Device for high speed sensing of RF signals from RF plasma processing equipment

Abstract

This teaching relates to a sensing device for monitoring electromagnetic radiation emitted from a plasma processing system, the sensing device comprising at least two of the following: (i) a first probe for detecting a time-varying RF electric field, (ii) a second probe for detecting a time-varying RF magnetic field, and (iii) an optical probe for detecting modulated light emission; and the sensing device further comprising a signal processing unit configured to receive a signal from each probe and monitor the electromagnetic radiation at only a single frequency relative to each signal.

Description

[Technical Field]

[0001] This invention is generally related to the analysis of alternating electromagnetic fields from plasma systems operating at RF frequencies. [Previous Technology]

[0002] Plasma processing of materials is ubiquitous in modern industrial manufacturing. A common example is the etching and deposition of layers to form transistors during the manufacturing of integrated circuits in the semiconductor industry. Plasma processing is also used in the manufacture of solar panels, flat panel displays, thin film coatings, and medical devices.

[0003] Plasma is typically formed within a vacuum chamber. Once the air is purged, a gas formulation is added to the chamber at a selected pressure. Energy, typically electrical energy, is supplied to the vacuum chamber to excite the gas into a plasma state. The plasma state provides the ions needed to modify the workpiece surface.

[0004] Electrical energy in the radio frequency (RF) band is typically used to power plasma reactors. The RF range is typically between tens of kilohertz and hundreds of megahertz. Radio waves are coupled from the RF generator to the plasma chamber via an RF power delivery subsystem, which includes a matching network to maximize power delivery. Power can be coupled to the plasma in several different ways. In one configuration, an RF power supply electrode can be used to excite the plasma via an electric field (E-field) formed between the power supply electrode and the counter electrode. The counter electrode can be another electrode or the chamber body, typically kept at ground potential. When the E-field strength is sufficient, electrical breakdown occurs, and plasma forms. An RF current sustains the discharge and flows between the power supply electrode and ground. Electrons stripped from the parent atoms and molecules oscillate back and forth in the RF E-field, ionizing the background gas in the process, thus sustaining the plasma.

[0005] In another configuration, RF power is coupled to the plasma via an RF antenna. The antenna does not need to be in direct contact with the background gas. The RF current flowing through the antenna induces a time-varying magnetic field (H-field), which is perpendicular to the direction of current flow. The H-field is typically coupled to the chamber via a dielectric window. Once breakdown occurs, the H-field induces an E-field in the plasma, thereby driving the RF current. Free electrons oscillate in the RF field, ionizing the background gas and thus sustaining the plasma. Many other plasma reactor configurations and plasma generation mechanisms exist.

[0006] Regardless of the type of RF coupling mechanism, an interface region inevitably exists between the electrode or antenna and the plasma. This region is called the plasma sheath. This sheath has a nonlinear RF impedance. Therefore, harmonics at the fundamental drive frequency are excited. As a result, the plasma voltage and current can have a rich harmonic spectrum. The RF harmonic characteristics of a plasma process are determined by many variables, including fundamental plasma parameters, plasma chemistry, chamber geometry, chamber surface conditions, and chamber mechanical properties. Therefore, the harmonic spectrum contains important information about the health and performance of the plasma process.

[0007] When plasma electrons oscillate in an RF field, they ionize and / or excite neutral gas particles. Electron collision excitation events are accompanied by light emission. Most of the emitted light comes from the visible region of the electromagnetic spectrum. Depending on the type of gas used, light emission occurs at a specific wavelength. In RF plasma, the intensity of light emission is modulated by electron motion using the RF drive frequency and its higher harmonics.

[0008] In some plasma reactor configurations, there are one or more drive frequencies. This generates a harmonic spectrum for each fundamental drive frequency and intermodulation frequency. There is a great need for a sensing device to accurately measure the characteristics of the spectrum generated in an RF-powered plasma reactor. Accurately measured frequency / harmonic spectrum characteristics can be used for real-time monitoring of process performance.

[0009] Due to their design, plasma processing systems typically "leak" electromagnetic fields. RF time-varying electric and magnetic field radiation can penetrate any unshielded or improperly grounded area. Optical emission radiation can penetrate any opaque area, such as an observation port. This provides three methods for sensing the RF spectrum radiated from a plasma processing system using the following: a) a time-varying E-field sensor, b) a time-varying magnetic field (point B) sensor, and c) a time-varying light intensity sensor.

[0010] RF spectra are typically represented as a graph of signal amplitude versus frequency. The signal amplitude can correspond to the time-varying E-field or B-field intensity. The time-varying optical signal amplitude will closely correspond to the B-field intensity, as both are driven by the movement of electrons within the plasma system. These three signals are actually vector quantities with phase components and amplitudes. The phase angle between the fundamental frequency and harmonic frequencies of a single spectrum is an important set of data to be measured, as it is extremely sensitive to changes in the movement of electrons within the plasma. Furthermore, the phase relationship between corresponding frequency components of the E-field and B-field spectra has been found to be highly sensitive to certain physical phenomena in the plasma process. By processing and analyzing the sensed RF spectra, the plasma process can be monitored in real time. To avoid the need for detailed calibration of the measured RF spectra, statistical methods can be used to determine a baseline or fingerprint of certain process conditions. This is usually performed for known "healthy" processes. For the same setpoint, subsequent processes can be compared to the baseline to check for statistically significant changes. Significant changes can be easily detected in a single data channel; for example, a sudden decrease or increase in the E-field amplitude may indicate an RF generator problem. Other changes may be more subtle and may require multivariate analysis using many data channels from the RF spectrum to detect, such as minor air leaks or slight wafer misalignments—though these problems can be detrimental to the workpiece. Furthermore, some changes may only be detected in higher-level data channels, such as the phase measurements provided by this invention. For example, phase measurements may be highly sensitive to certain process endpoints, such as when a layer etching is complete—even if that layer occupies less than one percent of the workpiece surface area.

