed frequency components produced by applying high frequency power to a load modulated at low frequency (e.g., at low generator frequency and high generated 59.6 MHz and 60.4 MHz components are filtered out when the frequency of the filter is 400 kHz and 60 MHz respectively). But when the low pass filter is utilized, the apparent complex impedance locus shrinks to a point and, misleadingly, it appears that the high frequency generator 102 is delivering power down to 50 ohms.

Referring to FIG. 2 , there is shown a graph depicting how power may be perceived by measuring power using different measurement system filter bandwidths. Measurement system filtering is applied after down-conversion or demodulation of the measurement signal; thus, the measurement system filters frequency components centered on the generator output frequency. For example, a measurement system bandwidth of 100 kHz applied to a generator producing a 60 MHz output will suppress frequency components below 59.9 MHz and above 60.1 MHz. As shown, when the filter bandwidth of the measurement system is chosen to be smaller than the modulation frequency of the plasma, it appears that there is much less reflected power than actually exists (thus, it appears that only forward power is going into the plasma load), but that's not actually the case.

In contrast, when power is measured with sufficient bandwidth (e.g., by one or both of broadband measurement systems 116, 120), only a fraction of the apparent power (e.g., only half of the power ) is entering the plasma load. Therefore, one aspect of the present disclosure includes adjusting the measurement system so that its filter bandwidth exceeds the modulation frequency to capture the mixing products at higher frequencies. U.S. Patent No. 7,970,562, entitled "System, Method, and Apparatus for Monitoring Power," which is incorporated herein by reference, discloses various types of sensors (e.g., directional couplers or voltage/current (VI ) sensors) that can be used to implement the sensors 114, 118, and can be utilized by the broadband measurement systems 116, 120 to achieve filter bandwidths capable of capturing information about the mixing products at higher frequencies Sampling and processing techniques. It should be noted that the filter bandwidths of the measurement systems 116 , 120 should not be confused with the filter 104 .

Another problem is that the high frequency generator 102 needs to deliver power to a time-varying load (modulated plasma load), where the time-averaged load reflection coefficient magnitude is high. Referring to FIGS. 3A and 3B , for example, graphs depicting the load reflection coefficient seen by the high frequency generator 102 over the period of one cycle of the low frequency generator 108 are shown, and FIG. Graph of the resulting reflected power seen by the high frequency generator 102 for the filter 104 depicted in 1. As shown, the magnitude of the peak load reflection coefficient seen by the high frequency generator 102 can be close to 1 (and can even exceed 1, meaning that net power is flowing from the plasma load to the high frequency generator 102), while by The average load reflection coefficient value seen by the high frequency generator 102 may be 0.76. The relatively high load reflection coefficient magnitude means that, in general, the high frequency generator 102 may have difficulty applying the desired power level and be more prone to failure. Therefore, the high frequency generator 102 may require more power devices (bipolar transistors, MOSFETs, etc.) than would normally be required to deliver the required amount of power to the plasma chamber 100 .

Aspects of the disclosure herein relate to solutions to remove or mitigate the effects of plasma modulation. The aspect depicted in FIG. 1 is filter 104 depicted. As discussed above, in the absence of the depicted filter 104, the modulated plasma load presents a time-varying non-linear load to the high frequency generator 102, which presents a challenging problem.

In many embodiments, the filter 104 depicted in FIG. 1 may be implemented as a very narrow bandwidth, high power filter disposed between the high frequency generator 102 and the plasma chamber 100 . The filter 104 can have relatively low loss at the frequency of the high frequency generator 102 and suppress the mixing products sufficiently to limit the change in the load reflection coefficient presented at the input of the filter 104 to the high frequency generator 102 while under high power application for stable. When implemented, filter 104 may have a narrow bandwidth to filter sideband frequencies. As used herein, bandwidth is defined as the frequency range that exists between the lower cutoff frequency and the upper cutoff frequency, where each of these cutoff frequencies is 3 dB below the maximum center or resonant peak, while making both Other frequencies outside the point are attenuated or attenuated by more than 3 dB.

In some embodiments, for example, the low frequency generator 108 is implemented with a 400 kHz generator and the high frequency generator 102 is implemented with a 60 MHz RF generator; thus, a frequency ratio of 1 to 150 is present. Thus, in these embodiments, filter 104 may reject power at frequencies less than one percent away from the center frequency. As a specific example, the low frequency generator 108 may be a bias voltage supply that applies a function of voltage to the substrate support, and the high frequency generator 102 is a source generator that ignites and maintains the plasma.

And in many embodiments, the rejection of the power at the frequency of the high frequency generator 102 is at most 2 dB, and the rejection of the power from the frequency of the high frequency generator 102 at frequencies greater than the frequency of the low frequency generator 108 The rejection is at least 2 dB higher than the rejection of power at the high frequency generator 102 frequency. In some implementations, the bandwidth of filter 104 is 2% (or less) of the frequency of high frequency generator 102 . If the high-frequency generator 102 is realized by a 60 MHz RF generator, then, for example, the bandwidth of the filter can be 1.2 MHz or less.

Referring briefly to FIG. 4A , there is shown a graph depicting performance profiles for an exemplary design of filter 104 . In FIG. 4A, the bandwidth of filter 104 has a center frequency of about 60 MHz, and the power is rejected by 8 dB at fractions of a megahertz from the center frequency. Figure 4B shows the net power that can be delivered to the plasma load by the high frequency generator when the filter depicted in Figure 1 is not utilized. 4B shows that a filter with a response such as that shown in FIG. 4A will allow power delivered to the plasma load at the fundamental frequency of 60 MHz to be delivered from the high frequency generator 102 to the plasma load with relatively high efficiency, and will automatically The power reflected by the plasma load is suppressed back to the high frequency generator 102 .

But those of ordinary skill in the art have not yet led to implement a filter 104 with characteristics similar to the filter characteristics in FIG. 4A. A lack of awareness of the fundamental issues of plasma modulation is one reason. But in addition, designing a filter with the characteristics depicted in Figure 4A is challenging (even at low power levels). In many embodiments, however, filter 104 handles high amounts of power (eg, several kW of power), and the combination of high power and narrowband is not a combination that one of ordinary skill in the art might try.

As discussed above, FIGS. 3A and 3B depict the load reflection coefficient seen by the high frequency generator when the filter 104 is not utilized. And FIGS. 5A and 5B depict the load reflection coefficient seen by the high frequency generator 102 when the exemplary filter 104 is implemented. As shown in FIG. 5A , when filter 104 is deployed, the reflection coefficient is compressed to stay closer to the center of the graph over cycles of plasma modulation (compared to the loaded reflection coefficient in FIG. 3A ).

FIG. 3B depicts reflection coefficient magnitudes in the time domain without utilizing filter 104 . The corresponding level of forward power (nearly 100 watts) depicted in FIG. 3C is much lower than that utilized during plasma processing, but the reflection coefficient depicted in FIG. 3B and the forward and reflected power in FIG. 3C The relative magnitude is prescriptive. As shown, the forward power is 99.8 watts and the reflected power is 63.4 watts. In contrast, as shown in Figure 5C, with the filter 104 in place, there is 99.9 watts of forward power and 3.4 watts of reflected power; thus, the high frequency generator 102 is placed under much less stress Down. And on the load side of the filter 104, as shown in Figure 6C, the filter 104 can increase the average forward power.

Referring to FIG. 7 , there is shown a flowchart 700 depicting a method for plasma processing in a modulated plasma system. As shown, power is supplied to the plasma chamber 100 by the high frequency generator 102 to ignite and sustain the plasma (block 710). Additionally, power is supplied to the plasma chamber 100 by the low frequency generator 108 (block 720). Power transfer between the high frequency generator 102 and the plasma chamber 100 is suppressed at frequencies corresponding to the mixed product of the high frequency and the low frequency by a filter 104 disposed between the plasma chamber 100 and the plasma chamber 100. between high frequency generators (block 730). The tuning of the matching network 106 may be adjusted (eg, optimized) to balance the requirement to provide a well-matched impedance to the high frequency generator 102 and the efficiency of power delivery to the plasma chamber 100 (block 740 ).

Referring back briefly to Figure 6A, it should be noted that the locus of the load reflection coefficient is asymmetric about the origin, as is the case in Figure 3A. This is a characteristic of the required impedance on the load side of the filter 104 in order to match the input of the filter 104 to a load reflection coefficient close to zero and to obtain efficient power transfer from the high frequency generator 102 to the plasma load. The average load reflection coefficient on the load side of filter 104 is indicated by a "+" in FIG. 6A. The average value of the load reflection coefficient on the load side of the filter 104 as indicated in FIG. 6A is approximately -0.23 to j0.00. The average value of the load reflection coefficient on the high frequency generator 102 side of the filter 104 as indicated in FIG. 5A is approximately 0.04 to j0.02. This depicts the use of this filter 104, ie the load reflection coefficient on the load side of the filter 104 is not tuned to match the load (50 ohms in most systems), but is typically set so as to achieve Lower time-averaged load reflection coefficient values measured by a broadband measurement system. Thus, in many embodiments, broadband measurement component 116 or 120 is used to capture at least first order mixing products. The broadband measurement system 116 or 120 may be implemented as an integral component of the matching network 106, the high frequency generator 102, or may be implemented as individual components. Therefore, the step of adjusting the matching network at block 740 is different from the steps normally required by the matching network 106 .

In many implementations, the impedance presented by the plasma chamber 100 to the filter 104 is adjusted to optimize the efficiency of power transfer from the high frequency generator 102 to the plasma chamber 100 . For example, the time average of the absolute value of the load reflection coefficient presented to the filter can be minimized, and the load reflection coefficient can be measured using a bandwidth at least equal to the frequency of the low frequency generator 108 (e.g., by broadband metrology system 116 or 120). It is also considered that the time average value of the load reflection coefficient is optimized away from 0+j0.

Referring again to FIG. 7 , the length of the cable between the matching network 106 and the filter 104 may be adjusted (eg, optimized) to control the impedance by which the power mixing product is terminated (block 750 ). While the cable length (between the matching network and the plasma processing chamber) is adjusted (e.g. for stability) in other plasma processing systems, when using the filter 104 there are Additional considerations, namely: the terminating impedance provided by the filter 104 to the plasma system at the frequency of the hybrid product; the cables connecting the filter 104 to the matching network 106; and the matching network 106. Changing the cable length changes the nature of the modulation on the load side of the filter 104 . This cable length also affects frequency tuning in multi-state applications; therefore, the selection of this cable length can be more complex than in previous plasma processing systems.

