Detection and mitigation of anomalous plasma events in semiconductor processing

Abstract

In particular embodiments, abnormal plasma events may be detected and mitigated, which may include forming an arc in a semiconductor processing chamber. In certain embodiments, a method may include detecting a light signal emitted by a plasma, converting the light signal to a voltage signal, and forming an adjusted voltage signal. In response to determining that a change associated with the adjusted voltage signal exceeds a threshold, an output power of an RF signal coupled to the chamber may be adjusted. This adjustment may mitigate the formation of anomalous plasma events occurring within the chamber.

Description

This application is a divisional application of the invention patent application with application number 202080028432.7, application date February 7, 2020, and invention name “Detection and Mitigation of Abnormal Plasma Events in Semiconductor Processing”. Incorporated by Reference

[0001] The PCT application form is filed concurrently with this specification as a part of this application. Each application to which this application claims the benefit or priority as identified in the concurrently filed PCT application form is incorporated herein by reference in its entirety and for all purposes. Background Art

[0002] Semiconductor processing often involves the use of plasma in a manufacturing chamber; during such operations, abnormal plasma events may occur in response to the electrical conditions used to generate the plasma. Abnormal plasma events may include, for example, the formation of an arc, or other types of visible events resulting from a discharge between two or more localized plasma regions within a processing station. Abnormal plasma events may result in excessive heating and / or excessive current, which may damage the semiconductor wafer being processed. In some instances, abnormal plasma events may cause damage to a semiconductor processing station. Summary of the invention

[0003] An implementation may include a method of detecting an abnormal plasma event in a semiconductor processing chamber. Such an implementation may include: detecting an optical signal emitted by a plasma in the semiconductor processing chamber, wherein the plasma is formed in response to an RF signal from an RF generator. The method may also include: converting the optical signal to a voltage signal; and adjusting the voltage signal to form an adjusted voltage signal. The method may also include: determining whether a change associated with the adjusted voltage signal exceeds a threshold; and adjusting an output parameter of the RF signal from the RF generator based at least in part on the determination.

[0004] In a specific implementation, the method may include: adjusting the voltage signal to form the adjusted voltage signal by filtering the voltage signal with a low pass filter. Determining whether the change associated with the adjusted voltage signal exceeds a threshold includes: comparing the voltage signal with the change of the adjusted voltage signal. Adjusting the voltage signal to form the adjusted voltage signal may include: applying an offset to the adjusted voltage signal. Adjusting the voltage signal to form the adjusted voltage signal may include: obtaining a first derivative of the voltage signal. Adjusting the output parameter of the RF generator may include: reducing the output power of the RF generator from a first power level to a second power level. The method may also include: maintaining the output power of the RF generator at the second power level for a first time period after reducing the output power of the RF generator from the first power level to the second power level. The method may also include: increasing the output power of the RF generator from the second power level to the first power level for a second time period after the maintaining. The second power level may be a non-zero power level. Alternatively, the second power level is zero. The method may also include determining a first amount of RF power reduction based at least in part on light intensity or process parameters within the semiconductor processing chamber, and reducing the output power level of the RF generator by the first amount, wherein the process parameters may include one or more process parameters selected from the group consisting of the following process parameters: DC power level, RF bias power level, RF power variation between stations, frequency tuning parameters, pressure and temperature.

[0005] In one or more implementations, a semiconductor processing tool may include a semiconductor processing chamber. The semiconductor processing tool may include: an RF generator configured to provide RF power to the semiconductor processing chamber to generate and maintain plasma. The semiconductor processing tool may include: a photodetector configured to detect a light signal indicating that the plasma in the semiconductor processing chamber emits light. The photodetector may be configured to convert the light signal into a voltage signal using a slope change detection unit. The semiconductor processing tool may include: receiving the voltage signal. Adjusting the voltage signal to form an adjusted voltage signal. The semiconductor processing tool may be configured to: determine whether a change associated with the adjusted voltage signal exceeds a threshold. In one implementation, in response to the determination, the semiconductor processing tool may cause a signal configured to cause an adjustment to the RF generator to be sent to the RF generator.

[0006] In some implementations, the semiconductor processing tool may include: a lens on the semiconductor processing chamber, which is configured to allow light within the semiconductor processing chamber to pass through the lens. The semiconductor processing tool may also include a fiber optic cable between the lens and the photodetector, and may be configured to transmit at least a portion of the light passing through the lens to the photodetector. The slope change detection unit of the semiconductor processing tool may also include a filter for filtering the voltage signal to convert the voltage signal into the adjusted voltage signal. The semiconductor processing tool may also include a comparator for comparing the change between the adjusted voltage signal and the voltage signal. The comparator in the semiconductor processing tool may include transistor-transistor logic (TTL) configured to send a TTL signal to the RF generator. The RF generator may be configured to reduce RF power from a first power level to a second power level in response to receiving the signal from the slope change detection unit. The slope change detection unit may include a differentiator configured to obtain a derivative of the voltage signal to convert the voltage signal into the adjusted voltage signal. The semiconductor processing tool may further include a comparator configured to determine whether the change in the adjusted voltage signal exceeds the threshold. The semiconductor processing tool may further include: an optical fiber cable between the semiconductor processing chamber and the photodetector, configured to transmit light generated by the plasma in the semiconductor processing chamber to the photodetector. The photodetector may be electrically connected to the RF generator, wherein the optical signal from the semiconductor processing chamber is fed from the semiconductor processing chamber to the RF generator. The photodetector and the slope change detection unit are disposed outside the semiconductor processing chamber. Alternatively, the photodetector and the slope change detection unit may be disposed between the semiconductor processing chamber and the RF generator. The photodetector and the slope change detection unit may be disposed on the semiconductor processing chamber.

[0007] In a specific implementation, a method of detecting an abnormal plasma event may include: forming a plasma using a signal from an RF generator; and detecting a fluctuating optical signal generated by the plasma. The method of detecting may include: calculating a spectral density of the fluctuating optical signal; and determining that the spectral density of the fluctuating optical signal differs by a threshold amount from one or more reference spectral densities of the fluctuating optical signal generated by the plasma.

[0008] In a specific implementation, calculating the spectral density of the fluctuating optical signal may utilize a fast Fourier transform (FFT). The method may include the one or more reference spectral densities of the fluctuating optical signal corresponding to the spectral density of the fluctuating optical signal generated by the plasma maintained under nominal conditions. The threshold amount may be equivalent to one standard deviation relative to the spectral density of the fluctuating optical signal from the plasma maintained under nominal conditions. The threshold amount may be equivalent to two standard deviations relative to the spectral density of the fluctuating optical signal from the plasma maintained under nominal conditions. Calculating the spectral density of the fluctuating optical signal may include determining the spectral density of the fluctuating optical signal at a frequency of about 400 kHz. Calculating the spectral density of the fluctuating optical signal may include determining the spectral density of the fluctuating optical signal at a frequency between about 400 kHz and about 4 MHz. Calculating the spectral density of the fluctuating optical signal may include determining the spectral density of the fluctuating optical signal at multiple frequencies generated by one or more RF generators.

[0009] In a specific implementation, an apparatus may include a signal processor configured to analyze an output signal from a light detector that receives a light signal from a multi-station processing tool. The signal processor is further configured to: detect a fluctuating light signal generated by a plasma formed in a semiconductor processing chamber; calculate a spectral density of the fluctuating light signal; and determine that the spectral density of the fluctuating light signal differs by a threshold amount from one or more reference spectral densities of the fluctuating light signal generated by the plasma.

[0010] In a specific implementation, the one or more baseline spectral densities of the fluctuating optical signal may be equivalent to spectral densities calculated when a plasma is maintained at nominal conditions. The threshold amount may be equivalent to one standard deviation relative to the spectral density of the fluctuating optical signal from the plasma maintained at nominal conditions. Alternatively, the threshold amount may be equivalent to two standard deviations relative to the spectral density of the fluctuating optical signal from the plasma maintained at nominal conditions. The spectral density of the fluctuating optical signal may be calculated when an RF signal having a frequency of about 400 kHz is coupled to the multi-station processing tool. The spectral density of the fluctuating optical signal may be calculated to determine the spectral density of the fluctuating optical signal at a frequency between 400 kHz and 4 MHz. The spectral density of the fluctuating optical signal may be calculated at the frequency of the RF signal coupled to the multi-station processing tool. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Various implementations disclosed herein are depicted by way of example, and not by way of limitation, in the figures of the accompanying drawings in which like reference numerals refer to similar elements.

[0012] Figure 1 A first example of an abnormal plasma event detection apparatus is depicted.

[0013] Figure 2 A first exemplary slope change detection unit capable of detecting an abnormal plasma event is depicted.

[0014] Figure 3A and 3B Waveforms representing exemplary detection responses of a slope change detection unit are depicted.

[0015] Figure 4 A schematic diagram of an exemplary circuit capable of detecting an anomalous plasma event is depicted.

[0016] Figure 5 A waveform representing RF power coupled to a processing station of a process chamber as a function of time is shown.

[0017] Fig. 6A Another waveform is shown which represents the RF power output and the signal output of the slope change detection unit.

[0018] Figure 6B A set of contours representing the spectral density of light energy fluctuations under nominal operating conditions within a processing station is depicted.

[0019] Figure 6C Depicted are a set of contours representing spectral density of light energy fluctuations under nominal operating conditions within a process station, and contours representing spectral density of light energy fluctuations when an abnormal plasma event occurs within the process station.