[0011] E-field and B-point probes are commonly used in so-called VI sensors to measure voltage and current flowing in transmission lines. A VI sensor device is described in publication no. WO2014016357A2. This sensor is designed to be connected in series with an RF power feed line. For this purpose, it includes a section of transmission line, as is typical for this type of sensor. A wideband capacitance pickup (E-field probe) is used to determine the voltage signal on the RF line connected to the plasma. An induction loop (B-point probe) is used to determine the RF current in the RF line connected to the plasma. The voltage and current pickups are embedded in the RF transmission line section of the VI sensor structure. The signals representing current and voltage are passed to an analog-to-digital converter (ADC) and digitized in a field-programmable gate array (FPGA). While online VI sensors are very important tools, they are difficult to install without significant modifications to the plasma system configuration.

[0012] In Publication No. WO2018177965A1, McNally et al. describe a device in which a specially designed magnetic loop antenna senses plasma current flowing near an observation aperture in a plasma chamber from an external location. The antenna is carefully designed and calibrated. The antenna output is coupled to a spectrum analyzer to view the spectrum detected by the antenna. The inventors describe a frequency analysis technique for detecting the resonant characteristics of plasma. Operating the antenna in an isolated near-field environment is crucial for its proper functioning. Shielding is used to prevent the antenna from detecting signals from a matching network and other far-field signal sources.

[0013] In Publication No. WO2004006298A2, Parsons describes an invention that uses an RF antenna to sense RF radiation from a plasma system at a distance. The antenna can detect harmonic signals and is coupled to a processing unit for analysis. The processing unit is coupled to a plasma tool controller, in which the sensed RF signals are used to adjust and maintain parameters of the plasma process based on the measured signal level.

[0014] In US Patent No. 2007022766, Yamazawa et al. described a device consisting of two magnetic loop antennas located inside the plane of a plasma chamber wall. The antennas are placed near the two electrodes of a capacitively coupled plasma reactor. A voltage signal generated by the magnetic flux passing through each loop is coupled to a signal processing unit. The current flowing from the plasma to the chamber wall is then calculated.

[0015] In US Patent No. 6,441,620 B1, Scanlan et al. describe a method for identifying faults in a plasma process using data from online VI sensors. For a given baseline plasma process, the changes in the values ​​of a plurality of Fourier components relative to the baseline are determined, these changes being due to variations in a plurality of process input parameters. These changes in values ​​are stored as reference data. During any subsequent production operation, the plasma process is monitored for faults, and if a fault is detected, the baseline process is repeated using the same input parameter values ​​nominally labeled as the original baseline values. The changes in the Fourier components relative to the original baseline values ​​are determined and these changes are compared with the reference data. This comparison is used to determine which plasma subsystem is most likely to have caused the fault.

[0016] As can be seen from the above, the prior art has many shortcomings. These shortcomings need to be addressed. [Summary of the Invention]

[0017] This document discloses a sensing device for monitoring electromagnetic radiation emitted from a plasma processing system, the sensing device comprising at least two of the following: (i) a first probe for detecting a time-varying RF electric field; (ii) a second probe for detecting a time-varying RF magnetic field; and (iii) an optical probe for detecting modulated light emission, and the sensing device further comprising a signal processing unit configured to receive a signal from each probe and monitor the electromagnetic radiation at only a single frequency relative to each signal.

[0018] The signal processing unit can be further configured to determine the amplitude of each signal from each probe to generate amplitude data.

[0019] The signal processing unit can be further configured to analyze the amplitude data to determine a varying average amplitude.

[0020] The signal processing unit can be further configured to identify an event as corresponding to an amplitude that exceeds a predetermined limit with respect to the average amplitude.

[0021] Depending on the circumstances, the signal processing unit may be further configured to store amplitude data corresponding to the identified event.

[0022] Depending on the circumstances, the signal processing unit may be further configured to store amplitude data corresponding to a predetermined time period of the identified event.

[0023] The signal processing unit can be further configured to discard amplitude data that is determined not to correspond to an identified event.

[0024] The signal processing unit can be further configured to average the amplitude data of each signal within a predetermined time period.

[0025] Depending on the circumstances, the signal processing unit may be further configured to store the average amplitude data corresponding to the predetermined time period.

[0026] The signal processing unit may include an orthogonal module and a local oscillator, the orthogonal module being configured to multiply each signal from each probe by one of the signals from the local oscillator in order to select a single frequency.

[0027] The signal processing unit may include a phase-locked loop configured to tune the local oscillator such that the local oscillator tracks each signal from each probe between a lower frequency limit and an upper frequency limit.

[0028] The signal processing unit may further include a filtering and averaging module configured to convert the output signal from the quadrature module into an average value and remove modulation from the output signal to generate a signal vector for each signal.

[0029] The signal processing unit may further include a vectorization block configured to uniformly rotate the phase of the signal vectors to generate a voltage signal vector with zero phase on the real number axis.

[0030] Depending on the circumstances, the signal processing unit may further include a first-in-first-out buffer configured to store the signal vectors after processing by the vectorization block.

[0031] Depending on the circumstances, when a pulse is applied to the plasma process, the signal processing unit is configured to analyze amplitude data of at least one of the signals and store parameters of a pulse based on the amplitude data.

Implementation Method

[0033] This teaching provides an apparatus / device and method for performing the following operations: detecting electromagnetic signals radiated from a plasma processing system; processing sensing data in the form of an RF spectrum; and analyzing the spectrum to detect statistically significant plasma process changes to identify fault conditions.

[0034] The sensor of the sensing device may include two or more of the following three sensing elements: a first probe for sensing a time-varying electric field, a second probe for sensing a time-varying magnetic field, and an optical probe (e.g., a high-frequency photodiode) for detecting RF modulated light emission. The second probe and the optical probe sense the RF current within the plasma system. The first probe may be used with one of the other two to detect the phase change between the two fields. When measurements are performed simultaneously and synchronously, this allows the determination of the phase angle between the electric field spectral components and the magnetic field spectral components. That is, the sensing device according to this teaching samples the signal synchronously in a manner that maintains the phase data between the frequencies and harmonics of the signal. This phase measurement is particularly sensitive to changes in the RF impedance of the plasma. The sensing device also includes a signal processing unit. This unit is designed to monitor the phase angle between the fundamental frequency and its harmonics of each individual spectrum. These phase measurements are particularly sensitive to chemical changes within the plasma.

[0035] In this instruction, a remote E-field and B-point probe can be used to detect the RF field emitted from the plasma system. Photodiodes or other optical sensors can be added to obtain additional sensitivity. Hereinafter, these three element sensors according to this instruction can be referred to as VIO sensors. The specific configuration of the VIO sensors according to this instruction will be described below.