8A and 8B are equivalent circuits of an embodiment of the filter 104 described with reference to FIG. 1 . Figure 8A shows the equivalent circuit of the lossless prototype, and Figure 8B shows the equivalent circuit of filter 104 when simplified to be implemented using realizable lossy components. Other ways of implementing this narrow-band, high-power filter exist (for example, using large ring resonators or cavities), but in all cases, careful attention must be paid to the high voltage, high current, and high power present in these filters dissipation.

Referring next to FIG. 9 , there is shown a perspective view of the exterior of a water-cooled filter 904 designed with two parallel spiral resonators. The filter contains two water connections 910 and 920 for passing water through the filter for cooling, an input connector 930 and an output connector (not visible in this view).

Figure 10 is a view of the interior of a filter 904 design with two parallel spiral resonators. As shown, each of the helical resonators includes hollow helical coils 1020 , and each hollow helical coil 1020 is coupled to a copper block 1024 . Extending from the copper block 1024 is a copper strap 1026 and insulating the copper strap 1026 from the copper block 1024 is a ceramic insulator 1028 . In this implementation, metallization 1030 is disposed on ceramic 1028 to form input capacitor 810 and output capacitor 820 . In addition, each hollow helical coil 1020 includes a ground terminal 1022 . Filter 904 also includes a potted cylindrical housing 1032 (shown transparently for purposes of viewing the internal components of filter 104 ) surrounding hollow helical coil 1020 and copper block 1024 .

FIG. 11 shows a cross-sectional view of filter 904 . This view shows how copper strap 1026 is connected to input connector 1110 and output connector 1140 to capacitors formed on ceramic insulators 1120 and 1150 . This view also shows how hollow helical coils 1130 and 1160 are connected to copper block 1024 .

Figure 12 shows copper block 1240 (1024 in Figure 10) in more detail. This assembly provides the required capacitive coupling from the input and output to the spiral resonator. Implementing capacitors on ceramic substrates is used in the design of filters due to the small value of the required capacitors, the high voltage the capacitors must withstand, and the power the capacitors must dissipate. The copper block contains a water channel 1210 into which the hollow helical coil is attached (by eg welding). The capacitors formed on ceramic insulators 1220 and 1260 are thus water cooled. The ceramic insulators have front metallization 1280 and rear metallization 1250 respectively. The size of the front metallization 1280 controls the capacitance implemented by the assembly. A ceramic insulator may be attached to copper block 1240 using conductive epoxy. Straps 1270 and 1230 may be soldered to the front metallization and to connectors 1110 and 1140 .

FIG. 13 shows an exploded view of filter 904 . An insulating bracket 1310 holds the hollow helical coil in place and provides mechanical stability to the assembly. The bracket is made of a suitable low loss dielectric material such as PTFE plastic or ceramic and contains holes to allow the potting material to flow through. Due to the high voltages that may be encountered in this design, the high voltage area of the filter is potted (for example, with a silicone dielectric gel) to reduce the risk of failure due to air breakdown. Alternatively, the entire assembly can be evacuated to high vacuum, filled with a high quality dielectric liquid, or filled with an insulating pressurized gas such as but not limited to sulfur hexafluoride (SF6).

It should be appreciated that, in view of this disclosure, one of ordinary skill in the art can design aspects of the helical coil 1020 (e.g., number of turns, radius, length, pitch, inner and outer coil diameters, and outer coil diameters of the coils). diameter) to achieve the desired bandwidth and heat dissipation. It should also be appreciated that variations in the design of the filter 904 depicted in FIGS. 9-13 are of course contemplated.

Using a spiral resonator instead of an inductor, which is close to resonance on the low frequency or inductive side of the resonance, achieves a similar bandwidth compared to a design using an inductor, but compared to a design with an inductor, the spiral resonator provides Smaller effective inductance. Additionally, the use of two resonators in parallel allows water cooling of the ground connection of the entire assembly, where the water system can remain grounded. More specifically, water provided from a ground-connected water system is fed through the hollow helical coil 1020, enabling a large amount of heat to be dissipated. For example, filter 904 (and variations of filter 904) may operate at relatively high power levels (eg, in the 1 kW to 30 kW power range). By virtue of its design, filter 904 (and variations thereof) can operate at relatively high power levels while operating at an efficiency of at least 75%.

Figure 14 shows a filter 1404 with tuning chips. Tuning may be required for setting the passband frequency of the filter due to component manufacturing tolerances, but may also be actively adjusted to compensate for component value variations due to, for example, self-heating of the filter 1404 . Tuning pellets 1420 and 1440 may be, for example, ferrite rods movable within hollow helical coil 1020 along the depicted Y-axis, but more typically, tuning pellets may be made of copper. Cups 1410 and 1430 made of a suitable insulator (eg, PTFE plastic) provide an area free of potting compound in which to move the tuning pellets.

Filters 104, 804B, 904, 1404 are used to compress the frequency range to a very small frequency range within which frequency tuning (for impedance matching) is possible. This requires a different approach to multi-state operation of generators. An example of multi-state operation may be switching between multiple power levels, where each power level represents a state and where the high frequency generator 102 sees different values in each state due to the non-linear nature of the plasma load. load impedance, and wherein the high frequency generator 102 may operate at a different frequency in each state in order to improve impedance matching or stability for that state. To facilitate multi-state operation in systems using filter 104, one may need to ensure that the impedance presented to the load side of filter 104 for the different states lies along or close to And the matching impedance line. This can be done by adding a fixed or variable time delay, such as delay element 112, on the load side of the filter.

An aspect of the present disclosure is frequency tuning of the high frequency generator 102 to adjust the impedance presented to the high frequency generator 102 . Referring to FIG. 15 , for example, a block diagram of a high frequency generator 1502 that may be used to implement the high frequency generator 102 described with reference to FIG. 1 is shown. As shown, high frequency generator 1502 includes exciter 1505 , power amplifier 1510 , filter 1515 , sensor 1520 and frequency tuning subsystem 1525 . An exciter 1505 (which may include an oscillator) generates an oscillating signal, typically in the form of a sine or square wave, at an RF frequency. The power amplifier 1510 amplifies the signal generated by the exciter 1505 to generate an amplified oscillation signal. For example, the power amplifier 1510 can amplify the exciter output signal from 1 mW to 3 kW. Filter 1515 is optional (and different from filter 104) and may filter the amplified oscillating signal to produce a signal consisting of a single RF frequency (sinusoidal).

Sensor 1520 measures one or more parameters indicative of plasma loading in plasma chamber 100 . In one embodiment, sensor 1520 measures a power parameter indicative of the impedance Z of the plasma load. Depending on the particular embodiment, sensor 1520 may be, for example but not limited to, a VI sensor or a directional coupler.

其中Γ（γ）係阻抗Z相對於所要阻抗 Z ₀之反射係數。反射係數（|Γ|）之量值係用以表現阻抗Z接近於所要阻抗 Z ₀之程度的極便利方式。Z及 Z ₀一般均為複數。 A measure of how close the load impedance is to the desired impedance can take many forms, but typically it is expressed as the reflection coefficient Γ =

where Γ(γ) is the reflection coefficient of the impedance Z relative to the desired impedance Z ₀ . The magnitude of the reflection coefficient (|Γ|) is a very convenient way of expressing how close the impedance Z is to the desired impedance Z ₀ . Z and Z ₀ are generally plural.

In general, the frequency tuning subsystem 1525 receives measurements from the sensor 1520 that are indicative of the impedance Z of the plasma load and processes these measurements to generate a frequency that is fed to the exciter 1505 via a frequency control line 1530 to adjust the frequency generated by the exciter 1505. frequency adjustment of the frequency.

As an alternative to sensor 1520 (discussed below), sensor 114 may be used to measure a power parameter on the load side of filter 104, and broadband measurement system 116 may be indicative of the impedance Z of the plasma load. The signal is provided to frequency tuning subsystem 1525 .

Frequency tuning subsystem 1525 performs calculations (based on frequency tuning methods) to generate frequency adjustments (eg, frequency steps) that are fed to exciter 1505 via frequency control line 1530 . In some use cases, the goal is to adjust the frequency of exciter 1505, thereby changing the impedance of the plasmonic load in a way that minimizes |Γ| (ie, it achieves Γ as close to zero as possible). The frequency at which this minimum |Γ| is achieved may be referred to as the target frequency. As understood by those of ordinary skill in the art, an ideal complex reflection coefficient of zero corresponds to a matching condition where the plasma load impedance perfectly matches the desired impedance Z ₀ . In other embodiments, the target is not minimum |Γ|. Instead, the frequency tuning subsystem 1525 intentionally tunes the exciter 1505 to produce frequencies other than the frequency that produces the minimum |Γ|. This embodiment may be referred to as a "detuned" implementation, and the target frequency may not minimize Γ.

Referring next to FIG. 16 , there is shown a flowchart 1600 depicting a method that may be performed in conjunction with embodiments disclosed herein. As shown, power having a multi-state waveform is applied to the plasma chamber 100 by the high frequency generator 102, 1502 (block 1610), and power is also applied to the plasma chamber by the low frequency generator 108 100 (block 1620). Mixed products of high and low frequencies are suppressed by the filter 104 (block 1630). And as shown, the power signal between the filter 104 and the plasma chamber 100 is delayed (block 1640), and the frequency of the high frequency generator 102 is adjusted during each of these states to adjust the presented Impedance to high frequency generator (block 1650).

Accurate power measurements may require measuring the power on the load side of the filter 104 with a bandwidth sufficient to capture a sufficient number of mixing products. This is because the efficiency of the filter 104 depends on the locus of the load impedance presented to the filter 104 . Measurements on the high frequency generator 102 side of the filter 104 may not provide an accurate measurement of the power delivered to the plasma load because it is difficult, if not impossible, to take into account the efficiency of the filter 104 .

A variety of different frequency tuning methods may be used to adjust the frequency of the high frequency generator 102 at block 1650 . In general, frequency tuning methods determine in which direction to adjust the frequency (increase or decrease the frequency) and determine the magnitude of the frequency steps used in making the change in frequency.

Assuming that the ideal operating frequency is the frequency at which the magnitude of the load reflection coefficient is at or substantially close to its minimum value, it should be noted that the relationship between the controlled variable (frequency) and the error is not necessarily monotonic. Furthermore, the optimum point of operation is at the point where the gain (defined as the change in error divided by the change in frequency) is zero.

To add to the challenge, it is also possible that local minima may exist in regions where control methods may be trapped. In some special cases where a priori information about the load is known, it is possible to configure the error function as a monotonic function of frequency, so that a simple linear controller can be used. For example, such a system is disclosed in US Patent No. 6,472,822, issued October 29, 2002 to Chen et al., entitled "Pulsed RF Power Delivery for Plasma Processing." This linear control is rarely applicable due to the non-monotonic relationship between frequency and error, except in these special cases where a priori information about the load is available.