[0020] Fig.6D and 6E Depicted Figure 1 An alternative configuration of a first exemplary abnormal plasma event detection apparatus.

[0021] Fig. 7A A flow chart of a first exemplary technique for detecting and mitigating the effects of an anomalous plasma event is depicted.

[0022] Figure 7B A flow chart of a second exemplary technique for detecting and mitigating the effects of an anomalous plasma event is depicted.

[0023] Figure 8 Block diagrams are provided of exemplary devices that may be used to implement the disclosed embodiments.

[0024] Fig. 9A schematic diagram of an embodiment of a multi-station processing tool is shown.

[0025] Fig.10 A substrate processing apparatus for depositing films on semiconductor substrates is shown.

[0026] Fig.11 An implementation of a multi-station processing tool is depicted.

[0027] Fig.12 The Paschen curve for helium is depicted. DETAILED DESCRIPTION

[0028] In the following description, many specific details are described to provide a complete understanding of the embodiments. The disclosed embodiments can be practiced without some or all of these specific details. In other cases, well-known process operations are no longer described in detail to avoid unnecessarily obscuring the disclosed embodiments. Although the disclosed embodiments will be described in conjunction with these specific embodiments, it should be understood that this is not intended to limit the disclosed embodiments.

[0029] In this application, the terms "semiconductor wafer", "wafer", "substrate", "wafer substrate", and "semi-finished integrated circuit" are used interchangeably. It should be understood by those skilled in the art that the term "semi-finished integrated circuit" can refer to a silicon wafer during any of the many stages of integrated circuit manufacturing on the silicon wafer. The wafers or substrates used in the semiconductor device industry may have a diameter of 200 mm, or 300 mm, or 450 mm. The following detailed description assumes that the present invention is implemented on a wafer. However, the present invention is not so limited. The workpiece may have a variety of shapes, sizes, and materials. In addition to semiconductor wafers, other workpieces that can utilize the present invention include various objects, such as printed circuit boards, glass panels, and the like.

[0030] Many semiconductor processes use plasma during one or more aspects of processing. Plasma is typically generated in a processing chamber between two electrodes, where one electrode is connected to a radio frequency (RF) power source and the other electrode is connected to an RF ground. The RF power supplied to the electrodes may be between about 200 watts and about 3000 watts. The frequency used to drive the plasma formation may include a high frequency (HF) component, a low frequency (LF), or both a HF and a LF component. The HF frequency may be about 13.56 MHz or about 27 MHz; the LF frequency may be between about 300-400 kHz. Other frequencies of high or low frequency RF power may also be used, but "high" frequency generally means a signal frequency that is several orders of magnitude higher than a "low" frequency, such as megahertz compared to kilohertz.

[0031] One type of abnormal plasma event that can occur during plasma formation is a high voltage breakdown, "lightup," or arcing within a processing chamber, for example, in the area surrounding a showerhead or substrate support structure such as an electrostatic chuck (ESC). Arcing or high voltage breakdown may occur under several circumstances when two structures that can serve as electrodes are separated by a gap. These circumstances generally depend on the voltage difference between the electrodes, the gas composition between the electrodes, the gas pressure between the electrodes, and the size of the gap between the electrodes. The interaction between these factors can be characterized by Paschen's law, which provides a high voltage breakdown voltage as a function of pressure, gap distance, and two gas-dependent parameters. The voltage required to initiate an abnormal plasma event, such as a lightup or arc, can be referred to as a "breakdown voltage" and is a function of, for example, the gas filling the volume between the two electrodes, the pressure of the gas, and the distance between the potential endpoints of the arc. This relationship is determined by Paschen's law, which is stated as Among them, "V B ” is the breakdown voltage, “p” is the pressure, “d” is the distance, and “a” and “b” are empirically derived constants associated with the gas involved.

[0032] Fig.12 (Implementation 1200) depicts a Paschen curve for helium. The logarithmic vertical axis represents the breakdown voltage (e.g., the voltage required to produce an arc or other type of abnormal plasma event) in volts, while the horizontal axis (also logarithmic) represents the pressure multiplied by the distance in Torr-cm. Fig.12 As shown, the lowest, or minimum, breakdown voltage of about 150 V is at the lowest point of the Paschen curve, which occurs at about 4 Torr-cm. This characteristic establishes two regions where semiconductor processing can occur before the lowest breakdown voltage is reached—the "left" of the lowest point of the curve and the "right" of the lowest point of the curve. Fig.12 The vertical dashed line in approximates the dividing line between these two sides.

[0033] Some semiconductor processes utilize higher RF power to cause a reduction in deposition time and increase substrate productivity. However, the use of higher RF power may increase the risk of voltage breakdown, arcing, or other types of abnormal plasma events in the processing chamber. These events may cause damage to the substrate, to the electrodes (e.g., the showerhead or substrate support structure), and to the chamber itself. Unacceptable damage to the substrate may cause the substrate to be scrapped and / or may cause damage to the electrodes. In addition, a damaged chamber may require long, expensive, and / or extensive repairs of the electrodes (e.g., replacing the showerhead or susceptor), which may result in additional costs and additional downtime of the semiconductor processing chamber.

[0034] Therefore, it is desirable to detect arcs or other types of abnormal plasma events in a processing chamber and then mitigate such events. Many conventional abnormal plasma event detection techniques may result in missed and erroneous arc detections. For example, a conventional method may include a voltage detection method that measures the RF voltage delivered to a driven electrode to generate a plasma and determines whether the voltage reaches a threshold voltage or is close to a specific amount different from the threshold voltage; if so, such a system assumes that an arc action may occur in the processing chamber. In response to detecting the threshold voltage, the RF power is reduced, or completely turned off, which in turn reduces the voltage in the processing chamber and prevents abnormal plasma events from occurring. The threshold voltage is generally a predetermined value obtained using experimental measurements. However, this detection method has disadvantages, such as being limited to a single predetermined voltage threshold. In addition, the detection method may be difficult to implement because the method may require a very fast voltage monitoring speed that is computationally intensive and may not accurately reflect the real-time occurrence of abnormal plasma events. The inability to accurately reflect the real-time occurrence of abnormal plasma events may be due to the potential time displacement between the occurrence of the arc event and the measured voltage. Furthermore, the detection method may be potentially insensitive to smaller arcing events, suffer from system noise, and may require extensive time and experimentation to determine and optimize the appropriate threshold voltage.

[0035] Threshold voltage detection methods may have other detection limitations. For example, such methods typically detect whether the measured voltage has increased above the specified threshold. However, the RF drive signal comprises a time-varying voltage driven at various frequencies (e.g., a periodic waveform, such as a sinusoidal signal operating at 13.56 MHz), and the intensity of light emitted by the plasma may oscillate with a similar frequency component in the detector output signal (attributable to the normal intensity of light emitted by the plasma in response to the RF signal at the nominal peak), then, if the threshold voltage is appropriately set, the abnormal plasma event that causes the additional light emission may be detectable. However, if the abnormal plasma event occurs between successive peaks of the plasma cycle, the abnormal plasma event may not be detected, especially if the component in the detector output signal attributable to the abnormal plasma event has a smaller amplitude than the amplitude of the component in the detector output signal attributable to the RF signal. For example, assuming that the measured voltage of the detector output signal has a peak amplitude of 0.2V with an average of 2V due to the application of the RF drive signal, the voltage may increase to 2.2V at the peak and then decrease to 1.8V at the bottom (so that the peak to peak is 0.4V). If an arc occurs during the cycle when the voltage is at a peak of 2.2V, which produces a 0.3V increase in the detector output signal, the measured voltage will be 2.5V, which is above the threshold of 2.4V (which may indicate the occurrence of an arc, as determined from experimental results). However, if an arc occurs during the cycle when the voltage is at a bottom of 1.8V, the measured voltage will be 2.1V (1.8V plus 0.3V), which is below the threshold of 2.4V. The threshold in such a system cannot be set below the peak output signal amplitude produced by applying a normal RF drive signal. Accordingly, this detection method may not be useful for detecting small voltage abnormal plasma events that occur during portions of an alternating (e.g., sinusoidal) voltage cycle.

[0036] Another similar technique can optionally monitor abnormal plasma events. Here, similar to the above method, the voltage is measured. However, the voltage can be obtained from the output of a photodetector that is configured to measure the light intensity in the chamber and generate a voltage signal indicative of the light intensity. For example, the plasma formed in the chamber may emit different amounts of light depending on the plasma behavior. The photodetector can therefore generate a voltage output signal representing the light intensity emitted by the plasma. However, an arc event may generate additional light, which causes an increase in the voltage generated by the photodetector that generates the corresponding voltage. The voltage may then be monitored in the manner previously discussed for direct voltage monitoring. However, this technique suffers from many of the same disadvantages as the direct voltage monitoring previously discussed, particularly insensitivity to small intensity discharge events that may not be detected if they occur during the local minimum of the photodetector signal.

[0037] Another limitation of these detection methods is that they may not detect abnormal plasma events that are smaller than the difference between the peak voltage and the threshold voltage. For example, the threshold voltage may shift beyond the peak voltage by a certain amount so that the peak voltages, and / or small insignificant voltage deviations, are not close to the threshold voltage and therefore do not result in an alarm or other corrective action. However, harmful abnormal plasma events may occur at voltages below the shift between the peak voltage and the threshold. Therefore, abnormal plasma events that are smaller than the difference between the peak voltage and the threshold voltage may not be detected. For example, if an abnormal plasma event of 0.15V occurs at a peak voltage of 2.0V, the measured voltage will be below the 2.2V threshold and thus remain undetected and unmitigated.