[0036] It should be understood that although the terms B-point probe and E-field probe are used herein, these terms are not intended to limit the sensing device to a "perfect" B-point probe that measures only variations along the B-axis or a "perfect" E-field probe that measures only variations along the E-axis. In fact, the sensing device of this teaching can work with probes not aligned with the E or B axes, provided that the probes are sufficiently independent of the electromagnetic field plane. In this teaching, each probe is susceptible to a certain amount of other fields; that is, the E-field probe detects a time-varying RF magnetic field, and the B-point probe detects a time-varying RF electric field. The probes of this teaching can be more broadly described as two distinct electromagnetic field probes, each located in a different vector plane.

[0037] One of the three VIO sensor signals is selected as a reference for synchronizing the other sensor signals to perform phase measurements. The optimal configuration uses the E-field as a reference for synchronizing the B-field and / or the optical RF spectrum. However, any signal from any of the three probes can be selected. Synchronization is necessary to allow determination of the phase relationship between corresponding frequency components of the E-field, B-field, and RF spectra.

[0038] The phase of the B-field RF frequency component relative to the corresponding E-field RF frequency component is related to the cavity geometry and its mechanical structure. The phase of the optical RF frequency component relative to the corresponding E-field RF frequency component is sensitive to the gas chemistry rather than the geometry. Electrons oscillating in the electric field excite gas atoms to higher energy states. The atoms remain in the excited state for a period of time. When the excited electrons decay back to the ground state, they emit photons. The time between the atomic excitation and its subsequent return to the ground state is considered the phase shift between the optical RF frequency component and the corresponding electric field frequency component. Therefore, it is important to measure the RF magnetic field and the phase shift of the RF optical signal to understand the health and performance of the plasma process to the greatest extent possible.

[0039] In modern plasma tools, the viewing aperture size is miniaturized, and RF shielding is added to minimize RF leakage through the viewing aperture. RF shielding is generally more effective at preventing E-field leakage, but less effective at B-field leakage. As long as the internal chamber is visible, optical RF emission is not obstructed by RF shielding. Although optical detectors always require a viewing aperture, as shown in Figure 1, electrical RF signals can be detected remotely in many areas of the plasma system. Examples of areas of the plasma processing system in which radiated RF signals can be detected include, but are not limited to: (101) Coaxial transmission lines between the RF generator and the matching network (unit); whether in line or through a grounded shield (or an area where the shield is removed) (102) Inside the matching unit / circuit, for example at the housing wall (103) Outside the matching unit, via a cooling fan (104) Through the RF housing between the matching unit and the plasma chamber (which may require a slot) (105) Via a turbopump (106) Inside the ICP source area, for example at the housing wall (107) Inside the plasma chamber

[0040] Assuming that multiple different locations in the plasma processing system can be selected to monitor electromagnetic radiation, the sensor of the sensing device according to this teaching has two preferred physical configurations. In one configuration, the sensing elements are co-located in an "antenna" module. The E-field probe, the B-point probe, and the high-frequency photodiode are co-located within this module or housing. This module is ideally suited for mounting at a large viewing aperture that can simultaneously detect two or all three signals. This configuration is shown in Figure 2. This shows the sensor 200 according to this teaching within module 201. Module 201 is located at viewing aperture 204 of the plasma processing system 100.

[0041] Figure 2 also shows the sensor 200 connected to the signal processing unit 202 via a signal coupling cable, as explained in more detail below. The combination of the sensor 200 and the signal processing unit can be regarded as a sensing device.

[0042] In another configuration, the individual sensing elements of sensor 200 are distributed across different areas of plasma processing system 100, where these sensing elements are best exposed to their sensitive radiation signals. This configuration is shown in Figure 3. In this example, optical detector / sensor 301 is located at the observation port (or fiber optic port), B-point sensor 302 is mounted inside the ICP plasma source housing, and E-field probe 303 is mounted on the inner wall of the matching box or in line with the coaxial cable between the generator and the matching box. When mounted between the generator and the matching unit, E-field probe 303 is exposed to significantly lower harmonic signal levels because it is severely attenuated by the matching unit. Nevertheless, a stable E-field reference can be obtained using the highly sensitive signal processing unit 202 to allow accurate measurement of the phase changes of the B-point and / or optical RF spectrum (if the E-field is selected as the reference).

[0043] Point B probe 302 includes an inductive pickup loop with multiple windings. A time-varying magnetic flux passing through the loop induces a voltage at the loop output. The loop voltage is coupled to a signal conditioning circuit (also referred to herein as a frequency response leveling circuit). The signal conditioning circuit adjusts the output voltage of the inductive loop so that magnetic flux of equal amplitude over a wide frequency range induces the same or similar voltage levels at the output of the signal conditioning circuit.

[0044] The E-field probe 303 includes a capacitance pickup. A time-varying electric field charges a capacitor and induces a voltage thereon. The output of the capacitance pickup is coupled to another signal conditioning circuit (frequency response leveling circuit) to achieve the same frequency leveling effect as described with respect to the B-point probe. The signal conditioning circuit is used to adjust the output voltage of the capacitance pickup so that equal amplitude current flux over a wide frequency range induces the same or similar voltage level at the output of the signal conditioning circuit.

[0045] The optical detector 301 may be a photodiode. The intensity of light reaching the photosensitive region of the optical detector induces a voltage at the output. The optical detector 301 is typically sensitive to light in the wavelength range of 200 nm to 1000 nm, but other ranges may be used if necessary. The detector is designed to have a flat response to light of the same intensity within a specified optical bandwidth. In other words, for different wavelengths with the same intensity, the voltage generated at the output of the optical detector is approximately the same. The optical detector has a high frequency bandwidth. This means that the output of the optical detector can respond to time-varying light intensity in the frequency range of 10 kHz to 100 MHz, and is designed to have a flat frequency response in a manner similar to that of point B and field E sensors. Maintaining modulation of the sensor's response within the frequency range of interest prevents saturation problems such as those attributable to resonance effects at certain frequencies.