It has been found that two common problems with plasma loading are: (1) the nonlinear nature of the loading, since the plasma loading impedance is a function of power level; and (2) the changing chemistry of the loading impedance due to the nonlinear plasma loading. Properties, pressure, temperature, and other physical properties change over time. Another problem specific to plasma (or plasma-like) loads is that the plasma can go out if the power delivered to the plasma falls below the minimum for a long enough time. Consequently, frequencies that do not deliver sufficient power to the plasma load cannot be applied for very long periods of time, or the plasma will go out.

Furthermore, frequency tuning becomes even more problematic when power (eg, RF power) to the load is pulsed. Due to the nonlinear nature of the load and the relatively high figure of merit (the ratio of stored energy to delivered energy per cycle (e.g., RF cycle), often denoted "Q") used by the impedance matching network, the load impedance in The applied pulse (eg RF pulse) changes very rapidly during the preceding few microseconds.

The disclosure of U.S. Patent No. 7,839,223, entitled "Method and Apparatus for Advanced Frequency Tuning," issued by van Zyl et al. on November 23, 2010 (the '223 patent), incorporated herein by reference, may be incorporated herein by reference Various frequency tuning methods used in the embodiment. In one approach described in the '223 patent, if the error (e.g., the difference between the desired value of γ and the actual value of γ) is decreasing stepwise, the frequency step is allowed to increase, and if the error is increasing stepwise, Then the frequency step can be reduced (or kept constant). This approach can be used in conjunction with embodiments disclosed herein to help maintain time-varying loads (eg, to limit or reduce changes in time-varying load reflection coefficients).

A method for simultaneous application of multi-state waveforms (eg, pulses) and frequency tuning discards information at the beginning of the pulse while the impedance is still changing rapidly and uses only the information to effectively control the frequency when the load impedance is stable. This approach avoids the need for in-pulse tuning, but manages to obtain a good average operating frequency.

To avoid aliasing effects, measurement and control can be synchronized to the rising edge of the pulse. By delaying the start of the measurement and control cycle from the start of the pulse, it is possible to operate properly on plasma-type loads. Typically, discarding the first 10 microseconds after the start pulse is sufficient to achieve reasonable results.

In some cases it is not possible to completely discard information at the beginning of the pulse, but due to the danger of aliasing effects, either due to insufficient control of the bandwidth, or due to the high bandwidth requirements of the frequency control system Risk of unstable operation, use of in-pulse information is undesirable. By using memory, it is possible to design a system with similar performance to a true intra-pulse control system, but implemented using a slower speed stabilizing controller that controls inter-pulse information.

Since the measurement and control loops can be synchronized to the pulses, it is possible to control frequency on a pulse-to-pulse basis using the same time slot in sequential pulses and using a slower (compared to intra-pulse controller) control system. Additionally, measurements of the same time slots of sequential pulses may be combined with measurements of time slots adjacent to these time slots. Not only frequency, but other control parameters can also be stored and used by the control system to, for example, control the power delivered to the load. These other control parameters may include the DC voltage supplied to the power device, the gate bias voltage in the case of a MOSFET (base emitter in the case of a bipolar device), and the RF drive level. Graphs depicting the operation of pulse-to-pulse controlled systems for high pulse repetition frequencies are shown in FIGS. 17 , 18 and 19 . If the pulse on-time becomes extremely long, it may be more advantageous to just ignore the information from the first few time slots, or to switch to intra-pulse control sometime later in the pulse.

Figures 17, 18 and 19 together illustrate the disclosed pulse-to-pulse frequency tuning. In this scheme, f _a2 is a function of only (or predominantly if adjacent time slots are also considered to have some weight) e _a0 , e _a1 and f _a1 . Similarly, f _b2 is a function of only (or primarily) e _b0 , e _b1 and f _b1 , and so on.

Another problem is entrapment in local non-optimal minima. Using the fact that there is a fixed time during which a plasma can operate at substantially reduced power without extinguishing, it is possible to sample and store information about operation at a frequency completely different from the current operating point. Assuming that the plasma will not be extinguished if the power is substantially reduced for a short enough time T, the method operates by operating at the optimal frequency (as judged by the frequency tuning method) e.g. 99% of the time and for a duration not exceeding The remaining 1% of the time slot at T is spent exploring operations at other frequencies. In some implementations, power may be delivered at the selected frequency for at least 90% of the total time, and power may be delivered at the test frequency during a test period no longer than 10% of the total time.

In some implementations, the value of the test frequency can vary according to different test periods. In other implementations, the same test frequency is accessed multiple times, each time adjusting the power delivered to the plasma load toward a desired power level.

While many variations are possible, the following methods are exemplary and illustrative. Consider operating at the optimal frequency for a time equal to 99T and then switching to a different test frequency for T time. The entire frequency range from f _min to f _max can be divided into eg 16 equally spaced frequencies f ₀ to f ₁₅ . The number of frequencies used to divide the entire frequency range is a function of the known quality factor of the matching circuit employed. Sixteen is the typical number to ensure that no true sweet spot will be missed in subsequent searches for the best frequency.

The method may start by sequentially searching f ₀ to f ₁₅ in time slots of duration T to find a rough optimum. Several searches of space may be required because the power control system may not be able to correctly adjust the power within the time T. Due to the non-linear nature of typical loads encountered, it is beneficial to measure the load reflection coefficient (or other error metric used by the method) at or near the desired power level. By storing the control value and power level each time a frequency is accessed, the correct power level can be obtained after accessing the same frequency several times.

Once a rough optimum has been found, for example at fk, where k _is an integer from 0 to 15, the method can start using time slots of length T to find the optimum. One option would be f ₁₆ =0.5( f _k-1 + f _k ) with the constraint that k >0, and f ₁₇ =0.5( f _k + f _{k +1} ) with the constraint that k <15. The frequency at which the error is at the minimum value between f ₁₆ , f _k and f ₁₇ then becomes the new desired frequency. Divide the left and right intervals of the new optimum value into two again, and select the minimum value among the previous minimum frequency and the two latest test frequencies. And only generate a new frequency when the minimum frequency happens to be f _min or f _max . Due to the fact that the interval is bisected each time, the optimal frequency is found with sufficient accuracy within only a few iterations. And because the load is generally time-varying, once the optimum frequency has been found, the method must generally start to ensure that conditions have not changed and a new global optimum has not been produced.

While this method to find the global optimum is being performed, the previously described local tuning method can be run during the 99T time slot to maintain operation at the current local minimum. And 99% of the time staying at the current optimum frequency ensures that the average delivered power to the load remains almost constant. 20 graphically depicts exemplary operating characteristics that may be associated with a method of searching for an overall best frequency using a smaller percentage of time and a maximum time slot T. FIG.

Another method of frequency tuning the high frequency generator 102 to adjust the impedance presented to the high frequency generator 102 is described with reference to FIGS. 21-38 . For example, in some implementations, the power supply 2110 , circuitry 2114 , sensors 2112 , power generation systems 2200 , 2300 , 2400 , 2500 , 2600 , 2700 described further herein may be integrated with the high frequency generator 102 . In other embodiments, one or more components of the power source 2110, the circuit 2114, the sensor 2112, and the power generation systems 2200, 2300, 2400, 2500, 2600, 2700 may be distributed. For example, circuits (e.g., circuits 2114, 2214, 2314, 2414) and/or sensors (e.g., sensors 2112, 2312, 2412) may be implemented in a centralized control housed separately from high frequency generator 102 device. As an additional example, circuits (eg, circuits 2114 , 2214 , 2314 , 2414 ) may be implemented as part of frequency tuning subsystem 1525 .

As described with reference to FIGS. 21-38 , plasma sustaining power can be applied at a primary frequency while a secondary power signal (e.g., comprising one or more frequencies) at a much lower power than the plasma sustaining power is used to detect alternative primary frequency. Advantageously, application of the secondary power signal enables monitoring of one or more aspects of the plasma load without adversely affecting the plasma load itself. Additionally, when the plasma sustaining power is applied to the plasma load via the matching network, the application of the low level signal can be applied at one or more specific frequencies that produce detectable Frequency of measurement (for example, mixing and intermodulation frequencies). Additionally, the information obtained about plasma loading can be used to control one or more aspects of the generator.

For example, for generator control, automatic frequency tuning can be performed using information about plasma loading. For example, a global optimum for some measure of performance may be obtained, and the high frequency generator 102 may be tuned towards this global optimum - without extinguishing the plasma. Exemplary methods include processing noise generated by the generator's primary operating frequency to efficiently perform low power sampling sweeps of relevant frequency ranges or to generate low power signals other than the primary power signal, where the low power signal is used for exploration or detection Global best value.

In both cases, the low-power nature of the noise or probe signal enables exploration of the frequency range of one or more probe frequencies while the generator's main power signal remains at one frequency (e.g., in local optimal value), where enough power can be delivered to the plasma load to maintain the plasma. For example, the primary power signal can be kept at or close to a local optimum while the search signal or noise (both will be referred to as "secondary power signal" or "probing signal" hereinafter) finds global optimum, thereby continuing to allow substantial power to reach the plasma load while detection occurs.

Where the secondary power signal is noise, the noise may be inherent noise due to the primary power signal, or noise may be added to the primary power signal. The noise can appear at secondary probing frequencies that are sometimes limited to the bandwidth governed by the filter applied to the primary power signal. Where the secondary power signal is a low level signal, this probe signal can be orders of magnitude lower than the amplitude of the primary power signal (e.g. -3 dB, -5 dB, -10 dB, -20 dB, -50 dB, -100dB). The low level signal may be sinusoidal or any other type of periodic signal and may be generated at RF or other frequencies. A signal that begins in finite time and eventually becomes sinusoidal or periodic is considered sinusoidal or periodic, respectively. The low-level probe signal can be swept across a fixed range of secondary probe frequencies. Alternatively, the low-level probing signal may "hop" between secondary probing frequencies according to a tuning algorithm that searches for a global optimum.

The global optimum can be found by comparing the optimum of different frequencies and selecting the best frequency. For example, if the measure of optimality is the minimum load reflection coefficient value, then the estimated load reflection coefficient values at different frequencies explored by the secondary power signal source are compared and the load reflection coefficient minimum value is selected. The hourly frequency is used as the global best frequency. Measurements and comparisons to find the optimum can take place sequentially or, for example, in the case of noise used as a secondary power signal, the optimum at different frequencies can be calculated simultaneously and at different frequencies Then choose the best frequency.