[0038] These conventional detection methods may also be undesirably sensitive to parameter settings of the processes occurring within the processing chamber. For example, each time a process parameter is changed (e.g., RF power, RF frequency, gas species, chamber pressure, etc.), multiple new threshold voltages may be required in response to the effect of the modified process parameter on, for example, the voltage threshold at which arcing occurs. Thus, modification of parameter settings may require additional experimentation and optimization time, which may also result in undesirable downtime of the semiconductor processing tool.

[0039] Thus, the present disclosure includes novel techniques and apparatus for abnormal plasma event detection in semiconductor processing. In general, a plasma may emit light detectable by a light detector (e.g., a light sensor). The light emitted by the plasma may oscillate or fluctuate at various frequencies based on plasma parameters (e.g., chemical species, RF power parameters (voltage, etc.), RF frequency components (e.g., high frequency components, low frequency components, or both), etc.). Some exemplary plasma illumination oscillation frequencies may be on the order of one cycle per tens of milliseconds, e.g., about 1 cycle / 20 ms. In contrast, the occurrence of abnormal plasma events generally occurs at a relatively fast, high frequency relative to other light emitted by the plasma; for example, on the order of cycles of several nanometers.

[0040] The techniques discussed below utilize the difference between light energy fluctuations that occur during abnormal plasma event frequencies and light energy fluctuations that occur during nominal plasma operation. This difference can be used to identify abnormal plasma events. In some implementations, because abnormal plasma events may cause sudden (e.g., high frequency) changes in slope in the photodetector output signal due to their higher frequencies, the slope change of the photodetector signal at a certain moment can be determined. In other implementations, the photodetector signal can be normalized by removing the lower frequency portion of the signal, leaving only the higher frequency portion that can indicate an abnormal plasma event. In other implementations, the photodetector signal can be subjected to, for example, Fourier analysis (e.g., fast Fourier transform), or other types of frequency spectrum analysis, to determine the intensity amplitude of various frequency components. In a particular implementation, if the analyzed photodetector signal exhibits high frequency components greater than a certain amplitude, these frequency components can indicate an abnormal plasma event. These abnormal plasma event detection techniques may differ from conventional techniques, such as those described above, because in a particular implementation, the abnormal plasma event detection technique considers the rate of change of the optical signal. This is different from conventional techniques, which rely only on the amplitude of the detected signal.

[0041] Figure 1 A first exemplary abnormal plasma event detection apparatus according to an embodiment 100 is depicted. As shown, the embodiment 100 includes a semiconductor processing chamber 102, a lens 104 (which may be omitted in several embodiments), a fiber optic cable 106 connected to the lens 104 and a signal processor 108, and an RF generator 110. The semiconductor processing chamber 102 is discussed in more detail below, but it can be used for any semiconductor processing that utilizes plasma, such as plasma enhanced chemical vapor deposition (PECVD), atomic layer deposition (ALD), atomic layer etching (ALE), etc. The RF generator 110 is also discussed in more detail below and may include a LF power generator, a HF power generator, or both. The lens 104 is configured within the semiconductor processing chamber 102 so that light generated within the chamber (included within any processing station in the semiconductor processing chamber 102) can pass through the lens 104, via the fiber optic cable 106, and reach the signal processor 108.

[0042] The signal processor 108 may include a photodetector, such as a photodetector 112 (which may include a photodiode), which may receive a light signal from the semiconductor processing chamber 102 and convert the received light signal into a voltage signal. The photodetector 112 may include any other type of light detection electronics capable of detecting any number of photons of energy delivered at various wavelengths of infrared light, visible light, and / or ultraviolet light, and may include, for example, a spectrometer. The signal processor 108 may also include a slope change detection unit 114, which is configured to receive a voltage signal from the photodetector 112, as represented by the dashed line between the photodetector and the slope change detection unit 114. The signal processor 108 may also be configured to determine a change in the voltage signal from the photodetector 112, and determine whether the change in the voltage signal indicates the occurrence of an abnormal plasma event occurring within the semiconductor processing chamber 102.

[0043] In some implementations, the slope change detection unit 114 can compare the raw photodetector signal to a filtered version of the same signal to determine high frequency changes between the raw signal and the filtered signal. The comparison between the raw signal and the filtered signal can enable a determination that an abnormal plasma event has occurred. In some such implementations, a signal having a high frequency component caused by an abnormal plasma event is compared to a signal that does not have the high frequency component (due to the use of a low pass filter applied to the raw voltage signal, etc.). The comparison can enable a determination of relative changes between the filtered signal and the unfiltered version of the signal.

[0044] Figure 2 A first exemplary slope change detection unit according to embodiment 200 is depicted. Here, the photodetector 112 is shown with a first exemplary slope change detection unit 214 surrounded by a dashed rectangle. The raw voltage signal 215 from the photodetector 112 is shunted and transmitted to a filter, which in some embodiments may include a buffer amplifier 216, a low pass filter 218, and a feedback amplifier 220 with a threshold / enable signal 222. The filtered signal is then transmitted to a high speed comparator 224 along with the raw voltage signal 215. The buffer amplifier 216 is operable to isolate the raw analog data from the low pass filter 218 (which filters high frequency signals), and the threshold / enable signal 222 causes an offset in the filtered signal.

[0045] exist Figure 2In an embodiment of the invention, the high speed comparator 224 operates to compare the raw photodetector voltage signal to the filtered photodetector voltage signal and determine whether there has been a change between the signals. For example, a determination may be made as to whether the difference between the two signals at a given moment exceeds a threshold amount. In response to the difference between the filtered signal and the unfiltered signal being greater than the threshold amount, this may indicate a slope change in the raw signal, thereby indicating an abnormal plasma event. In a particular embodiment, the filtered signal as discussed above may be offset relative to the unfiltered signal by a particular offset amount, which may operate to avoid a scenario where the two signals would crosstalk with each other (unlike the situation due to an abnormal plasma event). The high speed comparator 224 may also include transistor-transistor logic (TTL), which may change the output signal 226 from a first voltage to a second voltage in response to the detected change between the filtered signal and the unfiltered signal being greater than a first threshold.

[0046] Figure 3A and 3B A waveform representing an exemplary detection response of a slope change detection unit is depicted according to one embodiment. Figure 3A In (Embodiment 300A), the analog raw photodetector voltage signal (305) is represented as a dashed line, the filtered and offset voltage signal (310) is represented as a dotted line, and the output TTL signal 315 is represented as a solid line. The vertical axis is scaled in volts, and the horizontal axis is scaled in time (in seconds). As shown, the detected plasma luminescence oscillates as a sine wave with an amplitude of about 0.2V. Figure 3B In the example, Figure 3A A detailed view of a portion of is depicted. Here, an abnormal plasma event has occurred, and because the high frequency arc signal has been filtered out, the filtered signal remains constant. However, in this example, the abnormal plasma event, which may correspond to an arc, has caused a change in the photodetector voltage signal. In this example, the change in the photodetector output voltage signal corresponds to an increase in the slope of the photodetector output voltage signal. In response to a comparison between the simulated raw photodetector voltage signal 305 and the filtered output signal 310 being offset from each other by a first threshold value (e.g., at Figure 3B The high speed comparator causes the output TTL signal 315 to change from about 0.2V to about 3.8V. Figure 3BAs further shown in (embodiment 300B), in response to an abnormal plasma event (e.g., an arc), the analog raw photodetector voltage signal 305 changes again (e.g., decreases in slope). In response to the analog raw photodetector signal 305 decreasing in slope, such that the analog raw photodetector signal 305 and the low-pass signal 310 are no longer offset from each other by a value greater than the first threshold, the high-speed comparator then causes the output TTL signal 315 to change from about 3.8V back to about 0.2V.

[0047] Figure 4 An exemplary circuit capable of detecting an abnormal plasma event according to an embodiment 400 is depicted. The circuit of embodiment 400 may represent Figure 1 The signal processor 108 and Figure 2 Aspects of the first exemplary slope change detection unit 214. Figure 4 In the embodiment of the present invention, the "analog photodetector input" may represent the Figure 2 The photodetector 112 generates a photovoltage signal. Figure 4 The "threshold value from EIOC" (V 3 ) may represent a threshold / monitoring enable signal 222, which may be used as Fig. 9 A portion of the control program executed by the system controller 950 is transmitted from the Ethernet input and output controller (EIOC). Figure 4 The "buffer" can represent Figure 2 The buffer amplifier 216. The feedback amplifier 220 (depicted as Figure 4 The "feedback amplifier" in the 3 ) signal and Figure 4 The output signals from the buffers are added (or summed). Figure 4 The output signal of the buffer is filtered by the combination of R1 / C1 / C2, which forms a low pass filter with Figure 2 The low-pass filter 218 corresponds to. Figure 4 The output signal of the buffer is passed through Figure 4 The comparator (corresponding to Figure 2 of the comparator 224) and the combined analog photodetector input (V 2 ) and sine wave input (V 4 In response to a comparison of the set of combined analog photodetector input and sine wave input with the output signal from the feedback amplifier, Figure 4 The comparator detects an abnormal plasma event. In certain embodiments, an amplitude of the combined analog photodetector input and the set of sine waves that is greater than the output signal from the feedback amplifier by a threshold amount results in a detected abnormal plasma event.