[0046] Turning to Figure 4, this figure illustrates a remote, non-invasive sensing device according to the present teaching, having a sensing element and a signal processing unit with common positioning. That is, Figure 4 shows the internal configuration of the sensing device (i.e., the sensor 201 and signal processing unit 202 of Figure 2).

[0047] Sensing elements (optical sensor 301, E-field sensor 303, and B-point probe 302) and frequency response leveling circuits 401, 402, and 403 constitute the analog front end of the VIO sensor according to the present teaching. Each leveling circuit outputs an analog voltage signal proportional to the amount sensed by the respective sensing elements 301, 302, and 303. The analog voltage output is an alternating current (AC) signal in the RF band. To extract the spectrum in a form that can be analyzed and visualized in a useful manner, a signal processing unit 202 is incorporated into the VIO sensing device. A separate coaxial cable (which may be bundled in a common sheath) carries the AC signal to the signal processing unit 202. The signal from each sensing element is sampled using a multi-channel ADC. A reference signal (such as the E-field sensor, but any sensor may be selected as the reference) is coupled to channel one of the ADC. The other two channels are synchronized with channel one. It should be understood that the selected reference signal can be fixed, for example, always a signal from the E-field sensor, or the signal processing unit can dynamically select the reference signal for each measurement or sensing procedure. The strongest or highest intensity signal can be selected, or the signal processing unit can use some other criterion to select the reference signal. Any suitable criterion can be selected by someone skilled in this technique.

[0048] The high-speed ADC samples the AC waveform generated by the analog front end. Data sampling is performed simultaneously or synchronously on all channels. Sampling is performed synchronously in a manner that preserves the phase information between the frequencies and harmonics of the two signals. Those skilled in this technique can select appropriate components and specific techniques to ensure that phase information and harmonics are preserved where appropriate. As a first step, a data block of 512 samples is typically recorded. The block size is arbitrarily chosen and can be changed to meet different requirements. The data block is transmitted to a field-programmable gate array (FPGA), where a Fast Fourier Transform (FFT) is performed. The FFT transforms the time-domain AC waveform into a spectrum. The spectrum is sent to a microprocessor (not shown) in the signal processing unit for storage and further processing, including averaging. Multiple FFTs are averaged together to reduce the signal-to-noise ratio. Averaging also preserves channel phase information.

[0049] In this teaching, the AC waveform is sampled asynchronously between blocks but synchronously between channels. That is, whenever a new waveform is recorded on any channel, sampling does not begin at the same point in the AC cycle, but sampling is performed at exactly the same time on all channels. As a result, the phase of the fundamental frequency component and its harmonics relative to zero degrees changes with each new FFT processed. To perform averaging, a phase rotation operation is performed. The fundamental frequency component of the first ADC channel is selected as a reference. The phase of the complex fundamental frequency component is rotated to a predetermined phase angle ɸ. The phase shift Δɸ used to rotate the complex fundamental frequency component from its measured value to the predetermined value is recorded. The complex fundamental frequency components in all individual ADC channels are also rotated by Δɸ. All harmonic frequency components (N) are rotated by N×Δɸ, where N represents the harmonic order, such that N=1 is the first harmonic or fundamental frequency. This serves a dual purpose: aligning each FFT in the phase space to enable averaging of consecutive samples while preserving the phase relationships between harmonics within individual spectra and between corresponding frequency components of the spectra from different ADC channels.

[0050] After averaging the pre-configured number, the signal processing unit 202 outputs the dataset to the user for viewing and further analysis. That is, the dataset can be output by the sensing device to an external computer resource, such as a PC, in the form of a result table. However, further analysis can also be performed by the sensing device (onboard).

[0051] The data set or result table contains the frequency of the fundamental wave of each signal, the magnitude of each signal, the phase of each signal relative to the selected reference signal, the magnitude of each signal harmonic, the phase of each signal harmonic relative to its corresponding harmonic on the reference signal, and the phase of the reference signal harmonic relative to the fundamental wave of the reference signal.

[0052] The exemplary result tables below correspond to two signals—signal frequency 1 and signal frequency 2, respectively. These tables show that, for more than one signal, the result table for the first signal is substantially repeated for any other signal (signal frequency). Table 1 Table 2

[0053] When outputting data, the amplitude and phase information of each frequency component of each sensor channel are reported. Typically, the phase difference between relevant frequency components is calculated and output to the user. The signal processing unit is designed to process multiple fundamental frequency components simultaneously. This is particularly useful when monitoring plasma processes powered by more than one frequency. Depending on the application requirements, the signal processing unit can typically output a complete data set at a selectable rate ranging from approximately 1 millisecond to 1 second.

[0054] Not all RF plasma processes operate in continuous wave (CW) mode. Some RF plasma processes apply pulses at frequencies ranging from a few hertz to tens of kilohertz. For this purpose, the signal processing unit 202 may have a built-in external synchronization port. Unit 202 receives a TTL signal input from the pulse RF generator of the plasma processing system. The boxcar (averaging) technique captures the RF waveform at a specific time during the pulse RF cycle. In this case, multiple pulses are averaged. If needed, this technique can be used to establish an average pulse profile at a one-microsecond resolution. In addition to synchronization, signal processing is performed in the same manner. The pulse RF signal may have one or more power levels during the repetition cycle, but once the TTL synchronization signal is available, the signal processing unit can perform analysis without hindrance.

[0055] In recent years, pulsed RF plasma processes have been developed, enabling the fundamental carrier frequency to be dynamic rather than fixed. This allows for greater control over power matching within the pulse period, where mechanical movement of the matching position at the desired speed is impractical. Typically, a dynamic range of + / - 10% of the carrier frequency is sufficient. The signal processing unit according to this teaching is also dynamically carrier frequency enabled for continuous wave and pulsed RF process monitoring. The intelligent FFT process tracks carrier frequency movement in frequency intervals adjacent to where the fundamental frequency is expected to occur. Therefore, the carrier frequency spectrum can be obtained at any point within its + / - 10% dynamic tuning range.

[0056] The sensing device (sensor and signal processing unit) according to this teaching can use another signal processing method for high-speed event monitoring. The aforementioned spectrum analysis requires the acquisition of complex waveforms and further FFT processing. FFT is a relatively slow process. Therefore, root mean square (RMS) detection technology can be used, with a narrow bandpass filter used to isolate the fundamental frequency. It should be understood that the sensing device according to this teaching can operate in two modes—(i) the spectrum mode described above and (i) the high-speed detection mode described below.