Once the global optimum has been found, the main power signal can be shifted to the frequency of the global optimum. This shift may involve a sharp switch from one frequency to another, or may involve a power ramp-up of the secondary power signal while a power ramp-down to the primary power signal is made such that the secondary power signal becomes the primary power signal Signal.

Once the main power signal is operating at the frequency of the global optimum, further fine tuning can occur. For example, the secondary power signal can go off again while searching for a global optimum because the global optimum at the power level of the primary power signal differs from the global optimum at a lower power for the secondary power signal , or as the global optimum changes and as the first iteration of tuning occurs.

For the purposes of this disclosure, a "low level signal" is substantially lower, eg, at least an order of magnitude smaller, than the primary signal delivered to the plasma chamber.

For purposes of this disclosure, a "circuit" may include any combination of electrical components that generate an output signal based on an input signal. The circuitry may be or include a digital, analog, or part of a processor or central processing unit (CPU). The circuitry may comprise or be readable from a non-transitory tangible computer-readable storage medium having processor-readable instructions for performing the methods described below.

For the purposes of this disclosure, components are in communication, which in some cases includes electrical communication (eg, capable of sending signals between them). However, those of ordinary skill in the art will recognize that communications may also include optical and wireless radio communications, just to name two non-limiting examples.

For purposes of this disclosure, a "global optimum" may include, for example, the minimum or maximum value of a characteristic sampled across a frequency range. For example, where reflected power is the characteristic, the global optimum may be the global minimum, while where delivered power is the characteristic, the global optimum may be the global maximum.

Figure 21 depicts a power supply system configured for automatic frequency tuning of power delivered to a plasma load. The power generation system 2100 is configured to provide radio frequency (RF) power to the plasma 2106 or the plasma load via an RF impedance matching circuit, which may be an optional filter 2122 internal to the power supply 2110 and/or Matching network 2104 external to power supply 2110 . Filtering and impedance matching are usually performed by the same physical network. Thus, a filter such as optional filter 2122 can perform both filtering and impedance matching functions.

Power generation system 2100 may include power supply 2110 that converts external power 2140 to RF power, and power supply 2110 may be a 13.56 MHz generator, although this is of course not required. Consider other frequencies and other power supplies. Power generation system 2100 is configured to provide RF power (eg, RF voltage) at a sufficient level to ignite and maintain plasma 2106 contained in plasma chamber 2108 . Plasma 2106 is typically used to process workpieces or substrates (not shown), but is well known to those of ordinary skill in the art.

The power supply 2110 may apply a primary power signal to the output 2111 primarily at the primary frequency. Output 2111 may be configured for coupling to optional matching network 2104 and to plasma chamber 2108 . In particular, the primary power signal may be delivered to the plasma 2106 or to a load of the plasma 2106 (also referred to as a plasma load). The connection 2130 from the power source 2110 to the optional matching network 2104 is typically a coaxial cable, although other cable types and connection types are possible. The connection 2131 from the matching network 2104 to the plasma chamber 2108 is typically made via a custom coaxial connector, although other cable types and connection types are possible. In some applications, the matching network 2104 is absent and the output 2111 of the power supply 2110 is directly connected to the plasma chamber 2108 . In this case, RF impedance matching is performed inside the power supply 2110 by an optional filter 2122.

In some applications, other optional RF or DC generators 2150 may be connected to plasma chamber 2108 via optional matching network 2104 . And in some applications, other optional RF or DC generators 2151 (e.g., low frequency generator 108) may be connected to plasma chamber 2108 via other components (e.g., other optional matching network 2105) . Other generators connected to the plasma load via the matching network 2104 or via other means (eg, connections to different electrodes to deliver power to the same plasma) often complicate the frequency tuning problem. In the following description, the possibility of selecting one or more generators 2150 and 2151 and other components connected to the plasma (for example, matching network 2105) as appropriate is not excluded, but for the sake of simplicity, they will not be To illustrate or describe.

The sensors 2112 may monitor characteristics indicative of the power delivered by the generator or the ability to deliver power, such as reflected power, delivered power, or impedance mismatch, just to name three non-limiting examples. Other non-limiting examples of characteristics indicative of delivered power or the ability to deliver power include power delivered to matching network 2104, power reflected from matching network 2104, power delivered to plasma chamber 2108, seen by power generation system 2100 load impedance, and properties of the plasma chamber 2108 such as plasma density. Sensors 2112 may also monitor characteristics indicative of the stability of the plasma system, such as fluctuations in load impedance. Sensor 2112 may also monitor characteristics indicative of nonlinear properties of the plasmonic load, such as the generation of mixing and intermodulation products.

Using a secondary signal source to perform frequency tuning of the generator has an additional benefit: measurements of plasma properties can be made by the generator. Optional matching network 2104 typically acts as a bandpass filter. This property of the matching network 2104 makes it difficult to make reliable measurements of the plasma at harmonics of the generator output frequency, but this information can be useful. However, the modulation of plasmonic impedance can be characterized by observing mixing and intermodulation products produced by secondary signal sources. For example, if the primary signal source is at 13.56 MHz and the secondary signal source is at 13.57 MHz, then one would expect the mixing products to be at 13.55 MHz and the intermodulation products to be at 13.56 plus multiples of 10 kHz, such as at 13.53, 13.54, 13.58, etc. Measuring the amplitude and phase relationship of mixing and intermodulation products and deriving, for example, the amount of amplitude and phase modulation present can provide information about the properties of the plasma. The processing of the information can be done in several ways: simply analyzing the time series of measurements from the sensors and performing higher order statistics on the time series using a dedicated receiver tuned to the frequency of the mixing and intermodulation products; extracting the amplitude and phase The relationship can use any number of mathematical transformations including, but not limited to, the discrete Fourier transform. Monitoring mixing and intermodulation products and detecting changes in properties of the plasma indicated by, for example, the amount of phase modulation is just to name one attribute that can be used in, for example, endpoint detection in semiconductor manufacturing, such as in etching operations.

The sensor 2112 can be a directional coupler, a current-voltage sensor, or other multi-port network, and can monitor the voltage between the power supply 2110 and the matching network 2104 or between the matching network 2104 and the plasma chamber 2108. A combination of current and voltage or voltage and current (for example, incident and reflected signals). In another non-limiting example, sensor 2112 may be an optical detector directed into plasma chamber 2108 to optically measure the density of plasma 2106 . These examples in no way describe the range or limits of the sensor 2112 or the location in which the sensor 2112 can be configured, but in fact demonstrate that the sensor 2112 can take many forms and can be coupled to the system in a variety of ways (various non-limiting See Figures 22 to 27 for examples). Additionally, the sensor 2112 may be one or more sensors already resident in the optional matching network 2104 or plasma chamber 2108 .

Signals from sensors 2112 or sensors already resident in matching network 2104 and plasma chamber 2108 may be provided to one or more circuits 2114 that are also in communication with power supply 2110 and control the power. One or more circuits 2114 may use information from sensors 2112 and/or sensors already resident in matching network 2104 and plasma chamber 2108 to tune the primary and/or secondary for power supply 2110 operation. The frequency is probed to optimize the power delivered to the plasma 2106 or to optimize another measure of optimum such as plasma stability.

In some cases, such tuning results in operation at a local optimum (e.g., a local minimum of reflected power or a local maximum of delivered power, to name two examples), so some tuning algorithms are able to further tune the main frequency in order to find the global optimum (for example, via a series of rapid frequency "hops"). However, this search can harvest power through regions of the frequency spectrum with insufficient impedance matching (e.g., near fa in FIG. , near fa in Figure 28).

To avoid this, the search for the global optimum can be performed with one or more secondary signals, thus enabling the high power primary power signal to remain at sufficient power to be delivered to the plasma while the search for the global optimum continues at a frequency of 2106 (for example, at a local optimum). Figures 31-33 show plots of the monitored characteristic as a function of frequency and how a secondary power signal having a substantially lower amplitude than the primary power signal can be used to search for an overall optimum. These curves will be discussed in depth later, once the related systems and devices have been described.

Figure 21 depicts a power generation system for automatic frequency tuning of power delivered to a plasma load. A power supply 2110 may provide the main power signal to the plasma load of the plasma 2106 in the plasma chamber 2108, wherein the impedance seen by the power supply 2110 is compatible with the matching network 2104 and the power supply disposed between the power supply 2110 and the plasma chamber 2108 2110 frequency tuning impedance matching. The power supply 2110 can be frequency tuned to find the optimal frequency, typically where the delivered power is optimized, although other measures of optimality can be used. This tuning can sometimes cause the main power signal from the power supply 2110 to be tuned to a local optimum rather than a global optimum. In such cases, a probe signal comprising one or more probe frequencies may be generated by the power supply 2110 and processed to identify the global optimum without having to use the primary power signal to search for the global optimum.

In other cases, the secondary power source may provide a secondary power signal (also referred to as a probe signal) (see, eg, FIGS. 24 and 26 ). One or more secondary power signals may be provided at an amplitude or power level lower than that of the primary power signal (or substantially lower than the primary power signal, a portion of the primary power signal, or at a substantially lower power The level is provided so as to have a negligible effect on the plasma 2106 compared to the main power signal). The probing signal may comprise a plurality of secondary probing frequencies all generated simultaneously (eg, FIGS. 31-33 ). In an alternative, one or more secondary power signals may be tuned to two or more different frequencies at different times (eg, as depicted in FIGS. 31-33 ).

One or more secondary power signals may be used to sample power delivery at frequencies other than that of the primary power signal without applying too much power at these secondary frequencies to affect the plasma. In other words, the primary power signal may be maintained at a frequency at which the plasma is sustainable (e.g., at or near a local optimum) while one or more secondary power signals are used to Search for the global best value.

In particular, sensor 2112 or two or more sensors and/or sensors already present in other components of power generation system 2100 may monitor the frequency of the primary power signal as well as at the secondary frequency measure of effectiveness. One or more sensors (eg, sensor 2112 ) may also measure at frequencies where mixing and intermodulation products are expected to extract information about the nonlinear properties of plasma 2106 . For example, changes in mixing and intermodulation products can be used to sense plasma ignition or endpoint detection for plasma processing. The injection of one or more secondary frequency components and the measurement of the properties of the mixing and intermodulation products can sense the nonlinear characteristics of the plasma 2106 at harmonics of the primary power signal, even though the matching network 2104 and filter 2122 may Direct measurement of harmonics is not allowed.