[0048] Although Figure 2 The slope change detection unit function can be utilized Figure 4 The analog electronic circuit is implemented in this way, but in some other implementations, Figure 2 The function of can be implemented digitally, for example, by a field programmable gate array (FPGA). In some other implementations, the slope change detection unit can use a differentiator, which is operable to obtain a derivative of the original optical data signal (e.g., an optical data signal from the optical detector 112) to determine whether a change in the slope of the output signal from the optical detector has occurred. The output signal from the differentiator can be proportional to the rate of change of the input signal voltage (e.g., a derivative with respect to time). The derivative of the original optical data signal (e.g., from the optical detector) can be transmitted to an input port of a comparator (e.g., a high-speed comparator 224). The high-speed comparator can implement a comparison between the original data signal and a threshold to determine whether the difference between the slopes of the original optical data signal exceeds the threshold. In some instances, the threshold sensitivity can be adjusted depending on process changes during deposition to avoid false alarms. If the change has exceeded the change threshold, the comparator 224 can again output a TTL signal at a second voltage with an increased amplitude relative to the first voltage.

[0049] In some implementations, the slope change detection unit can use a fast Fourier transform (FFT) to convert the raw photovoltage signal to the frequency domain. The signal represented in the frequency domain can be used to determine whether a number of frequency components (such as those associated with arc occurrence (or other abnormal plasma events)) are present at an intensity greater than a certain threshold amount. In some implementations, if an unexpected frequency is detected (or detected to a certain extent exceeding the threshold) in response to an abnormal plasma event, an output signal is generated.

[0050] In response to the detected abnormal plasma event, for example, by the generation of a TTL signal, various reactions and actions may be taken, such as adjusting the RF power coupled to the process chamber. In some examples, the adjustment may include reducing (or completely interrupting) the RF power (e.g., LF, HF, or both). In some such examples, after the amplitude of the RF power has been reduced to zero, the RF power may be increased to the previous level after a period of time sufficient to allow arcing (or other abnormal plasma events in a process chamber) to cease. The increase in RF power may correspond to a linear or non-linear increase. In some other examples, the adjustment of the RF power may include reducing the RF power (e.g., LF, HF, or both) to a non-zero level for a period of time, and then increasing the power to a level equivalent to the original level.

[0051] Figure 5A waveform 500 is shown which represents the RF power coupled to a processing station of a chamber as a function of time. Figure 5 As shown in the figure, at time t1, the RF power (given by Figure 1 The power supplied by the RF generator 110 of the embodiment of the present invention corresponds to the power level P1. At time t1, in response to the detected abnormal plasma event, the RF power is reduced to the power level P2 and maintained for a period of time from time t1 to time t2. At time t2, the RF power is gradually increased (e.g., ramped) between time t2 and time t3 until the RF output power reaches the level P1. In some embodiments, the RF power reduction can be a percentage reduction, such as a 10% or 30% reduction. However, it should be understood that the disclosed implementations are not limited to such examples and are intended to include any percentage reduction in RF power, such as 5%, 10%, 20%, 30%, etc.

[0052] In some implementations, the magnitude of the detected abnormal plasma event can be determined, for example, by subtracting the low pass filtered data points of the photodetector signal from the corresponding unfiltered data points to determine the intensity of the abnormal plasma event. The low pass filtered data points can be subtracted from the corresponding unfiltered data points to estimate, for example, the arc light intensity. This determination (or other determinations that can provide some indication of the intensity of the abnormal plasma event), in some implementations, can be used to determine the extent of RF system adjustments in response to the abnormal plasma event. For example, if the detected abnormal plasma event has a reduced amplitude, the system can cause the RF generator to switch to a lower power mode at a first level. However, if the amplitude of the detected RF event reaches a certain threshold amplitude, the system will cause the RF generator to switch to another lower power mode at a second level lower than the first level. In such a system, there can be multiple power levels supported, where the supported power levels can correspond to different threshold amplitudes of the abnormal plasma event intensity. In these examples, the system can select a power level associated with the intensity amplitude range (the detected arc event falls within the range).

[0053] Fig. 6A Another waveform is shown according to embodiment 600A, which represents the RF power and the output signal of the slope change detection unit. As shown, the RF power output of the RF generator oscillates, and the output signal of the slope change detection unit corresponds to a relatively flat constant line. Under a detected abnormal plasma event, the amplitude of the output signal of the slope change detection unit increases, and the RF power unit interrupts the RF power output signal. In some examples, the response time between the increase in the slope change detection unit signal and the adjustment in the RF power output can be equivalent to about 2 microseconds.

[0054] In some embodiments where an abnormal plasma event (e.g., the formation of an arc) occurs above or below an edge portion of a semiconductor wafer, the arc intensity and process parameters may define the amount by which the RF power should be reduced. For example, in some embodiments, a high RF power produces a more dispersed RF plasma. Therefore, reducing the RF power coupled to the processing station may operate to confine the plasma to the region between the electrodes. For example, in response to the occurrence of an arc at an edge portion of a semiconductor wafer, reducing the RF power may reduce the plasma density at the edge of the wafer, and thus concentrate the generated plasma in a region between the electrodes and closer to the center of the wafer. Reducing the RF power may also operate to reduce the voltage at the electrodes, which in turn supports reducing the occurrence of abnormal plasma events (e.g., arc occurrence events), even if such events occur near the center of the wafer (away from the edge portion). The intensity of the abnormal plasma event and the process parameters may also be used to define the amount by which the RF power is to be reduced. For example, the higher the arc intensity, the greater the reduction in RF power. Additionally, for very low pressure processes, the plasma may be dispersed compared to high pressure processes (for the same power, gas, and other parameters). Therefore, in such an example, the power may need to be reduced more to result in greater confinement of the plasma.

[0055] Abnormal plasma event detection equipment, such as the above-described apparatus, may be configured in alternative configurations. More specifically, the signal processor 108 may be configured in different locations and may include different configurations. Figure 1 In some implementations, the signal processor may be disposed at a location other than within the process chamber, such as in a different portion of the tool. In some implementations, the optical signal may be transmitted from a location within the process chamber 102 to an optical detector of the signal processor, such as via a fiber optic cable 106.

[0056] Figure 6B A set of contours 600B are depicted according to one embodiment, representing the spectral density of light energy fluctuations under nominal operating conditions within a processing station. Figure 6B The vertical axis of indicates a measure of spectral density (e.g., percentage of the total spectrum), while Figure 6B The logarithmic scale horizontal axis indicates frequency. Thus, the profiles in the set of profiles 600B represent spectral density fluctuations in optical energy in response to RF power coupled at frequencies of about 400 kHz (i.e., ±10%) and about 13.56 MHz (i.e., ±10%) during a predetermined time period. Figure 6BIn a specific example of FIG. 1 , the spectral density of the light energy fluctuations is shown at four sampling periods, where each profile represents the spectral density of the light energy fluctuations at one of the four sampling periods. Thus, in response to coupling of RF power (e.g., including frequencies of 400 kHz and 13.56 MHz) to the processing chamber, the light signal from the plasma may fluctuate or oscillate at a first peak corresponding to a 400 kHz fundamental frequency (±10%), and at harmonics of 400 kHz (e.g., at about 800 kHz (±10%) and 1600 kHz (±10%). Furthermore, it should be understood that the spectral density of the fluctuating light signal from the plasma varies only a small amount at frequencies between about 100 kHz and about 50 MHz. For example, at a frequency of about 100 kHz, the spectral density of the light signal generated by the plasma may be in a range of about 0.8% to about 1.2% of the total spectrum of the fluctuating light signal from the plasma.

[0057] According to implementation 600C, Figure 6C A set of contours representing the spectral density of light energy fluctuations under nominal operating conditions within a process station and contours representing the spectral density of light energy fluctuations when an abnormal plasma event occurs within the process station are depicted. Figure 6B , Figure 6C The vertical axis indicates a measure of spectral density (e.g., percentage of total light energy), while Figure 6C The logarithmic scale horizontal axis indicates frequency. The profiles in the set of profiles 610 represent the spectral density of the optical energy of the fluctuating plasma in response to the coupling of RF power (e.g., 400 kHz and 13.56 MHz) to the process chamber during a predetermined time period. Figure 6C In the specific example of FIG. 6 , the spectral density of the light energy fluctuations is shown at three sampling periods, each corresponding to one of the profiles 610. In response to the coupling of RF power (e.g., 400 kHz and 13.56 MHz) to the process chamber, the light signal generated by the plasma may fluctuate or oscillate at a first peak frequency corresponding to a 400 kHz fundamental frequency. The light signal generated by the plasma may additionally fluctuate or oscillate at harmonics, such as at about 800 kHz and (1600 kHz) 1.6 MHz. Additionally, it should be understood that the spectral density of each of the profiles 610 varies only a small amount at frequencies between about 100 kHz and about 50 MHz.

[0058] However, in addition to the profile 610, Figure 6C Also depicted is a profile 620 representing the spectral density of optical energy fluctuations in response to an abnormal plasma event (e.g., arcing) occurring within the process chamber. Figure 6CAs shown, during an abnormal plasma event, the spectral density of the optical energy fluctuations increases significantly between about 600 kHz and about 3.6 MHz. In other embodiments, during an abnormal plasma event, the spectral density of the optical energy fluctuations may increase significantly over a different frequency range, such as between about 400 kHz (±10%) and about 4 MHz (±10%). For example, at about 1 MHz (1000 kHz) the spectral density of profile 620 (which may be observed during an abnormal plasma event in a chamber) corresponds to a value of about 2%. This can be compared to the spectral density observed under nominal conditions (e.g., as indicated by profile 610), which corresponds to a value of about 1%. Thus, as Figure 6C As shown in , an abnormal plasma event (eg, as shown by profile 620) may be indicated by observing a significant increase in the spectral density of plasma fluctuations between certain frequency ranges.