[0057] In high-speed detection mode, the root mean square (RMS) amplitude of each data channel can be determined instantly and at high speed (typically a report rate of one microsecond or less), and high-speed amplitude data can be captured. This high-speed method does not allow for harmonic spectrum measurements. However, real-time, high-speed processing enables pulse profile monitoring without requiring synchronization with TTL signals. Individual pulse profiles can be captured and analyzed against the aforementioned average pulse profile obtained in boxcar technology. Important pulse characteristics, such as duty cycle and pulse repetition frequency, can be determined pulse-by-pulse to establish defect or fault correlation metrics. Pulse profile monitoring is only one type of high-speed measurement that high-speed detection methods can be used for. Another important type of high-speed event is RF arcing. Arcing is common in industrial plasma processing chambers, and it occurs for many reasons. A common type is micro-arcs. For example, dielectric particles form a nanolayer on the chamber wall. When the outer surface of this layer is charged until a certain point is reached, electrical breakdown of the layer occurs. The arc burns through the layer to the grounded wall behind it. When this is done, particles are emitted from the surface, which can pose a significant risk to the wafer being processed. These arcs cause sudden and abrupt changes in plasma voltage and current. The severity of these changes can be classified according to their duration and the relative changes in voltage or current levels induced. Measuring the onset of micro-arcs is crucial, as this can often be the prelude to a major or catastrophic arcing event that could damage the processing tool. Signal processing units operating in high-speed detection mode are well-suited for detecting such micro-arc events.

[0058] Additionally, the high-speed mode provides complete "time coverage" of signals from sensors / probes. Known methods use block techniques that monitor the temporal portion of a signal, similar to looking through a fence, potentially seeing what lies behind it even when not 100% covered. That is, prior art relies on discrete sampling, while this teaching provides continuous monitoring. For event detection, 100% time coverage of the signal is important so that no event is missed. That is, continuous monitoring of the entire plasma process can be provided, and amplitude data for the entire process can be obtained. The signal processing unit is configured to continuously determine the amplitude of each signal from each probe to (continuously) generate amplitude data.

[0059] Figure 5(a) illustrates the data flow from the ADC to memory in the high-speed operation mode. This all occurs within the VIO sensing device and is executed by a combination of the FPGA and onboard CPU of the signal processing unit 202. In this exemplary embodiment, all high-speed amplitude data is transmitted from the FPGA to the CPU in blocks at a resolution of one microsecond. That is, high-speed amplitude data is transmitted from the preprocessed data memory 501 to the event detection and classification module 502, pulse measurement module 503, and data averaging module 504 of the CPU. The CPU checks the events and stores up to 5 ms of data locally in RAM (event data memory 505) for each detected event. It should be understood that any time length can be selected where appropriate, and 5 ms is merely an example. The event detection technique uses a moving average to compare with the most recent value. In the exemplary embodiment, the moving average is only the average of the most recent 1000 values, that is, the most recent 1 ms. Events are classified as corresponding to amplitudes that exceed user-defined limits regarding the moving average.

[0060] The CPU also averages the measured RMS signal level and stores the average data every 100 ms (in context data memory 506). Similarly, the choice of 100 ms is arbitrary and any time period can be selected. This "context" average differs from the aforementioned moving average and, in the exemplary embodiment, is the average of points within each 100 ms block.

[0061] The CPU checks for arc and RF pulse events and stores a snapshot of the arc and / or pulse for each event in the event data memory 505. That is, the CPU processes all data in the first-in-first-out buffer (FIFO) 509 and only stores data corresponding to the events in the event data memory. Average "context" data (i.e., 1 data point every 100 ms) is stored so that the signal processing unit records the average RMS value.

[0062] Regarding the pulse (profile) measurement module 503, it provides high-resolution capture of non-compliant pulses for storage in event data memory 505. The pulse measurement module 503 provides pulse statistics (i.e., period, duty cycle, average "on" value, maximum and minimum "on" value) to the context data memory every 100 ms.

[0063] The software running on the connected PC can retrieve data from the event data memory and context data memory of the sensor and store it in a file on the connected PC.

[0064] It should be understood that the signal processing unit does not need to capture and store event data and context data. Only one such process may be executed upon user request. However, capturing context data and event data is advantageous. Having "context" data provides a way to determine what else happened when the event occurred, i.e., whether an event occurred after power was applied, and whether the voltage and current before and after the event differed from normal conditions (relative to the previously operated plasma process).

[0065] Referring to Figure 5(b) provides a more detailed overview of how the ADC 506, RMS detector 507, and filtering and averaging block 508 are used to capture and process high-speed amplitude data stored in memory 501.

[0066] The sensing device may also have a communication interface. This can be used to stream data to external computing resources (e.g., a PC). The interface also allows the computing resources to request data instead of streaming it.

[0067] The block diagram shown in Figure 5(b) illustrates the FPGA logic used for the aforementioned high-speed (RMS) detection. The FPGA logic shown in Figure 5(b) for RMS detection is very different from the logic used for harmonic spectrum detection (spectral mode). It should be understood that the configuration described herein is merely exemplary, and those skilled in the art can use alternative logic configurations to achieve the same high-speed detection mode of the sensing device according to this teaching.

[0068] Referring to Figure 5(b), in high-speed mode, the signal processing unit of the sensing device tracks only one fundamental frequency and determines the amplitude and phase of the sensor signal at an update rate of one microsecond or faster. This data is stored in a first-in-first-out (FIFO) buffer 509 in preparation for being retrieved by the CPU. The CPU interface allows the FPGA registers to be memory-mapped in the CPU memory space. The CPU sets up direct memory access (DMA) transfers from the FPGA's FIFO directly to the memory.

[0069] The RMS FPGA logic has two clock domains, one related to the ADC clock sampling rate and the other related to the CPU interface. The high-level diagram shown in Figure 5(b) illustrates the signal path from ADC 506 to FIFO memory 509. The quadrature (IQ) block 510 multiplies the V, I, or O signal (signals from individual sensing elements) by the local oscillator signal, thus effectively selecting the frequency to be monitored.