For example, sensor 2112 may be a reflected power sensor or a delivered power sensor, and the characteristic may be reflected power or delivered power, respectively. Other characteristics can also be monitored and used to identify local and global optima (e.g., load impedance seen by power source 2110, voltage and current of power on supply cable 2130 to matching network 2104, and plasma 2106 density , just to name a few non-limiting examples). Sensors 2112 and/or other sensors may provide information descriptive to one or more circuits 2114 (e.g., logic circuits, digital circuits, analog circuits, non-transitory computer readable media, and combinations thereof ). One or more circuits 2114 may be in communication (eg, in electrical communication) with sensor 2112 and power source 2110 . One or more circuits 2114 may adjust the primary frequency of power supply 2110 in order to tune power supply 2110 to optimize the delivered power to the plasma load.

In some embodiments, optimizing the measure of performance includes controlling a feedback loop using the secondary power signal in order to explore or search for a global optimum. In this case, one or more circuits 2114 may be based on information from sensor 2112 (or two or more sensors, and/or sensors already present in other components of power generation system 2100) regarding The secondary power signal and its one or more secondary frequencies are controlled in response to the measured performance. For example, the frequency of the secondary power signal may be swept across a fixed frequency range covering the primary frequency of the primary power signal, and one or more circuits 2114 may monitor a measure of performance as a function of frequency of the secondary power signal . Based on this sweep, one or more circuits 2114 may identify a global optimum and then instruct power supply 2110 to adjust its primary frequency in order to move the primary power signal to the identified global optimum. Frequency hopping or other tuning schemes can be used to find the global optimum via one or more secondary power signals.

The secondary power signal can take several different forms. In one instance, one or more circuits 2114 may instruct power supply 2110 to apply a low-level signal at one (eg, as depicted in FIG. 11 ) or multiple (eg, as shown in FIG. 32 ) secondary frequencies. The secondary power signals, or apply low level signals at these secondary frequencies in a specific order (eg, Figure 31), or according to an algorithm to optimize the measurement of performance (eg, Figure 32). In another case, one or more circuits 2114 may instruct power supply 2110 to apply a secondary power signal in the form of noise. This noise may be intrinsic to the main power signal, in which case one or more circuits 2114 may not necessarily supply commands to the power supply 2110, or the noise may be extrinsic noise added to the output of the power supply 2110 (e.g. , as shown in Figures 26 and 27).

Regardless of the form in which the secondary power signal occurs, in many embodiments its amplitude is one or more orders of magnitude lower than that of the primary power signal. For example, the secondary power signal may be 1 dB to 100 dB lower than the primary power signal. In other embodiments, the secondary power signal may be 1 dB, 5 dB, 10 dB, 20 dB, 50 dB or 100 dB lower than the primary power signal.

As shown, one or more circuits 2114 may include a global optimum identification module 2116 and a frequency control module 2118 . The global optimum identification module 2116 may analyze information from the sensor 2112 at each of the one or more secondary frequencies and identify the frequency corresponding to the global optimum. This frequency may be referred to as the identified global best frequency, and it corresponds to the global best value for the characteristic of the power delivered by the generator. Frequency control module 2118 may adjust the primary frequency of the primary power signal during initial tuning of the primary power signal, which may result in the identification of a local optimum, as well as towards a global optimum when the global optimum is identified by the global optimum identification module 2116. Identify the best frequency in the world and adjust the main frequency.

In detail, once the identified global best frequency is identified, the frequency control module 2118 may instruct the power supply 2110 to adjust the main frequency to jump to the identified global best frequency or to reduce the amplitude of the main frequency while increasing The amplitude of the secondary frequency at the global optimum frequency is identified such that the primary frequency and the secondary frequency invert the effect. In this way, the dominant frequency can be shifted to a frequency corresponding to the global optimum of the power characteristics (e.g., low reflected power or low level oscillations) without suppressing or extinguishing the plasma in the frequency spectrum (e.g., Figures 28-28). 33 in the vicinity of fa ) to apply power in the area.

The operations of the global best value identification module 2116 and the frequency control module 2118 can be looped to iteratively improve the accuracy of tuning the primary frequency to the global best value. For example, where the characteristic being monitored (e.g., plasma impedance) is non-linear, the global minimum of the characteristic can be found when a low-level secondary power signal is applied, but when a large When there are many dominant power signals, different global optimal frequencies may exist for higher power signals. Thus, the secondary power signal can again be used to further guide the global optimum of the primary power signal and this can continue in a round-robin fashion for multiple iterations. Adjusting the frequency toward the global best value may include changing the frequency to a frequency associated with the global best value, or simply changing the frequency to a frequency that is closer to the global best value than the original frequency.

In some embodiments, once one or more secondary frequencies begin a sufficiently steep portion of the falling/rising frequency curve (eg, between fa and f0 in FIGS. 28-34 ), the primary frequency may switch to one or more one of the minor frequencies. When this steep portion of the curve is identified, the global optimum identification module 2116 can determine that it is approaching the global optimum and thereby instruct the power supply 2110 to switch the primary frequency to a frequency that is close to the frequency of the secondary power signal, thereby This enables the main power signal to skip and avoid regions of the frequency curve (eg, near fa ) that may suppress the plasma. Once the primary power signal switches frequency, one or more secondary power signals can continue to lead to the global optimum, or the primary power signal can be used to further steer to the global optimum.

In many embodiments, the power connection 2130 may be implemented by a pair of conductors or a set of two-conductor coaxial cables connecting the power source 2110 and the matching network 2104 . In other embodiments, cable 2130 is implemented with one or more twisted pair cable wires. In yet other embodiments, cable 2130 may be implemented by any cable network including, but not limited to, simple conductor connections and four-pole connections. Connections 2131 are typically made by connectors, but can take a variety of forms including simple conductor connections.

The matching network 2104 can be realized by various matching network architectures. The matching network 2104 may be used to match the load of the plasma 2106 to the power source 2110 as will be appreciated by those of ordinary skill in the art. With proper design of the matching network 2104 or 2105, it is possible to transform the impedance of the load of the plasma 2106 to a value close to the desired load impedance of the power source 2110. Proper design of the matching network 2104 or 2105 may include matching networks internal to the power supply 2110 (eg, via filter 2122 ) or external to the power supply 2110 as shown in FIGS. 21-27 .

One or more circuits 2114 may be original equipment of power generation system 2100, while in other embodiments one or more circuits 2114 may be retrofit components that may be added to power generation that were not originally capable of frequency tuning as described herein system.

In one embodiment, the power generation system 2100 may include an optional filter 2122 . Filter 2122 may be configured to attenuate a portion of the main power signal outside the selected bandwidth and to perform additional impedance matching. For example, since 50 ohms is the predominant impedance of the cables and connections 2130, the desired impedance seen at the output of the power supply 2110 is typically 50 ohms or some other convenient impedance. The impedance at the input of the filter 2122 (on the opposite side from the output of the power supply 2110) provides the impedance required by the active components of the power supply (e.g., MOSFETs) and is typically very different from 50 ohms, such as 5+j6 ohms typical for single MOSFET amplifiers. For such a system, the filter 2122 is then designed to match 50 ohms at the output to 5+j6 ohms at the input. In addition to impedance matching, filters are also typically designed to limit harmonics generated by active components. For example, a filter can be designed to match 50 ohms at the output to a value close to 5+j6 over the frequency range over which the generator is expected to operate, such as from 12.882 to 14.238 MHz, and at frequencies above 25 MHz The signal is suppressed by some amount, typically at least 20 dB at the second or third harmonic of the output.

The sensors 2112 may be deployed in a variety of locations, including locations that are part of the power generation system 2100 and locations external thereto. Where the sensor 2112 monitors characteristics may also vary from embodiment to embodiment, as will be seen in FIGS. 22-27 .

FIG. 22 illustrates one embodiment of a power generation system 2200 in which a sensor 2212 resides within the power generation system 2200 along with a power source 2210 and one or more circuits 2214 . Power generation system 2200 includes output 2220 configured for coupling to optional matching network 2204 or directly to plasma chamber 2208 in the absence of matching network 2204 . Accordingly, a primary power signal and one or more secondary power signals may be provided to output 2220 and thus configured for delivery to matching network 2204 .

FIG. 23 illustrates an embodiment of a power generation system 2300 in which a sensor 2312 resides external to the power generation system 2300 . Here, a power generation system 2300 includes a power source 2310 , one or more circuits 2314 , an optional filter 2322 , and an output 2320 to the power generation system 2300 . Sensors 2312 are coupled to one or more circuits 2314 and provide information describing performance measurements such as load reflection coefficient values or plasma density. If the matching network 2304 does not exist, the sensor 2312 monitors between the power generation system 2300 and the optional matching network 2304, between the matching network 2304 and the plasma chamber 2308, or at the plasma chamber 2308, Or the characteristics between the power generation system 2300 and the plasma chamber 2308 . Sensors 2312 may also perform monitoring at or within matching network 2304 .

Although Figures 21-23 depict a single power supply 2110, 2210, 2310, one of ordinary skill in the art will recognize that this power supply 2110, 2210, 2310 is capable of generating both primary and secondary power signals simultaneously. For example, the power supplies 2110, 2210, 2310 may supply both a high-power primary power signal (eg, using a primary oscillator) and a low-level secondary power signal (eg, using a secondary oscillator), or the power supplies 2110, 2210, The 2310 can supply a high power primary power signal (eg, by a single oscillator-amplifier combination) and use the noise inherent in that primary power signal as a secondary power signal, just to name two non-limiting examples. Alternatively, the power supplies 2110, 2210, 2310 may generate the main power signal (eg, by a single oscillator-amplifier combination) and combine this with generated or amplified noise. While each of these examples shows how a single power supply 2110, 2210, 2310 can generate both primary and secondary power signals, FIGS. Embodiment of secondary power signal.

24 illustrates an embodiment of a power generation system 2400 having a power source 2410, a low level signal source 2411, one or more circuits 2414, optional sensors 2412 that may be configured within the power generation system 2400 or may be An optional sensor 2413 disposed outside the power generation system 2400 and a combiner 2424 combining outputs from the power source 2410 and the low-level signal source 2411 . As will be appreciated by those of ordinary skill in the art, the combiner can be implemented with couplers known in the art.

FIG. 25 illustrates an embodiment of a power generation system 2500 in which the primary and secondary signals are combined before being amplified by a power amplifier 2550 .

Figure 26 illustrates an embodiment of a power generation system 2600 in which a power source 2610 generates a primary power signal and a noise source 2613 generates a secondary power signal in the form of noise. The primary power signal and the secondary power signal or noise may be combined in the power generation system 2600 and the combined signal may be provided to an output 2620 of the power generation system 2600 . As will be appreciated by those of ordinary skill in the art, the noise source 2613 can be implemented by many different types of devices including noise diodes. Advantageously, the noise source 2613 can generate a continuum of secondary frequencies, and the responses of the secondary frequencies can be processed in parallel at multiple different frequencies (eg, by multiple demodulation channels or fast Fourier transform modules). For example, reflection coefficients at multiple frequencies can be arrived at in parallel to identify frequencies that provide low reflection coefficients, stable frequencies, or a balance between stability and low reflection coefficients.