[0059] exist Figure 6C In a specific example of the present invention, an abnormal plasma event is indicated by observing or detecting an increase in spectral density of about 100% (e.g., from about 1% spectral density to about 2% spectral density) at a specific frequency or during a specific range of frequencies relative to the nominal plasma conditions of the processing chamber at the specific frequency or frequency range. Additionally, such an increase in spectral density may be observed at frequencies between about 400 kHz (±10%) and 4 MHz (±10%), but it should be understood that the disclosed implementations are not limited to such examples and may include spectral densities and / or spectral density profiles observed at other frequency ranges (e.g., frequencies below about 400 kHz and frequencies above about 4 MHz). Furthermore, in other implementations, an abnormal plasma event may be indicated by a smaller increase in spectral density of a fluctuation or oscillation signal from the plasma relative to nominal operating conditions. For example, in a specific implementation, an abnormal plasma event may be identified in response to an increase in spectral density of less than 100% (e.g., 75%, 50%, etc.), and the disclosed implementations are not limited to such examples and may include other increases in spectral density.

[0060] In certain implementations, an abnormal plasma event may be identified in response to an increase in spectral density relative to the standard deviation of the spectral density observed under nominal operating conditions. Thus, in one example, an abnormal plasma event may be identified by first calculating the standard deviation of the spectral density for a set of profiles (e.g., profile 610) representing nominal (or baseline) plasma conditions. In another example, a moving average (and moving standard deviation) may be calculated over a particular duration. In another example, the standard deviation may be calculated continuously using a time window that expands in size as a function of time. The method used to calculate the standard deviation, or other statistical measure, may depend on the particular process performed at the processing station and may vary during wafer processing.

[0061] In response to calculating a standard deviation using a baseline spectral density, an abnormal plasma event can be identified in response to observing a fluctuation signal from a plasma characterized by a spectral density profile having an amplitude greater than, for example, a calculated standard deviation (e.g., using the standard deviation of the baseline on the spectral density). In another example, an abnormal plasma event can be identified in response to observing a spectral density profile having an amplitude greater than two standard deviations from a calculated standard deviation (e.g., using the standard deviation of the baseline spectral density). However, the disclosed implementations are intended to include spectral density profiles that indicate abnormal plasma events that differ by any number of standard deviations relative to a spectral density profile of a fluctuation signal from a plasma operating under nominal conditions.

[0062] Short return Figure 1 In certain embodiments, the signal processor 108 may be programmed to perform analysis on the raw output signal from the light detector 112. In one or more of such embodiments, analysis of the output signal from the light detector 112 may be utilized to determine the spectral density of light energy fluctuations under various operating conditions within the processing station. Figure 6C , a profile representing plasma energy fluctuations during a predetermined time period may be analyzed to determine the spectral density at various frequencies as a fraction or percentage of the total spectrum of light energy fluctuations that are plasma-derived. In some examples, the analysis performed by the signal processor 108 may involve performing a fast Fourier transform (FFT) on the output signal from the light detector 112. After performing an FFT on the output signal from the light detector 112 during one or more predetermined time periods, the spectral density of the light energy fluctuations may be similar to Figure 6B / 6C. In response to the analysis of the spectral density of the light energy fluctuation, the signal processor 108 may determine that an abnormal plasma event has occurred. In a specific embodiment, the abnormal plasma event may be identified in response to detecting an increase in the spectral density at a specific frequency of the light energy fluctuation (e.g., between about 400kHz and about 4MHz). The abnormal plasma event may be identified in response to observing a spectral density with an amplitude greater than a threshold amount relative to the spectral density observed during the nominal plasma condition. In a specific embodiment, for example, the threshold amount may be equivalent to one or two standard deviations of the spectral density calculated during the nominal plasma condition. The abnormal plasma event may be identified in response to observing a spectral density with an amplitude different from one or two standard deviations with reference to the spectral density calculated during the nominal plasma condition, and the disclosed implementation is not limited to such an example. In addition, the abnormal plasma event may be identified by detecting an increase in the spectral density at a frequency other than between about 400kHz and about 4MHz, and the disclosed implementation is not limited to such an example. On the contrary, the implementation is meant to include the detection of the spectral density within any range of the oscillation or fluctuation of the light energy of the plasma.

[0063] Fig.6D and 6E Depicted Figure 1 An alternative configuration of a first exemplary abnormal plasma event detection device. Fig.6D In the embodiment, the abnormal plasma event detection device 600D may not include a fiber optic cable. In fact, the abnormal plasma event detection device 600D may include a signal processing unit 608A disposed at or on the process chamber 602A so that the light detector obtains the light signal of the plasma at the process chamber 602A. The output signal from the light detector 612A may be processed by a high-speed comparator 614A. The output signal from the high-speed comparator 614A is then transmitted to the RF generator 610A. In some embodiments, the device may include a lens 604 between the process chamber 602A and the light detector 612A. Fig. 6E , the abnormal plasma event detection apparatus 600E includes a lens 604B, which is coupled to a signal processing unit 608B as part of an RF generator 610B via an optical fiber cable 606B. In such an example, the photodetector 612B may also be a component of the RF generator 610B. In some such examples, the RF generator 610A may include a high-speed comparator 614B so that an additional high-speed comparator is not used. An advantage of the abnormal plasma event detection apparatus 600E is that noise near the processing chamber 602B and the signal processing unit 608B may not affect the signal generated and used by the signal processing unit 608B.

[0064] Various techniques may also be used to detect and mitigate arc events within a processing chamber. According to one embodiment 700A, Fig. 7A A flow chart of a first exemplary technique for detecting and mitigating the effects of an abnormal plasma event is depicted. The technique begins at 701, and at 703, a light signal generated by fluctuating plasma luminescence is detected. As described above, the plasma emits light of various wavelengths, and a light sensor or light detector, such as light detector 112, is configured to detect the light emitted by the plasma. Additionally, at 705, the light signal is converted to a voltage signal, and the conversion may be performed by the light sensor or light detector. At 707, the voltage signal is adjusted to form an adjusted voltage signal. The adjustment may be based on any of the techniques described above (e.g., filtering and / or offsetting the raw light voltage signal (e.g., with respect to Figure 2 As described above), obtain the derivative of the original photovoltage signal, or use FFT analysis of the original photovoltage signal).

[0065] In response to the raw photovoltage signal being adjusted, 709 determines whether the change associated with the adjusted voltage signal exceeds a threshold. The determination may be any of the comparisons described above, such as comparing the filtered and / or offset voltage signal (i.e., the adjusted voltage signal) to the raw photovoltage signal to determine whether any change between the two signals exceeds the threshold. As described above, in some embodiments, a determination is made as to whether the change between the two signals exceeds a threshold, which in turn may indicate an arc or other type of abnormal plasma event occurring within the processing chamber. The determination in these embodiments (i.e., embodiments using a filtered light signal) is not based solely on changes in the filtered and / or offset signal itself, but rather is based on changes in the raw voltage signal relative to the filtered and / or offset signal. For example, as in Figure 3B As shown in FIG. 1 , a change in the raw voltage signal relative to the filtered signal indicates the occurrence of an arc or other type of abnormal plasma event. In response to utilizing both signals, the comparison is still a determination associated with the adjusted voltage signal. The threshold can be selected based on various parameters, such as the rate of change and the amount of change.

[0066] In some other embodiments of 709, determining that an abnormal plasma event is occurring can be based on changes in the adjusted voltage signal relative to a threshold, such as using derivative and / or FFT methods. Here, the determination can be based on changes in the adjusted voltage signal. For example, the output signal of the derivative technique may involve a direct correlation method of the rate of change of the optical signal, so that the determination of the adjusted voltage signal can directly indicate the occurrence of an arc.

[0067] If it is determined that the change associated with the adjusted voltage signal exceeds the threshold, mitigation actions may be taken at 711. These mitigation actions may include, as described above, adjusting output parameters of the RF generator, such as by reducing the RF power to a lower power level for a period of time and then gradually ramping the RF power output back to the initial power level.

[0068] Although in Fig. 7A , but as described above, the technique may also include indicating that the determination of 709 causes a TTL signal to be generated and sent to the RF generator. In some embodiments, the RF generator is configured to receive the TTL signal and cause a response to one or more parameters of the RF generator, such as reducing the output power of the RF generator. Additional aspects of the technique may also include, prior to or simultaneously with 703, generating a plasma in the processing chamber by coupling RF power to the processing chamber.

[0069] Figure 7B 700B depicts a flow chart of a second exemplary technique for detecting and mitigating the effects of an abnormal plasma event according to an embodiment. The technique begins at 751, and at 752, a plasma is formed in response to a signal of sufficient amplitude from an RF generator. In some embodiments, the RF generator may, for example, generate signals of about 400 kHz (±10%) and about 13.56 MHz (±10%). It should be understood that the disclosed implementations are not limited to such examples and may include various additional frequencies. At 754, a fluctuating optical signal generated by the plasma may be detected. The detection may be performed by a light sensor or a light detector. At 755, a signal processor may calculate a spectral density of the fluctuating optical signal. In a particular embodiment, the calculation may be performed while operating the plasma under nominal conditions and may occur during several time periods. In a particular embodiment, at 756, it may be determined that the spectral density of the fluctuating optical signal is different from one or more previously calculated spectral densities of the fluctuating optical signal. However, in several other embodiments, 756 may include determining that the spectral density of the fluctuating optical signal is different from a baseline spectral density. In response to the spectral density of the fluctuating optical signal differing by perhaps a threshold amount from a previously calculated spectral density, or any other type of reference spectral density, the signal processor may indicate that an abnormal plasma event has occurred.