[0070] A phase-locked loop (PLL) control logic 511 is used to tune the local oscillator so that it tracks incoming signals between the lower and upper frequency limits. A filtering and averaging block 508 converts the instantaneous IQ signal into an average value. The IQ output will contain the desired "DC" signal, plus a modulation signal related to the sampling frequency and the signal frequency. The filtering and averaging logic 508 removes the modulation from the IQ output signal to generate a VIO signal vector. The VIO signal vector is then passed to a vectorization block 512 every microsecond, where the VIO signal vector undergoes a uniform phase rotation so that the V signal is real, i.e., the vector will be on the real axis and have zero phase. Subsequently, V, along with Ir and Ii and / or Or and Oi (where the subscripts r and i represent the real and imaginary components, respectively), are stored in a FIFO 509 for read by the CPU of the signal processing unit.

[0071] Those skilled in this art will understand that the high-speed (RMS) detection technique described herein is not limited to the VIO sensing device of this teaching, but can also be used with known prior art sensors (such as the aforementioned VI sensing device or any sensor capable of supplying appropriate signals). This high-speed detection technique does not require all signals (V, I, and O).

[0072] Although the VIO remote sensing device according to this teaching can be calibrated to provide absolute measurements of the E-field, B-point field, and modulated light intensity, this is not necessary for the type of application described herein. In the following description, a method for determining a baseline for a known healthy process is proposed, relying solely on relative signal strength. An acceptable process window is established using a process run sequence from one of the known healthy processes. The healthy process window follows the normal distribution of the measured variables. The baseline determination process involves measuring a sample of the distributions of all variables. Statistical methods are used to determine whether a new process run is within the acceptable process window. For ease of statistical analysis, a database is used, in which each measurement is timestamped and stored for later retrieval. This statistical analysis can be performed locally or on an external computer.

[0073] One example of a baseline determination sequence requires calculating the mean (μ) and standard deviation (s) for each variable for a selected sample size (n). These parameters are stored in memory. The distance from the mean can be determined for each new measurement, measured in terms of standard deviation. Other methods may also be used.

[0074] Individual variables may be insensitive to certain types of process variations. For each fundamental frequency, each spectrum may include 15 harmonic amplitudes and 15 associated phases. Using 3 ADC data channels and inter-spectral phase parameters, hundreds of variables are obtained for each measurement. Therefore, a multivariate model was implemented to obtain optimal sensitivity. When comparing with a baseline sample, using the distance from the mean analysis method is only one of many methods that can be used to determine bias within a multivariate dataset. Principal component analysis and neural networks can also be used. Derivatives from the measured dataset can also be used to further improve sensitivity.

[0075] Because VIO sensing devices can generate a large number of variables, it can sometimes be useful to include only the most sensitive variables in the calculation to obtain optimal sensitivity. Alternatively, all measured variables can be included. The spectral fingerprint is shown in Figure 6. In this example, the baseline of the plasma processing chamber is determined before a set of 5 wafers sequentially pass through the processing chamber. The X-axis shows the spectral component numbers, such as 1-15 being the E-field amplitude of the fundamental frequency and 14 subsequent harmonics, 16-30 being the B-point amplitude of the same frequency component, and 31-45 being the phase angle between channels of each frequency component. Element 46 in this spectrum is the result of the multivariate distance calculated from the mean for each wafer. For all 5 wafers, the calculated value is approximately one standard deviation, indicating that the process is always in a healthy state. To prevent false alarms, a threshold value can be implemented, for example, only a distance greater than 6 sigma from the mean is considered a fault (fault score).

[0076] The spectrum shown in Figure 6 has 45 spectral components (and the result of fault score calculation). Including the optical RF spectrum brings the number of spectral components to 75. Furthermore, adding harmonic phase measurements relative to its fundamental frequency adds an additional 14 spectral components to each effective data channel. This provides a dataset highly sensitive to most faults that may occur during plasma processing. The fault score method provides a relatively simple fault detection technique in which a large dataset is compressed into a single fault score with statistical significance. However, different faults may return the same score. More complex methods are required if fault classification is needed. For fault classification, spectral pattern recognition techniques can be used, enabling the training of intelligent algorithms to identify specific fault characteristics or fingerprints.

[0077] The sensing devices and methods described to date can be used to detect various process faults and events, including: a) chamber wall conditions; b) process endpoints; c) pressure control valve malfunctions, airflow problems, and leaks; d) wafer shift; e) RF arcing events; f) atypical RF pulse events.

[0078] The techniques described herein can be performed by a sensing device or by an external computer connected to the sensing device and using data from the sensing device.

[0079] It should be noted that categories e) and f) above do not require the same baseline determination method described earlier. Chamber wall conditions are an important consideration in many plasma processes. During material etching or deposition in wafer fabrication, layers are formed on the chamber walls. These layers, especially when non-conductive, can significantly alter plasma impedance. This can change plasma properties and push process conditions outside the acceptable process window. When the chamber walls are known to be clean, the previously described method can be used to determine the process baseline. When the chamber walls reach a predetermined contamination level, a chamber cleaning process is performed. Accurate feedback on when the chamber is sufficiently clean is highly desirable. Figure 7 shows the chamber failure score calculated during the cleaning process. A score of 2σ is determined to indicate a sufficiently clean chamber. This technique is used to alert the user when the chamber reaches a clean state. For the example shown in Figure 7, the chamber reaches a clean state after approximately 800 seconds.