FIG. 27 illustrates an embodiment of a power generation system 2700 in which the primary and secondary signals are combined before being amplified by a power amplifier 2750 . In this embodiment, the secondary signal is generated by noise source 2713 .

The systems depicted in FIGS. 21-27 can be more easily understood with reference to the curves seen in FIGS. 28-35 .

Figure 28 shows plots of measurements of performance as a function of frequency. Solid line 2801 shows a measure of actual performance (eg, load reflection coefficient magnitude) as a function of frequency that would result if the main power signal were tuned to each frequency and measured. Dashed line 2802 shows a measure of estimated performance obtained using one or more secondary power signals while the primary power signal is kept at a fixed frequency (eg, f1 ).

As discussed, the power level of the primary frequency affects measures of performance (e.g., load reflection coefficient); therefore, measures of performance estimated using a low-level power signal will differ from performance at higher powers of the primary signal. Measure. But as discussed further herein, the low level signal enables close estimation of the desired dominant frequency (eg, which results in low reflection coefficient and/or low instability). The frequency of the main signal can then be fine-tuned at higher power levels without testing for frequencies that would cause the plasma to go out.

Figure 29 depicts an aspect where an initial dominant frequency can be applied between f1 and fa, and how the frequency tuning algorithm (which relies on sweeping and testing the frequency of the dominant power) can be at a local optimum of the measure of performance can be intercepted without the information provided by the low power secondary signal. More specifically, the tuning algorithm may tune the primary frequency towards an optimal frequency considered to be at f1 . In particular, Figure 29A shows a measure of performance (e.g., reflection coefficient) as a function of frequency; the solid line of Figure 29B shows how an algorithm using only the main power can adjust the frequency of the main power signal to minimize the measure of performance and FIG. 29C shows the frequency spectrum (power per bandwidth, eg, watts per 3 kHz bandwidth) of the power generation system output 2220, 2320, 2420, 2520, 2620, or 2720 at time t2 in FIG. 29B. As shown by the dotted line in Figure 29B, the low level secondary signal can be used to identify the best frequency globally.

But as shown by the solid line, after reaching the local optimum at f1, if the dominant frequency is used to search for the global optimum, these attempts may result in the application of power around the frequency fa, as can be seen in Fig. 30A and that seen in Figure 30B results in the extinction of the plasma. Figure 30A shows a measure of performance as a function of frequency. The solid line in Figure 30A shows the measure of the efficiency by lighting the plasma, and the dashed line shows the measure of the efficiency of extinguishing the plasma. Fig. 30B shows how a global search using the main power signal may result in quenching plasma, since it may not be possible to deliver enough power near fa to sustain the plasma. Figure 30C shows the frequency spectrum of the power generation system output at time t2 in Figure 30B.

The fact is that one or more secondary power signals can be used to search for the global optimum, as shown in Figure 31 (showing one secondary power signal) and Figure 32 (showing multiple secondary power signals), while the primary power signal Stay at a fixed frequency (eg, at or close to a local optimum). In FIG. 31 , frequency tuning using a secondary power signal in the form of a low level signal at a single secondary frequency applied in a particular order is shown. 32 shows frequency tuning using a secondary power signal in the form of a low level signal with spectral components at multiple secondary frequencies adjusted according to an algorithm to optimize a measure of performance.

As shown, one or more secondary power signals may be applied at a power level substantially lower than that of the primary power signal, and may be applied at one or more secondary frequencies. The secondary frequencies may be fixed frequencies with equal or unequal spacing, or may be variable frequencies as shown in FIG. 32 . Additionally, primary and secondary power signals can be applied simultaneously.

The secondary signals (probing frequencies) can be applied continuously, as shown in FIG. 31 , or only when searching for global optimum, as shown in FIG. 32 . In addition, although a single characteristic is shown in the graphs of FIGS. 28-33 , in other embodiments multiple characteristics can be monitored simultaneously, such as the magnitude of the load reflection coefficient in conjunction with electrical current measured via (e.g., fluctuations in load impedance). slurry stability, and analysis of all monitored properties (or a plurality of monitored properties) can be used to identify global optimum values. In this way, a global optimum is identified without applying the full power of the primary signal around fa or any frequency that might extinguish the plasma.

In some modes of operation, the amplitude of one or more secondary power signals applied at one or more secondary probing frequencies is so small that it can be considered negligible compared to the primary power signal and therefore has no effect on Plasma has no significant effect. In other applications, the amplitude of one or more secondary power signals may be important compared to the primary power signal if the goal is simply not to extinguish the plasma while searching for a global optimum. In this case, care must be taken not to exceed the voltage and current ratings of the plasma system due to the high resulting amplitude at the beat frequency.

Figure 31 shows an embodiment where a single secondary detection frequency is swept continuously across the frequency range. The range over which the secondary sounding frequency is swept will typically be the frequency range over which the power generation system is expected to operate (eg, 12.882 to 14.238 MHz), but this is not necessarily the case. Examples where other frequency ranges may be considered include when the secondary power signal is used to extract information about plasma conditions by, for example, analyzing mixing and intermodulation products. In other cases, as shown in FIG. 32 , one or more secondary probing frequencies may be adjusted according to an algorithm to find an optimal frequency rather than sweeping in a predetermined pattern as shown in FIG. 31 . As also shown in FIG. 32 , once the global optimum has been identified, the secondary power signal may be turned off rather than continuously applied as shown in FIG. 31 .

As shown in Figures 31A and 32A, the estimate of the optimum frequency using one or more secondary power signals may not correspond exactly to the true optimum. Typically, this difference will arise from the non-linear nature of the plasma loading. As shown in Figures 3 IB and 32B, after determining the optimum frequency using the secondary power signal, the primary frequency can be adjusted to further optimize performance. Figures 31C and 32C depict the spectral components of the primary and secondary sounding frequencies of Figures 31B and 31C, respectively.

33A-33C show the case where the secondary power signal is noise. Figure 33C shows the frequency spectrum of the power generation system output at time t2 in Figure 33B. Noise can be inherent to the main power signal or can be added to the power generation system output (see, eg, Figures 26 and 27). Figure 33B shows the noise power as a function of time, assuming that the noise is added to the output of the generating system.

Once the global optimum has been identified, the main power signal may be adjusted or switched to (or towards) the frequency corresponding to the global optimum without requiring the main power signal to pass through regions of the spectrum where plasma may be suppressed (e.g., near fa ). For example, in FIG. 14, the amplitude of the primary power signal is ramped down, while the amplitude of the secondary frequency is ramped up at the global optimum. In this way, the primary power signal and the secondary power signal swap places. Figure 35 shows another variation of switching the primary frequency towards the global optimum, where the frequency of the primary power signal changes abruptly to the identified global optimum frequency.

In some embodiments, the identified global best frequency may be selected from one of the secondary frequencies, but this is not required. For example, the identified global best frequency may be between two of the two or more secondary frequencies. For example, interpolation between secondary frequencies can be used to identify the identified overall best frequency.

FIG. 36 illustrates a method for frequency tuning a power generation system to find the global optimum of a measure of performance using a secondary sounding signal to steer the global optimum. Method 3600 applies (eg, by high frequency generator 102 ) a primary power signal to a plasma system (eg, matching network 2104 connected to plasma chamber 2108 ) primarily at a primary frequency (block 3602 ). Simultaneously, the method 3600 applies a low level signal to the plasmonic system at one or more of the secondary detection frequencies or a continuum (eg, as in the case of noise) (block 3604 ).

The low level signal can be periodic or the sum of periodic signals, it can be noise inherent in the main power signal, or it can be noise added to the main power signal. The one or more secondary frequencies may be equally spaced in frequency or may have varying intervals. One or more secondary frequencies may be applied together or at different times, and may be adjusted over time. One or more secondary frequencies may be swept across a fixed frequency range. Alternatively, one or more secondary frequencies can be adjusted via feedback to detect and steer to a global optimum. One or more secondary or continuum frequencies of the secondary frequencies can be applied all the time or only when required.

Method 3600 monitors a characteristic as a measure of frequency-dependent performance (e.g., load reflection coefficient magnitude), specifically at one or more of the secondary frequencies or a continuum and/or at the primary frequency and/or Alternatively, the characteristic is monitored for expected mixing and intermodulation products of primary and secondary frequencies (block 3606 ). As shown, the mixing product of the dominant frequency and any low frequencies from any low frequency generator (eg, low frequency generator 108 ) is suppressed (eg, by filter 104 ) (block 3607 ). The method 3600 then identifies the best frequency corresponding to the global best value of the characteristic (block 3608). This can be done via minimization and maximization algorithms familiar to those of ordinary skill in the art. Finally, the method 3600 adjusts the primary frequency of the primary power signal to the optimal frequency identified in the identification operation (block 3610 ). This adjustment can be done in a number of ways. For example, this adjustment may be necessary to avoid applying major power only in regions where the reflected power is close to 100% (e.g., near fa in FIG. The slurry system is maintained by another power source 2150 or 2151). Thus, the primary power signal can be switched to an optimal frequency, or the power levels of the primary and secondary power signals can be gradually reversed such that the power signals are placed in reverse, just to name two non-limiting examples.

In some embodiments, method 3600 ends when the primary power signal has moved to a frequency identified as the global optimum using one or more secondary power signals. But in other cases, the method 3600 can be iterated to further refine the optimization or to account for global variations due to, for example, the nonlinear nature of the plasma load or parameters that can change over time (e.g., plasma chamber gas pressure). Variation of the best value.

The identification of the best frequencies (block 3608) can happen instantaneously because the samples are obtained from monitoring (block 3606), or the analysis can happen after the frequency range has been sampled. The movement of the primary frequency (block 3610) may only occur after a global optimum has been identified (block 3608), or it may occur after a frequency better than the current primary frequency has been identified.

Methods of monitoring properties using secondary power signals may also be used for the purpose of identifying plasma properties or changes in plasma properties. Instead of identifying an optimal frequency and adjusting the primary frequency toward the identified global optimum, outputting or monitoring characteristics (block 3608 ) may be used to identify plasma characteristics or changes in plasma characteristics. Monitoring mixing and intermodulation products can be used to monitor the nonlinear behavior of the plasma or simply to detect whether the plasma is lit. Instead of looking at specific mixing and intermodulation production, higher order statistics (eg, bispectrum) can be used to identify plasmonic properties or changes in plasmonic properties.