[0070] In some embodiments, the semiconductor processing chamber may include two or more processing stations, as described below, and the light emission of each station may be monitored separately. For example, the processing chamber may include four processing stations, and each of these processing stations may utilize a separate light detector configured to detect light from the corresponding processing station.

[0071] The devices and techniques described herein provide many advantages over conventional abnormal plasma event detection techniques. In some instances, the techniques described allow for faster detection times and mitigation response times because the optical signal can be processed using analog circuits that can directly generate signals to an RF generator (the RF generator can respond to the signal). The direct connection between the RF generator and the optical signal enables a fast and efficient response time. In contrast, conventional techniques may require sending the detected signal to a processor for analysis and another signal from the processor to the RF processor. Since some abnormal plasma events (such as arcs) may form quickly, the delays generated by some conventional techniques may cause undesirable damage to the wafer and / or other equipment. Additionally, as described above, the techniques and devices described are capable of detecting small-amplitude abnormal plasma events that are undetectable by conventional methods.

[0072] As mentioned above, the apparatus and techniques described herein can be applied to any semiconductor process that uses plasma, and any semiconductor processing chamber in which plasma is generated. Examples of these processes and apparatus are described below. Plasma assisted chemical vapor deposition (PECVD) equipment

[0073] In some embodiments, the wafer processing equipment may include one or more PECVD processing stations included in a processing tool. The processing tool may include Fig. 6A and 6B The features described above may, for example, be accompanied by the other features described below.

[0074] Figure 8 A block diagram of an exemplary device that can be used to implement the disclosed embodiments is provided. As shown, Figure 8The reactor 800 includes a processing chamber 824, which surrounds the other components of the reactor and is used to contain the plasma generated by, for example, a capacitor type system (which includes a showerhead 814, which operates in conjunction with a grounded heater block 820). A high frequency RF generator 804 (which is connected to a matching network 806) and a low frequency RF generator 802 are connected to the showerhead 814. The power and frequency supplied by the matching network 806 are sufficient to generate plasma from the process gas, for example, 400-700W of total energy. In one embodiment of the present invention, both HFRF generators and LFRF generators can be used during deposition, while in some other embodiments only HFRF generators or LFRF generators are used. In a typical process, the high frequency RF component is generally between about 2-60MHz; in at least one embodiment, the HF component is about 13.56MHz. The low frequency LF component is generally between about 250-400kHz. In some embodiments, the RF generator can be considered to have both HFRF generators and LFRF generators, and in some instances, a matching network.

[0075] Within the reactor, a wafer pedestal 818 supports a substrate 816. The pedestal generally includes a chuck, fork, or lift pins that are used to hold and transfer the substrate during and between deposition and / or plasma processing reactions. The chuck may be an electronic chuck, a mechanical chuck, or various other types of chucks available for use in industry and / or research.

[0076] Process gases are introduced via inlet 812. Multiple source gas lines 810 are connected to manifold 808. These gases may or may not be premixed. Appropriate valve control and mass flow control mechanisms are used to ensure that the correct gases are delivered during the deposition and post-deposition stages of the process. In the example where the chemical precursors are delivered in liquid form, a liquid flow control mechanism is used. The liquid is then vaporized and mixed with other process gases during delivery in the manifold heated above its vaporization point before reaching the deposition chamber.

[0077] Process gases exit chamber 824 via outlet 822. A vacuum pump 826 (e.g., one or more stages of mechanical dry pumps and / or turbomolecular pumps) typically draws the gases out and maintains a suitably low pressure within the reactor through a closed-loop controlled flow restriction (e.g., a throttle valve or pendulum valve).

[0078] The invention can be implemented on a multi-station or single-station tool. In a specific embodiment, a 300 mm Novellus Vector 3000 with a 4-station deposition scheme is used. TM tool or 200mm Sequel with 6-station deposition solution TM tool.

[0079] According to one embodiment 900, Fig. 9 A schematic diagram of an embodiment of a multi-station processing tool 900 having an entry load lock 902 and an exit load lock 904, one or both of which may use a remote plasma source. A robot 906 is configured to move a wafer from a wafer box loaded through a cabin 908 into the entry load lock 902 via an atmospheric port 910 at atmospheric pressure. The wafer is placed on a pedestal 912 in the entry load lock 902 by the robot 906, the atmospheric port 910 is closed, and the load lock is evacuated. In the case where the entry load lock 902 uses a remote plasma source, the wafer can be exposed to a remote plasma treatment in the load lock and then introduced into a processing chamber 914. In addition, the wafer can also be heated in the entry load lock 902, for example, to remove moisture and adsorbed gases. Then, the chamber transfer port 916 of the processing chamber 914 is opened, and another robot (not shown) places the wafer in the reactor on the pedestal of the first station shown in the reactor for processing. Although Fig. 9 The illustrated embodiment includes a load lock, but it will be appreciated that in some embodiments, wafers may be allowed to enter the processing station directly.

[0080] The depicted processing chamber 914 includes Fig. 9 There are four processing stations numbered 1 to 4 in the illustrated embodiment. Each station has a heated susceptor (shown as 918 for station 1) and a gas line inlet. Although the depicted process chamber 914 includes four stations, it should be understood that a process chamber according to the present disclosure may have any suitable number of stations. For example, in some embodiments, a process chamber may have five or more stations, while in other embodiments, a process chamber may have three or fewer stations.

[0081] Fig. 9 Also depicted is one embodiment of a wafer handling system 990 for transferring wafers within the processing chamber 914. In some embodiments, the wafer handling system 990 can transfer wafers between various processing stations and / or between a processing station and a load lock. It should be understood that any suitable wafer handling system can be used. Non-limiting examples include wafer conveyors and wafer handling robots. Fig. 9 Also depicted is one embodiment of a system controller 950 for controlling process states and hardware states of the processing tool 900. The system controller 950 may include one or more memory devices 956, one or more mass storage devices 954, and one or more processors 952. The processor 952 may include a CPU or computer, analog and / or digital input / output connections, a stepper motor controller board, and the like.

[0082] Although in Fig. 9Not shown, tool 900 may include any of the features of reactor 800, such as gases and piping for the various stations described above, and a vacuum pump.

[0083] In some embodiments, the system controller 950 controls all activities of the process tool 900. The system controller 950 executes system control software 958, which is stored in the mass storage device 954, loaded into the memory device 956, and executed in the processor 952. The system control software 958 may include instructions for controlling timing, gas mixtures, gas flow rates, chamber and / or station pressures, chamber and / or station temperatures, purge conditions and timing, wafer temperatures, RF power levels, frequencies, substrate, pedestal, chuck and / or susceptor positions, and other parameters of a particular process performed by the process tool 900. The system control software 958 may be configured in any suitable manner. For example, various process tool component subroutines or control objects may be written to control the operation of the process tool components used to perform various process tool processes. The system control software 958 may be encoded in any suitable computer-readable programming language.

[0084] In some embodiments, the system control software 958 may include input / output control (IOC) sequence instructions for controlling various of the above-mentioned parameters. For example, a PECVD process may include one or more instructions for execution by the system controller 950. Instructions for setting process conditions for a PECVD process stage may be included in a corresponding PECVD recipe stage. In some embodiments, the PECVD recipe stages may be sequentially configured so that all instructions for a PECVD process stage are executed simultaneously with the process stage.

[0085] Other computer software and / or programs stored in the mass storage device 954 and / or the memory device 956 in association with the system controller 950 may be used in some embodiments. Examples of programs or program segments used for this purpose include substrate positioning programs, process gas control programs, pressure control programs, heater control programs, and plasma control programs.

[0086] The substrate positioning program may include program code for processing tool components used to load a substrate onto the pedestal 918 and control the spacing between the substrate and other components of the processing tool 900 .

[0087] The process gas control program may include code for controlling gas composition and flow rate, and selectively flowing gas into one or more process stations prior to deposition to stabilize the pressure in the process station. The process gas control program may include code for controlling the gas composition and flow rate within any of the disclosed ranges. The pressure control program may include code for controlling the pressure in the process station by, for example, adjusting the gas flow into the process station (for example, a throttle valve in the process station exhaust system), etc. The pressure control program may include code for maintaining the pressure in the process station within any of the disclosed pressure ranges.

[0088] The heater control program may include code for controlling the current flowing to a heating unit for heating the substrate. Alternatively, the heater control program may control the delivery of a heat transfer gas (e.g., helium) to the substrate. The heater control program may include instructions for maintaining the substrate temperature within any of the disclosed ranges.

[0089] The plasma control program may include code for setting the RF power level and frequency applied to the processing electrodes in one or more processing stations (e.g., using any of the RF power levels disclosed herein). The plasma control program may also include code for controlling the duration of each plasma exposure.