[0080] Plasma processing typically involves multiple etching and deposition steps to create features and structures on a substrate surface. The etching step usually requires the complete removal of a layer of material to expose the underlying material. The etching process duration must be carefully selected to ensure features are created within dimensional tolerances. If the etching process does not run long enough, the layer will be under-etched, meaning it is not completely removed and the underlying layer cannot be exposed. If the etching process runs for too long, the layer will be over-etched, meaning the underlying layer may be damaged while the layer of interest is completely removed but the etching process continues. The term "etch endpoint" is used to define the point in time when the etching process has completely removed the layer of interest. After this layer is completely removed, the plasma process begins etching the underlying material. Optical emission spectroscopy (OES) is typically used to monitor the plasma composition and look for the presence of particles from the underlying material. Once these particles are detected, it can be assumed that etching of the layer of interest is complete, and the process can be properly terminated. OES is not always effective for very low concentrations of foreign particles. It has been shown that the fault scoring technique using the VIO sensing device according to this teaching may be more sensitive to low open area etched endpoints compared to OES technology. Figure 8 shows an example of an etched endpoint. Data at the start of the process is used to determine the process baseline. A 4σ score has been determined as the optimal time to terminate the process. RF harmonic spectra, especially harmonic phase components (relative to the fundamental frequency), are highly sensitive to minute changes in plasma chemistry during the etching of endpoints. Using RF spectra and the T-score method already described, very low levels of plasma composition changes due to the introduction of new material into the process can be detected. The optical RF spectrum detector of this invention should not be confused with the standard OES technology mentioned. In this invention, a high-frequency broadband photodiode is used to capture the RF spectrum by measuring the modulation of light intensity. Standard OES systems monitor the time-averaged light intensity emitted by the plasma process and decompose it into a series of light wavelengths. Specific wavelengths are used to identify different types of atoms or molecules. Therefore, OES can be used to determine the chemical composition of the plasma process. However, at low material concentrations, resolution limitations become an issue.

[0081] Many plasma processing chambers operate under vacuum. Chamber pressure is strictly regulated. Gas flow and concentration are also strictly controlled. Failure of the pressure and / or gas control system can have catastrophic consequences for the substrate being processed. To test the sensitivity of the sensing device according to this teaching in detecting gas flow problems, a controlled experiment was conducted in a plasma processing chamber operating silane plasma for silicon deposition applications. Baseline spectra were recorded. The silane concentration was increased by 10% and decreased by 10% relative to the baseline conditions, while keeping all other parameters constant. RF fingerprints were recorded under both conditions. The fingerprint fault score space of the 45-channel RF spectrum is shown in Figure 9. Channel 46 shows the multivariate fault score results. Significant differences exist in the spectral fingerprints under the two conditions, and the fault scores are greater than 5σ, making changes in silane concentration easily detectable. Much lower concentration changes can be detected in a wide range of plasma gas chemistry.

[0082] Wafer positioning on the pedestal (or electrostatic chuck) of the plasma processing unit is crucial for wafer processing. There are several reasons why wafers may shift on the chuck. The robotic arm may malfunction and place it in the wrong position. More commonly, debris from other parts of the tool may fall onto the chuck, causing the wafer placed on it to be lifted off the chuck surface. Regardless of the cause, undetected wafer shift can lead to incorrect wafer processing and scrapping on the production line. Wafer shift affects plasma impedance and the resulting harmonic spectrum. Figure 10 shows the impact on the B-point field measured remotely using a VIO sensor device. The main graph shows the relationship between the amplitude of the basic B-point field and time during the plasma process. Data for six correctly placed wafers are shown together with data for shifted wafers. The processing time for shifted wafers is shortened. The inset in Figure 10 also shows the harmonic spectra of the E-field and B-point signals for correctly placed and shifted wafers. The differences in this example are very obvious and no statistical analysis is required. Nevertheless, this example demonstrates the ability of a VIO sensing device according to this instruction to detect misaligned wafers to prevent wafer scrap events.

[0083] As previously stated, when the central signal processing module is configured for high-speed (RMS) detection, short-term (RF) events can be detected at a distance. Figure 11 shows an example of detecting events such as electric arcs in a plasma process. Specifically, the top view of Figure 11 illustrates the presence of an electric arc in a typical plasma process. The bottom view of Figure 11 shows a high-resolution view of an electric arc with a resolution of 1 microsecond.

[0084] It should be understood that the VIO sensing device described herein provides a non-invasive method for detecting RF plasma arcs. Early detection of arcs in the plasma processing chamber is highly desirable because remedial measures can prevent product scrapping. Arc characteristics are present in the E-field signal, the B-point signal, and the optical RF signal. Therefore, as shown in Figure 3, one or all sensing elements can be used to detect arcs at various locations around the plasma system at a distance.

[0085] Pulsed RF plasma processes are widely used in advanced semiconductor node manufacturing. Applying pulses to plasma enables the realization of different plasma chemistry properties that are not obtainable in continuous wave RF modes. These plasmas are typically pulsed in the frequency range of a few hertz to tens of kilohertz. Achieving highly repeatable plasma conditions from pulse to pulse is crucial for process yield. To achieve these repeatable plasma conditions, variations in pulse characteristics must be kept within certain thresholds. For example, the number of known wafer defects is related to pulse variations. The high-speed (RMS) detection mode of VIO sensing devices is well-suited for analyzing each pulse profile to extract key pulse characteristics, such as pulse repetition frequency, duty cycle, and pulse on-time. The signal processing unit (pulse profile measurement module 503) can sample the incoming data signal every microsecond (and other time intervals are also possible). This provides a detailed snapshot of each pulse profile within the pulse frequency range of interest. Each pulse profile is analyzed in real time, and important pulse parameters are stored. If a pulse profile with a 1-microsecond sample falls within the user-configured tolerance limits, it is discarded. On the other hand, if the pulse exceeds the tolerance, a high-resolution pulse profile is stored for further analysis. This "exceptional" approach limits the amount of data generated; that is, only high-resolution profiles of pulses that do not meet the tolerance criteria are stored. Figure 12 shows a pulse profile captured by the point B sensor of the VIO sensing device according to this teaching. This is an atypical pulse where the duty cycle is diagnosed as being outside the acceptable tolerance range. In particular, Figure 12 provides an example of capturing a single pulse profile at a resolution of one microsecond within a five-millisecond window.

[0086] It should be understood that the statistical analysis and processing of the detected data or RF spectrum described above with reference to Figures 6 to 12 can be performed locally on the sensing device of this teaching or remotely on a computing resource (e.g., a PC) connected to the sensing device.

[0087] In view of the foregoing, it should be understood that the present teaching provides an apparatus and method for performing the following operations: sensing RF signals radiated from a plasma processing system; processing the signals in the form of an RF spectrum; analyzing the spectrum to determine a plurality of amplitude and phase components; and performing statistical analysis based on changes in the amplitude and phase components to identify fault conditions of the plasma processing system.

[0088] The sensors of the sensing device include an E-field probe, a B-point probe, and / or an optical RF detector. The E-field probe (or any of these probes) can be used as a reference for the other probes and for sensing time-varying electric fields. The B-point probe and the optical RF detector are used to sense time-varying electron motion or plasma current within the reactor. The probes can be located together or dispersed at suitable locations around the plasma system.