FIG. 37 shows three exemplary implementations of sensors, such as sensors 2112 or 2412 . The sensor can be, for example, a directional coupler 3710 as shown in FIG. 37A or a voltage and current (VI) sensor as shown in FIG. 37B , and either implementation can include a directional coupler 3710 as shown in FIG. 37C filter 3730 and analog-to-digital converter 3720 .

Figure 38 shows an exemplary implementation of a global best value identification module (eg, 2116 or 2418). Part of the functionality shown in Figure 38 may also be part of a sensor. Figure 38 shows an implementation using multiple demodulators 3810 that allow simultaneous processing of multiple frequency components. Signals 3820 (labeled A) and 3830 (labeled B) may be, for example, forward and reflected power or voltage and current or some other related measurement. After multiplication 3850 and filtering 3840 by cosine and sine functions, the complex vector representations of A and B at different frequencies labeled A ₁ , B ₁ to _AN , B _N are used to calculate power at multiple frequencies and load reflection coefficient. Typically, one channel will be reserved for the primary frequency. Other channels can be set to one or more secondary frequencies or desired mixing and intermodulation products. As mentioned before, this is only one implementation, and many other implementations are possible using, for example, discrete Fourier transforms rather than dedicated demodulation channels.

The depicted configurations of the components shown in FIGS. 21-27 are logical, the connections between the various components are illustrative only, and the depictions of these embodiments are not meant to be actual hardware diagrams; therefore, the components may Combined or further separated in practical implementations, and components can be connected in various ways without changing the basic operation of the system.

Instead of a single secondary power supply, as seen in Figures 24-27, two, three, four or more secondary power supplies can be used to generate two or more secondary power signals.

For purposes of this disclosure, the secondary power signal may be periodic, such as an RF signal. However, in other embodiments, a non-periodic power signal (eg, noise) may be used.

Although the disclosure has repeatedly shown tuning to local and global minima, those of ordinary skill in the art will appreciate that tuning to local and global maxima is also contemplated and that the disclosure can be readily applied to monitored A characteristic where the dominant frequency of the delivered power is optimized for the global maximum of the monitored characteristic. Furthermore, there is no need to perform the frequency tuning described herein to achieve local or global maxima/minima. In fact, applications where it is beneficial to achieve a detuned frequency may be preferable in some cases, such as where the frequency of the high frequency generator (e.g., high frequency generator 102) to achieve a stable plasma is better than the minimum position of the reflected power standard case.

Those of ordinary skill in the art would understand that information and signals may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or magnetic particles, optical fields or optical particles, or any combination thereof.

Those of ordinary skill in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of both . A general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic designed to perform the functions described herein devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof to implement or execute the various illustrative logic blocks, modules, and circuits described in connection with the embodiments disclosed herein. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, eg, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in combination with a DSP core, or any other such configuration.

The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be directly embodied in hardware, in software modules executed by a processor, or in a combination of both. Software modules can reside in non-transitory memory, which includes RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, scratchpad, hard disk, removable Disk, CD-ROM, or any other form of storage media known in the art. An exemplary storage medium is coupled to the processor such that the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral with the processor. The processor and storage medium can reside in the ASIC. The ASIC may reside in a user terminal. In the alternative, the processor and storage medium may reside as discrete components in the user terminal.

Referring to FIG. 39 , there is shown an example of a computing system 3900 that may be used in conjunction with embodiments disclosed herein. As shown, display 3912 and non-volatile memory 3920 are coupled to bus 3922, which is also coupled to random access memory ("RAM") 3924, processing section (which includes N processing elements) 3926 , a field programmable gate array (FPGA) 3927 and a transceiver assembly 3928 comprising N transceivers. Although the components depicted in FIG. 39 represent physical components, FIG. 39 is not intended to be a detailed hardware diagram; thus, many of the components depicted in FIG. 39 may be implemented with common configurations or dispersed among additional physical components. In addition, it is contemplated that other existing and yet to be developed physical components and architectures may be used to implement the functional components described with reference to FIG. 39 .

This display 3912 is generally used to provide a user interface to the user, and in some embodiments, the display 3912 is implemented by a touch screen display. In general, non-volatile memory 3920 is memory used for storing (e.g., persistent storage) data and machine-readable (e.g., processor-executable) code (including executable code associated with implementing the methods described herein) Executing program code) in non-transitory memory. For example, in some embodiments, non-volatile memory 3920 includes boot loader code, operating system code, file system code, and non-transitory processor executable code to facilitate the methods described herein (including but not limited to the methods described with reference to the flowcharts of FIG. 7 , FIG. 16 to FIG. 20 and FIG. 36 ).

In many implementations, non-volatile memory 3920 is implemented by flash memory (eg, NAND or ONENAND memory), although it is contemplated that other memory types may be utilized. While it is possible to execute code from non-volatile memory 3920, executable code from non-volatile memory is typically loaded into RAM 3924 and processed by one of the N processing elements in processing section 3926 One or more to execute the executable code. Non-volatile memory 3920 or RAM 3924 may be used to store the frequency of the global optimum as described in FIGS. 28-34.

In operation, N processing elements in combination with RAM 3924 are generally available to execute instructions stored in non-volatile memory 3920 to implement aspects of broadband measurement systems 116, 120, frequency tuning subsystem 1525, global optimum Identify module 2116 , frequency control module 2118 , circuit 2114 and control the aspect of high frequency generator 102 , 1502 (eg, frequency tuning aspect), power supply 2110 and matching 106 . For example, non-transitory processor-executable instructions for implementing aspects of the methods described with reference to FIGS. N processing components execute in conjunction with RAM 3924 . As will be appreciated by those of ordinary skill in the art, the processing portion 3926 may include a video processor, a digital signal processor (DSP), a graphics processing unit (GPU), and other processing components.

Additionally or in the alternative, Field Programmable Gate Array (FPGA) 3927 can be configured to implement one or more aspects of the methods described herein (eg, see FIGS. 7, 16-20 and method described in Figure 36). For example, non-transitory FPGA configuration instructions may be persistently stored in non-volatile memory 3920 and accessed by FPGA 3927 (e.g., during startup) to configure FPGA 3927 to implement broadband measurement systems 116, 120 and controls the aspect (eg, frequency tuning aspect) and matching 106 of the high frequency generator 102 .

An input component can be used to receive a signal (eg, from a sensor 114, 118, 1520, 2112, 2312, 2412, 2413) indicative of one or more aspects of power. Signals received at the input component may include, for example, voltage, current, forward power, reflected power, and plasma load impedance. Output components are typically used to provide one or more analog or digital signals (eg, a frequency control signal on frequency control line 1530 ) to achieve the operational states of generators 102 , 108 , matching 106 and/or broadband measurement systems 116 , 120 Sample. For example, the output section may provide control signals utilized by generators 102 , 108 , matching 106 and/or oscillators and power amplifiers of broadband measurement systems 116 , 120 .

The depicted transceiver component 3928 includes N transceiver chains that can be used to communicate with external devices over a wireless or wired network. Each of the N transceiver chains may represent a transceiver associated with a particular communication scheme (eg, WiFi, Ethernet, Profibus, etc.).

The previous description of the disclosed embodiments is provided to enable any person of ordinary skill in the art to make or use the invention. It should be appreciated that the various depicted embodiments are not intended to be stand-alone embodiments. Rather, several of the embodiments depicted herein should be examined to convey aspects that can be combined. For example, the probing signal and signal detection techniques described with reference to FIGS. 21-38 may be used in conjunction with the frequency tuning algorithms described with reference to FIGS. 1-20. As another example, the filters 104, 904, 1404 and broadband measurement systems 116, 118 may be used in conjunction with the embodiments described with reference to FIGS. 21-38. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

100: Plasma chamber 102:High frequency generator/generator 104: filter 106:Matching Network/Matching 108:Low frequency generator/generator 110: match 112: delay element 114: sensor 116:Broadband measurement components/broadband measurement system 118: sensor 120:Broadband measurement components/broadband measurement system 122: connection 124: connection 700: Flowchart 710: block 720: block 730: block 740: block 750: block 804B: filter 810: input capacitor 820: output capacitor 904: Water-cooled filter/filter 910: Water connection 920: water connection 930: input connector 1020: hollow helical coil / helical coil 1022: ground terminal 1024: copper block 1026: copper strap 1028: ceramic insulator/ceramic 1030: metallization 1032: potted cylindrical shell 1110: input connector/connector 1120: ceramic insulator 1130: hollow spiral coil 1140: output connector/connector 1150: ceramic insulator 1160: hollow spiral coil 1210: waterway 1220: ceramic insulator 1230: strap 1240: copper block 1250: rear metallization 1260: ceramic insulator 1270: strap 1280: front metallization 1310: insulation bracket 1404: filter 1410: cup body 1420: tuning pellets 1430: cup body 1440: tuning pellets 1502: High frequency generator 1505: exciter 1510: Power Amplifier 1515: filter 1520: sensor 1525: Frequency Tuning Subsystem 1530: frequency control line 1600: Flowchart 1610: block 1620: block 1630: block 1640: block 1650: block 2100: Power generation system 2104: Matching network 2105: Matching network 2106: Plasma 2108: Plasma chamber 2110: Power 2111: output 2112: sensor 2114: circuit 2116: Global Best Value Identification Module 2118: Frequency Control Module 2122: filter 2130: Connection/supply cable/supply connection/cable 2131: connect 2140: External power 2150: RF or DC generator/generator 2151: RF or DC generator/generator 2200: Power generation system 2204: Matching network 2208: Plasma chamber 2210: power supply 2214: circuit 2220: Output/Generation System Output 2300: Power generation system 2304: Matching network 2306: Plasma 2308: Plasma chamber 2310: power supply 2312: sensor 2314: circuit 2316: frequency control module 2318: Global best value identification module 2320: output 2322: filter 2400: Power generation system 2404: Matching network 2406: Plasma 2408: Plasma chamber 2410: power supply 2411: Low level signal source 2412: sensor 2413: sensor 2414: circuit 2416: frequency control module 2418: Global best value identification module 2420: Power generation system output 2422: filter 2423: filter 2424: Combiner 2500: power generation system 2510: Main frequency source 2511: Secondary frequency source 2524: Combiner 2550: power amplifier 2600: Power generation system 2610: power supply 2613: noise source 2620: Output/Generation System Output 2624: Combiner 2700: Power generation system 2710: Main frequency source 2713: noise source 2724: Combiner 2750: Power Amplifier 2801: solid line 2802: dotted line 3600: method 3602: block 3604: block 3606: block 3607: block 3608: block 3610: block 3710: Directional coupler 3720: Analog to Digital Converter 3730: filter 3810: demodulator 3820:Signal 3830:Signal 3840: filter 3850: multiplication 3900: Computing Systems 3912:Display 3920: non-volatile memory 3922: busbar 3924: Random Access Memory/RAM 3926: processing part 3927: Field Programmable Gate Array/FPGA 3928: Transceiver Components T: duration/maximum time slot