[0090] In some implementations, the system controller 950 may be part of or coupled to a computer that is integrated with, coupled to, or connected to the system or a combination thereof via a network. For example, the controller may be in the "cloud" or all or part of a wafer fab host system that may allow remote access to wafer processing. The computer may enable remote access to the system to monitor the current progress of a manufacturing operation, examine the history of past manufacturing operations, examine trends or performance criteria for multiple manufacturing operations, change parameters of a current process, set processing steps to follow the current process, or start a new process. In some examples, a remote computer (e.g., a server) may provide a process recipe to the system via a network that may include a local network or the Internet. The remote computer may include a user interface that enables input or programming of parameters and / or settings, which are then communicated from the remote computer to the system. In some examples, the system controller 950 receives instructions in the form of data that specify parameters for each processing step to be performed during one or more operations. It should be understood that the parameters may be for the type of process to be performed and the type of tool that the controller is configured to connect to or control. Thus, as described above, the system controller 950 may be distributed, for example, by including one or more discrete controllers that are networked together and work toward a common goal (e.g., the processes and controls described herein). Examples of distributed controllers for these purposes may be one or more integrated circuits within the chamber that communicate with one or more remote integrated circuits (e.g., at the platform level or as part of a remote computer) that combine to control the processes within the chamber.

[0091] In some embodiments, there may be a user interface associated with the system controller 950. The user interface may include a display screen, a graphical software display of apparatus and / or process conditions, and user input devices such as a pointing device, keyboard, touch screen, microphone, and the like.

[0092] In some embodiments, the parameters adjusted by the system controller 950 may relate to process conditions. Non-limiting examples include process gas composition and flow rate, temperature, pressure, plasma conditions (e.g., RF power level, frequency, and exposure time), etc. These parameters may be provided to the user in the form of a recipe, which may be entered using the user interface.

[0093] Signals for monitoring the process may be provided from various process tool sensors by analog and / or digital input connections of the system controller 950. Signals for controlling the process may be output on analog and digital output connections of the process tool 900. Non-limiting examples of process tool sensors that may be monitored include mass flow controllers, pressure sensors (e.g., manometers), thermocouples, etc. Appropriately programmed feedback and control algorithms may be used with the data from these sensors to maintain process conditions.

[0094] Any suitable chamber may be used to implement the disclosed embodiments. Exemplary deposition equipment includes, but is not limited to, Product Series, product line, and / or The equipment of the product series, each of which can be purchased from Lam Research Corp., (Fremont, California) or various other commercially available processing systems. Two or more of these stations can perform the same function. Similarly, more than two stations can perform different functions. Each station can be designed or constructed to perform a specific function / method as desired. PECVD Processing

[0095] Many semiconductor processing techniques use plasma enhanced chemical vapor deposition ("PECVD") to deposit materials. In a typical PECVD reaction, a substrate is exposed to one or more volatile precursors, which react and / or decompose to produce a desired deposit on the substrate surface. The PECVD process generally begins by flowing one or more reactants into a reaction chamber. Reactant delivery can continue while the plasma is generated, and the substrate surface is exposed to the plasma, which in turn causes deposition to occur on the substrate surface. The process continues until the desired film thickness is reached, after which the plasma is typically extinguished and the reactant flow is terminated. Next, the reaction chamber can be purged, and post-deposition steps can be performed.

[0096] As described above, in some implementations, the frequency used to drive plasma formation during PECVD may contain a high frequency ("HF") component, a low frequency ("LF") component, or both. The HF frequency may be about 13.56 MHz or about 27 MHz. The HF RF power used to drive plasma formation may be between about 200-3,000 W. These power levels represent the total power delivered, which can be divided between multiple stations in a multi-station processing chamber. The plasma can be generated at a first power level, which can be any power within this range, such as 600 W for a single station, or 2,400 W for a four-station processing chamber, which results in 600 W for each of the four stations. The duration of plasma exposure depends on the desired thickness of the deposited film. In some implementations, pulsed PECVD methods can be used. These methods may involve pulsing the precursor and / or RF power level. In some embodiments, the frequency used to drive plasma formation during PECVD may contain both LF and HF components. The LF frequency may be between about 300-400 kHz. The LF RF power used to drive plasma formation may be between about 200-2,500W. Atomic Layer Deposition

[0097] In some embodiments, a semiconductor process involves multiple film deposition cycles, each producing a "discrete" film thickness. Atomic layer deposition (ALD) is one such film deposition method, but any technique that deposits thin layers of film in a repeated sequential manner may be considered to involve multiple cycles of deposition, and the methods and apparatus disclosed herein may also be generally used to control film thickness in such multi-cycle deposition operations.

[0098] As device and feature sizes continue to shrink in the semiconductor industry, and 3D device structures (e.g., Intel's three-gate transistor structure) become more prevalent in integrated circuit (IC) designs, the ability to deposit thin conformal films (films of material having uniform thickness relative to the shape of the underlying structure, even if non-planar) will continue to become more important. ALD is a film-forming technique that is well suited for the deposition of conformal films due to the fact that a single cycle of ALD deposits only a single thin layer of material, the thickness of which is limited by the amount of one or more film precursor reactants that can be adsorbed on the substrate surface (i.e., forming an adsorption-limited layer) prior to the film-forming chemical reaction itself. Multiple "ALD cycles" can then be used to build up a film of the desired thickness, and because each layer is thin and conformal, the resulting film substantially conforms to the shape of the underlying device structure. In certain embodiments, each ALD cycle comprises the following steps: The substrate surface is exposed to a first precursor. A reaction chamber in which the substrate is located is purged. The reaction on the substrate surface is activated, usually using plasma and / or a second precursor. A reaction chamber in which the substrate is located is purged.

[0099] The duration of each ALD cycle is typically less than 25 seconds or less than 10 seconds. One or more plasma exposure steps of the ALD cycle may have a short duration, such as a duration of less than 1 second. Due to the short duration, control of plasma consistency can promote process uniformity. Variations in plasma impedance and delivered power are two factors that may affect process uniformity.

[0100] An apparatus and method for controlling RF power used in a multi-cycle deposition operation in a semiconductor tool having multiple processing stations sharing an RF power source is provided. The frequency of the RF power is controlled as well as the power applied to the individual stations sharing the RF power source.

[0101] Fig.10 A substrate processing apparatus for depositing a film on a semiconductor substrate is shown according to an embodiment. Fig.10 The processing apparatus 1000 has a single process chamber 1002 having a single substrate holder 1008 in an interior volume that can be maintained under vacuum by a vacuum pump 1018. A gas delivery system 1001 and showerhead 1006 are also fluidly coupled to the chamber for delivering, for example, film precursors, carrier and / or sweep gases and / or process gases, secondary reactants, etc. Fig.10 Also shown is a facility for generating plasma within the processing chamber. Fig.10 The apparatus schematically illustrated in provides a basic facility for performing film deposition operations (eg, chemical vapor deposition (CVD) or ALD) on a semiconductor substrate.

[0102] For simplicity, the processing apparatus 1000 is depicted as an independent processing station having a processing chamber body for maintaining a low pressure environment. However, it should be understood that as described herein, multiple processing stations can be included in a common processing tool environment, for example, within a common reaction chamber. For example, Fig.11 An implementation of a multi-station processing tool is depicted according to embodiment 1100. Furthermore, it should be understood that in some implementations, one or more hardware parameters of processing apparatus 1000, including those discussed in detail above, may be programmatically adjusted by one or more system controllers.

[0103] The processing apparatus 1000 is in fluid communication with a reactant delivery system 1001 to deliver the process gas to the distribution showerhead 1006. The reactant delivery system 1001 includes a mixing vessel 1004 for mixing and / or conditioning the process gas for delivery to the showerhead 1006. One or more mixing vessel inlet valves 1020 can control the introduction of the process gas into the mixing vessel 1009.

[0104] Some reactants may be stored in liquid form prior to evaporation and subsequently delivered to the processing chamber 1002 . Fig.10 The implementation of the process station includes a vaporization point 1003 for vaporizing the liquid reactants to be supplied to the mixing vessel 1009. In some implementations, the vaporization point 1003 can be a heated liquid injection module. In some other implementations, the vaporization point 1003 can be a heated vaporizer. In other implementations, the vaporization point 1003 can be eliminated from the processing station.

[0105] In some implementations, a liquid flow controller (LFC) may be provided upstream of the vaporization point 1003 to control the mass flow rate of liquid used for vaporization and delivery to the process chamber 1002 .

[0106] The showerhead 1006 distributes process gases and / or reactants (e.g., film precursors) toward the substrate 1012 at the processing station, with the flow rate being controlled by one or more valves (e.g., valve 1020) upstream of the showerhead. Fig.10 In the illustrated implementation, substrate 1012 is positioned below showerhead 1006 and is shown resting on pedestal 1008. Showerhead 1006 may have any suitable shape and may have any suitable number and arrangement of ports for distributing process gases to substrate 1012.

[0107] The volume 1007 is located below the showerhead 1006. In some implementations, the pedestal 1008 can be raised or lowered to expose the substrate 1012 to the volume 1007 and / or to vary the volume of the volume 1007. Optionally, the pedestal 1008 can be lowered and / or raised during portions of the deposition process to adjust the process pressure, reactant concentration, etc. within the volume 1007.

[0108] exist Fig.10In the embodiment of the present invention, the showerhead 1006 and the base 1008 are electrically connected to the RF power source 1014 and the matching network 1016 for powering the plasma. In some implementations, the plasma energy can be controlled by controlling one or more of the processing station pressure, gas concentration, RF source power, RF source frequency and / or plasma power pulse timing (e.g., via a system controller with appropriate machine-readable instructions). For example, the RF power source 1014 and the matching network 1016 can be operated at any suitable power to form a plasma having a desired free radical species composition. Similarly, the RF power source 1014 can provide RF power at any suitable frequency.