[0089] Unlike prior art designs, the sensing device is not coupled to a spectrum analyzer or network analyzer. These analyzers are expensive and cumbersome to use. Furthermore, they do not enable the necessary advanced detection functions. Instead, a signal processing unit is used. The signal processing unit provides new information related to the remote sensing RF spectrum, which has not been reported in prior art remote RF plasma sensor designs. Novel measurements include: a) harmonic phase, phase between spectra from different sensors, spectral analysis in pulsed RF and frequency-tuned plasma processes, RMS detection of arcs, and RMS detection of individual pulsed RF profiles. A statistical method is described based on the spectral fingerprint of known "healthy" plasma process conditions. Changes in the phase and amplitude of spectral components are analyzed, and fault scores are assigned to each new process measurement. Therefore, threshold values ​​can be configured to alert the user to detected process faults. Phase measurements are particularly sensitive to minute changes in plasma chemistry and plasma impedance. As described, the ability to measure phase makes the sensing device of this teaching a very useful diagnostic tool for detecting subtle process changes that occur during the low open region endpoints, for example, where standard technology is currently inadequate.

[0090] This document provides three methods for sensing the RF spectrum of radiation from a plasma processing system using the following: a) a time-varying E-field sensor, b) a time-varying magnetic field sensor, and c) a time-varying light intensity sensor. The sensing device according to this teaching provides non-invasive sensing of the RF spectrum of radiation simultaneously and synchronously via these three mechanisms or pairs, thus providing additional data channels not provided by prior art designs.

[0091] The present invention is not limited to one or more embodiments described herein, but may be revised or modified without departing from the scope of the present invention. [Simplified Explanation of the Diagram]

[0032] The present application will now be described with reference to the accompanying drawings, in which: Figure 1 shows a plasma processing system, demonstrating various regions where electrical and photoelectromagnetic emissions can be detected; Figure 2 shows a sensing device according to the present teaching, wherein individual sensing elements are in a single module; Figure 3 shows a sensing device according to the present teaching, wherein individual sensing elements are distributed near the plasma processing system; Figure 4 shows in more detail the sensing device shown in Figure 2 with sensing elements in common location; Figure 5(a) shows the data flow from the ADC to the memory of the sensing device when operating in high-speed detection mode; Figure 5(b) shows the FPGA logic of the sensing device when operating in high-speed detection mode; Figure 6 shows an illustration of a fault score spectral fingerprint generated using the sensing device according to the present teaching; Figure 7 shows an illustration of fault score monitoring of chamber cleaning generated using the sensing device according to the present teaching; Figure 8 shows a graph illustrating an example of low open area endpoint detection using the sensing device according to the present teaching; Figure 9 shows multiple graphs illustrating how the sensing device according to the present teaching can detect pressure, gas flow, and gas concentration in a plasma processing system; The graphs shown in Figure 10 illustrate how a sensing device according to the present teaching can detect misaligned wafers in a plasma processing system; the graphs shown in Figure 11 illustrate how a sensing device according to the present teaching allows the detection of arcs in a plasma process; and the graph shown in Figure 12 illustrates how a sensing device according to the present teaching allows the detection of pulses in a plasma process.

Claims

1. A sensing device for monitoring electromagnetic radiation emitted from a plasma processing system, the sensing device comprising at least two of the following: (i) a first probe for detecting a time-varying RF electric field; (ii) a second probe for detecting a time-varying RF magnetic field; and (iii) an optical probe for detecting modulated light emission; and the sensing device further comprising: A signal processing unit configured to receive a signal from each probe; The electromagnetic radiation is monitored at only a single frequency relative to each signal; the amplitude of each signal from each probe is continuously determined to generate amplitude data; and the amplitude data is analyzed according to a high-speed detection mode to determine a varying average amplitude.

2. The sensing device of claim 1, wherein the signal processing unit is further configured to identify a high-speed event based on the analysis of the amplitude data according to the high-speed detection mode.

3. The sensing device of claim 2, wherein the signal processing unit is further configured to store amplitude data corresponding to the identified high-speed event.

4. The sensing device of claim 3, wherein the signal processing unit is further configured to store amplitude data corresponding to a predetermined time period of the identified high-speed event.

5. The sensing device of claim 4, wherein the signal processing unit is further configured to discard amplitude data determined not to correspond to the identified high-speed event.

6. The sensing device of any one of claims 2 to 5, wherein the signal processing unit is further configured to average the amplitude data of each signal over a predetermined time period.

7. The sensing device of claim 6, wherein the signal processing unit is further configured to store the average amplitude data corresponding to the predetermined time period.

8. The sensing device of claim 1, wherein the signal processing unit includes an orthogonal module and a local oscillator, the orthogonal module being configured to multiply each signal from each probe by a signal from the local oscillator in order to select a single frequency.

9. The sensing device of claim 8, wherein the signal processing unit includes a phase-locked loop configured to tune the local oscillator such that the local oscillator tracks each signal from each probe between a lower frequency limit and an upper frequency limit.

10. The sensing device of claim 8, wherein the signal processing unit further includes a filtering and averaging module configured to convert the output signals from the quadrature module into an average value and remove modulation from the output signals to generate a signal vector for each signal.

11. The sensing device of claim 10, wherein the signal processing unit further includes a vectorization block configured to uniformly rotate the signal vectors to generate a voltage signal vector having zero phase on the real number axis.

12. The sensing device of claim 11, wherein the signal processing unit further includes a first-in-first-out buffer configured to store the signal vectors after processing by the vectorization block.

13. The sensing device of claim 1, wherein when a pulse is applied to a plasma process, the signal processing unit is configured to analyze amplitude data of at least one of the signals and store parameters of a pulse based on the amplitude data.

14. The sensing device of claim 1, wherein the high-speed detection mode can achieve pulse profile monitoring without synchronization with a TTL signal.

15. The sensing device of claim 2, wherein a high-speed event includes an RF arc event or an RF pulse event.

16. The sensing device of claim 6, wherein averaging the amplitude data of each signal over a predetermined time period includes averaging 1000 samples over a time period of 1 ms.

Images