[Figure 1] is a block diagram depicting the power supply system and plasma treatment system; [Fig. 2] is a graph depicting how power can be perceived by measuring power using different measurement system filter bandwidths; [FIG. 3A] and [FIG. 3B] are graphs depicting the modulation of the load reflection coefficient; [FIG. 3C] is a graph depicting the resulting reflected power seen by the high frequency generator when the filter depicted in FIG. 1 is not utilized; [FIG. 4A] Contains two graphs depicting performance profiles for an exemplary design of the filter depicted in FIG. 1; [ FIG. 4B ] is a graph depicting the net power that can be delivered to the plasma load by the high frequency generator at the fundamental and mixed product frequencies when the filter depicted in FIG. 1 is not utilized; [FIG. 5A] and [FIG. 5B] are graphs depicting the modulation of the load reflection coefficient; [FIG. 5C] is a graph depicting the resulting reflected power seen by the high frequency generator when the filter depicted in FIG. 1 is not utilized; [FIG. 6A] and [FIG. 6B] are graphs depicting the modulation of the load reflection coefficient; [FIG. 6C] is a graph depicting the resulting reflected power seen by the filter depicted in FIG. 1; [ FIG. 7 ] is a flowchart depicting a method that can be studied in detail in conjunction with the embodiments disclosed herein; [ FIG. 8A ] and [ FIG. 8B ] are diagrams depicting an equivalent circuit of an embodiment of the filter described with reference to FIG. 1 ; [FIG. 9] is a perspective view of an exemplary water-cooled filter design with two parallel spiral resonators; [FIG. 10] is a view of the interior of a water-cooled filter design with two parallel spiral resonators; [FIG. 11] is a cross-sectional view of a water-cooled filter design with two parallel spiral resonators; [FIG. 12] is a detailed view of the capacitor block of a water-cooled filter design with two parallel spiral resonators; [FIG. 13] is an exploded view of a water-cooled filter design with two parallel spiral resonators; [FIG. 14] is a view of a filter including means for tuning the filter; [FIG. 15] is a block diagram depicting an exemplary high frequency generator; [ FIG. 16 ] is a flowchart depicting a method that can be studied in detail in conjunction with the embodiments disclosed herein; [FIG. 17] is a graph depicting multi-state waveforms that can be applied by the high-frequency generators of FIGS. 1 and 15; [FIG. 18] is a graph depicting an exemplary aspect of a frequency tuning method; [FIG. 19] is a graph depicting additional aspects of the frequency tuning method described with reference to FIG. 18; [FIG. 20] is a graph depicting the operation state of another frequency tuning method; [FIG. 21] depicts a power generation system configured for automatic frequency tuning of power delivered to a plasma load; [ FIG. 22 ] depicts an embodiment of a power generation system, wherein sensors reside within the power generation system along with a power source and one or more circuits; [FIG. 23] depicts an embodiment of the power generation system, wherein the sensors reside outside the power generation system; [Figure 24] shows another embodiment of the power generation system; [FIG. 25] An embodiment of a power generation system is shown, wherein the primary power signal and the secondary power signal are combined before being amplified by a power amplifier; [FIG. 26] depicts an embodiment of a power generation system in which a power source generates a primary power signal and a noise source generates a secondary power signal in the form of noise; [ FIG. 27 ] shows an embodiment of a power generation system, wherein the secondary signal is noise and the primary power signal and the secondary power signal are combined before being amplified by a power amplifier; [FIG. 28] Curves showing measurements of performance as a function of frequency; [FIG. 29A] is a graph depicting a measure of performance (eg, reflection coefficient) as a function of frequency; [FIG. 29B] is a graphical representation depicting how the main power signal frequency may be adjusted to minimize the measure of efficacy depicted in FIG. 29A; [FIG. 29C] Spectrum (power/span, e.g., Watts per 3 kHz bandwidth) of power generation system output depicted in FIG. 29B at time t2; [FIG. 30A] is a graph depicting the measure of performance versus frequency; [FIG. 30B] is a graph depicting how a global search using the main power signal can produce quenched plasma; [FIG. 30C] is a graph showing the frequency spectrum of the power generation system output at time t2 in FIG. 10B; [FIG. 31A] is a graph depicting the estimation of the optimal frequency using the secondary power signal; [FIG. 31B] is a graph depicting the adjustment of the primary frequency after determining the desired frequency using the secondary power signal; [FIG. 31C] is a graph showing the spectral components of the power at the main signal and the secondary signal of FIG. 31B; [FIG. 32A] is a graph depicting the estimation of the optimal frequency using the secondary power signal; [FIG. 32B] is a graph depicting the adjustment of the primary frequency after determining the desired frequency using the secondary power signal; [FIG. 32C] is a graph depicting the spectral components of the power at the primary signal and the secondary signal of FIG. 32B; [FIG. 33A] is a graph depicting the estimation of the optimal frequency using the secondary power signal; [FIG. 33B] is a graph depicting the power of noise as a function of time, where noise is added to the output of the power generation system; [FIG. 33C] is a graph depicting the frequency spectrum of the power generation system output at time t2 shown in FIG. 33B; [FIG. 34A] is a graph depicting aspects of a method for frequency tuning; [FIG. 34B] is a graph depicting additional aspects of the method for frequency tuning shown in FIG. 34A; [ FIG. 34C ] is a graph depicting other aspects of the method for frequency tuning depicted in FIGS. 34A and 34B ; [FIG. 34D] is a graph depicting yet additional aspects of the method for frequency tuning depicted in FIGS. 34A, 34B, and 34C; [FIG. 35A] is a graph depicting aspects of a method for frequency tuning; [FIG. 35B] is a graph depicting additional aspects of the method for frequency tuning depicted in FIG. 35A; [FIG. 35C] is a graph depicting other aspects of the method for frequency tuning depicted in FIG. 35A and FIG. 35B; [ FIG. 36 ] Illustrates a method for frequency tuning a power generation system that can be studied in detail in conjunction with the embodiments described herein. [FIG. 37A] is a diagram depicting an exemplary sensor. [FIG. 37B] is a diagram depicting another embodiment of a sensor. [FIG. 37C] is a diagram depicting yet another embodiment of a sensor. [FIG. 38] is a diagram depicting the aspect of an exemplary recognition module. [ FIG. 39 ] is a block diagram depicting components that may be used to implement embodiments disclosed herein.

3600: method

3602: block

3604: block

3606: block

3607: block

3608: block

3610: block

Claims (20)

A power generation system comprising: a high frequency generator configured to apply power to a plasma chamber at a primary frequency; a filter configured to suppress mixing products to limit variations presented to a time-varying load reflection coefficient of the high frequency generator; and a frequency tuning subsystem configured to: while the high frequency generator is applying power at the primary frequency, applying a probe signal comprising one or more probe frequencies, wherein the probe signal has a power lower than that generated at the primary frequency; and The primary frequency of the high frequency generator is adjusted in response to the one or more probe frequencies indicative of an improved measure of performance. The power generation system of claim 1, comprising a low frequency generator to apply power to the plasma chamber at a low frequency, wherein the filter is configured to suppress a mixture product of the dominant frequency and the low frequency . The power generation system according to claim 2, wherein a frequency ratio of the low frequency generator and the high frequency generator is between 0.0005 and 0.2. The power generation system as claimed in claim 1, wherein the power generated at the detection frequencies is 1 dB to 100 dB lower than the power at the main frequency. The power generation system of claim 1, comprising a delay element configured to be coupled between the filter and the plasma chamber. The power generation system of claim 1, wherein the measure of performance is a measure of performance selected from the group consisting of: a reflected power; a deviation of a load impedance seen by the high frequency generator from a desired impedance a measure of the degree; and a measure of the magnitude of the load reflection coefficient. The power generation system according to claim 1, wherein the one or more detection frequencies are noises. The power generation system according to claim 7, which includes a single oscillator-amplifier combination to generate the main power signal, and the noise is inherent to the single oscillator-amplifier combination. The power generation system according to claim 1, comprising a primary oscillator for generating a primary power signal and a secondary oscillator for generating the one or more detection frequencies. A method for automatic frequency tuning of a power generation system, the method comprising: applying a primary power signal to a plasma load at a primary frequency by a high frequency generator directly or via a matching network; applying a probe signal to the plasma load at one or more probe frequencies, wherein the power generated at the probe frequencies is lower than the power generated at the primary frequency; suppressing the mixing products by a filter to reduce variations in a time-varying load reflection coefficient present to the high frequency generator; and The primary frequency is adjusted based on the measure of performance in response to the probe signal. The method of claim 10, wherein the application of the detection signal and the adjustment of the main frequency are performed cyclically so as to repeatedly improve the accuracy of adjusting the main frequency to a global best value. The method of claim 10, wherein the detection signal is periodic or a sum of one of periodic signals. The method of claim 10, wherein applying the probing signal comprises sweeping the one or more probing frequencies across a fixed frequency range. The method of claim 10, wherein applying the detection signal comprises tuning a detection signal to a single one of a plurality of different detection frequencies at different times. A plasma treatment system comprising: a plasma chamber; a high frequency generator configured to apply power to the plasma chamber at a primary frequency; a low frequency generator for applying power to the plasma chamber at a low frequency; a filter configured to suppress the mixing product of the dominant frequency and the low frequency to limit changes in a time-varying load reflection coefficient presented to the high frequency generator; and Means for frequency tuning the high frequency generator using a probe signal applied simultaneously with power applied to the plasma chamber at the primary frequency. The plasma processing system according to claim 15, comprising a matching network coupled between the high frequency generator and the plasma chamber. The plasma processing system according to claim 15, wherein a frequency ratio of the low frequency generator to the high frequency generator is between 0.0005 and 0.2. The plasma processing system of claim 15, wherein the power generated at one or more detection frequencies of the detection signal is 1 dB to 100 dB lower than the power at the main frequency. The plasma processing system according to claim 15, wherein the means for performing frequency tuning comprises means for performing frequency tuning by using the probe signal containing noise. The plasma processing system of claim 15, wherein the means for frequency tuning includes a secondary oscillator, and wherein the high frequency generator includes a primary oscillator for generating the primary frequency and for generating the probe signal of the secondary oscillator.

Images