[0109] In some implementations, the plasma can be monitored in situ by one or more plasma monitors. In one case, the plasma power can be monitored by one or more voltage and current sensors (e.g., load sensors such as VI probes). Examples of such sensors include MKS VI-Probe-4100 and 350. Such sensors can measure voltage, current, and phase difference. In some implementations, the sensor can be electrically connected to the RF power source and can be located at or near the showerhead. In such an implementation, the impedance presented to the output port of the RF generator can represent the impedance of the plasma. In another case, the plasma density and / or the process gas concentration can be measured by one or more optical emission spectroscopy (OES) sensors. In some implementations, one or more plasma parameters can be adjusted programmatically based on the measured values ​​from such in-situ plasma monitors. For example, a load sensor can be used in a feedback loop to provide programmatic control of the plasma power. It should be understood that in some implementations, other monitors can be used to monitor plasma and other processing characteristics. Such monitors can include but are not limited to infrared (IR) monitors, acoustic monitors, and pressure sensors.

[0110] In some embodiments, plasma can be controlled via input / output control (IOC) sequencing instructions. In one example, instructions for setting plasma conditions for plasma activation can be included in the corresponding plasma activation recipe of the process recipe. In some cases, the process recipe can be arranged in order so that all instructions for the process are executed simultaneously with the process. In some implementations, instructions for setting one or more plasma parameters can be included in the recipe before plasma treatment. For example, a first recipe may include instructions for setting the flow rate of an inert gas (e.g., helium) and / or a reactant gas, instructions for setting a plasma generator to a power setting value, and time delay instructions for the first recipe. A second subsequent recipe may include instructions for enabling a plasma generator and time delay instructions for a second recipe. A third recipe may include instructions for disabling a plasma generator and time delay instructions for a third recipe. It should be understood that within the scope of the present disclosure, these recipes may be further subdivided and / or iterated in any suitable manner.

[0111] In some deposition processes, the plasma excitation lasts for a period of about several seconds or longer. In some implementations described herein, shorter plasma excitations may be applied during a processing cycle. These times may be less than 50 milliseconds, with 25 milliseconds being a specific example. Such short RF plasma excitations require rapid stabilization and tuning of the plasma. In order to achieve rapid stabilization and tuning of the plasma, the plasma generator may be configured by a two-step tuning process, which includes a coarse tuning component and a fine tuning component. In the coarse tuning component, the impedance matching may be preset to a specific impedance. The coarse tuning component may be preset so that the amplitude of the impedance is at a value of, for example, 50 ohms. In some implementations, the coarse tuning component may be limited to adjustments that affect the amplitude of the impedance. In the fine tuning component, the RF frequency may be allowed to float relative to the baseline frequency in order to try to match the phase to a target value (e.g., a zero phase value). Conventionally, a high frequency plasma is generated at a frequency of about 13.56 MHz. In various implementations disclosed herein, the frequency may be allowed to float to a value different from the standard value in order to match the phase to the target value. In some implementations, the fine tuning components may be limited to adjustments that affect the impedance phase. By allowing the frequency to float while fixing the impedance match to a predetermined impedance, the plasma may stabilize much more quickly. Very short plasma excitations, such as those associated with ALD or atomic layer etch (ALE) cycles, may benefit from rapid stabilization of the plasma.

[0112] The first 1-2 milliseconds of a typical deposition cycle involve ignition of the plasma. After the plasma is ignited, fine tuning of the RF frequency is then performed to match the plasma phase to the target value.

[0113] As described above, one or more processing stations may be included in a multi-station substrate processing tool. Fig.11 An exemplary multi-station substrate processing apparatus that can utilize plasma balancing hardware is shown. With respect to facility costs and operating expenses, by using a Fig.11 A multi-station processing apparatus such as the multi-station processing apparatus shown can achieve various efficiencies. For example, a single vacuum pump can be used to create a single high vacuum environment for all four processing stations by evacuating the waste processing gases of all four processing stations, etc. Depending on the implementation, each processing station can have its own dedicated nozzle for gas delivery, but can share the same gas delivery system. Similarly, certain elements of the plasma generator facility can be shared between processing stations (such as power supplies), but depending on the implementation, certain aspects may be specific to the processing station (for example, if a nozzle is used to apply the potential generated by the plasma). Once again, it should be understood that this efficiency can also be achieved to a greater or lesser extent by using a greater or lesser number of processing stations in each processing chamber (e.g., 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 or 16 processing stations per reaction chamber).

[0114] Fig. 9 The substrate processing apparatus 200 of FIG. 1 may also be used to perform ALD processing. Here, a single substrate processing chamber 914 contains multiple substrate processing stations, each of which can be used to perform processing operations on a substrate held in a wafer holder at the processing station. In this particular implementation, a multi-station substrate processing tool 900 having four processing stations 1, 2, 3, and 4 is shown. Other similar multi-station processing apparatuses may have more or fewer processing stations depending on the implementation and, for example, the desired level of parallel wafer processing, size / space constraints, cost constraints, etc.

[0115] exist Fig. 9 , RF power is generated at the RF power system 913 and distributed to each of stations 1-4. The RF power system may include one or more RF power sources, for example, high frequency (HFRF) and low frequency (LFRF) sources, impedance matching modules and filters. In some implementations, the power source may be limited to a high frequency or low frequency source. Unless otherwise stated, it is assumed that the deposition process described uses only high frequency power. The distribution system of the RF power system is symmetrical about the reactor and has a high impedance. This symmetry and impedance results in approximately equal amounts of power being delivered to each station. Slight differences in RF power between about 5% and 15% may be due to tolerances in the distribution system components, station alignment, temperature differences, and process conditions.

[0116] Small differences in RF power can cause wafer-to-wafer non-uniformity in various film properties, such as composition, thickness, density, amount of cross-linking, chemistry, reaction completion, stress, refractive index, dielectric constant, hardness, etch selectivity, stability, hermeticity, etc. The ability to fine-tune plasma power at a single station and dynamically respond to changes in station conditions can reduce wafer-to-wafer non-uniformity. Note that the present method and apparatus are not limited to multi-station reactors; the method and apparatus disclosed herein are applicable to other RF power systems in which multiple processing areas share an RF power source.

[0117] The system controller 950 may also be used to control process conditions and hardware states of the processing tool 900 and its processing stations to perform ALD. The system controller 950 may provide machine-readable instructions for implementing the deposition process. The instructions may control various process parameters, such as DC power levels, RF bias power levels, station-to-station variations, such as variations in RF power parameters, frequency tuning parameters, pressure, temperature, etc. These instructions may control these parameters to perform in-situ deposition of film stacks according to various implementations described herein.

[0118] Fig.11 1 is a schematic diagram showing various components in an exemplary multi-station plasma reactor having multiple stations, wherein the multiple stations utilize RF frequency tuning to share an RF power source. As shown, an RF power source 1101, which may be a high-frequency RF power source, is distributed to multiple stations 1151 via a distribution network 1121. The HFRF may have a frequency of about 2-60 MHz or about 13.56 MHz. In other implementations, a low-frequency RF power source may be used in addition to or in place of a high-frequency RF power source. The low-frequency RF power source may have a frequency of about 100 kHz to about 1 MHz or about 400 kHz. Certain commercially available RF power sources have the ability to tune the frequency of the RF power source. Examples of such RF power sources include Advanced Engineer's Paramount series, MKS's SurePower series, Comdel's CB, CLX and CDX series, and Huettinger's TruPlasma series.

[0119] The power from the RF power source 1101 can be delivered through an impedance matching system, which can include a fixed matching module 1103. In certain implementations that include both a high-frequency RF power source and a low-frequency RF power source, a high-pass filter and / or a low-pass filter can also be present. In addition, in certain implementations, the power from the RF power source can be delivered through an automatic matching module. In implementations that include a low-frequency RF power source (whether as a supplement or alternative to a high-frequency RF power source), the low-frequency power can be delivered through a fixed match or an automatic match. In certain implementations, an automatic matching module can be used to match the RF power frequency to the set impedance of the load of the plasma. In other implementations, a fixed matching module can be used that does not automatically match the RF power frequency to the set impedance of the load of the plasma.

[0120] exist Fig.11 In the illustrated implementation, the RF power source 1101 is connected to a distribution network 1121. The distribution network 1121 can distribute the RF power generated by the RF power source 1101 to each of the plurality of stations 1151. The distribution network 1121 is connected to an RF regulator 1123 for each of the plurality of stations 1151. For each of the plurality of stations 1151, the RF regulator 1123 is connected to a power parameter sensor 1133 before the showerhead 1153. The power parameter sensor 1133 can be any type of sensor previously disclosed, such as a load sensor or an OES sensor. The RF power source 1101 can obtain instructions from the RF controller 1143 and vary the frequency of the RF power distributed to each station. The instructions can be frequency adjustments based on impedance detected by one or more power parameter sensors 1133. In other implementations, additional sensors can measure a final phase representing the phase of the plasma in all stations 1151. Then, the RF controller 1143 can vary the frequency of the RF power distributed to the station based on the final phase measured by the additional sensor. In some implementations, the RF controller 1143 may include instructions (e.g., code) to vary the frequency of the RF power so that the phase of the impedance is zero or close to zero. Fig.11 In the illustrated implementation, the RF controller 1143 may vary the frequency of the RF power from the RF power source 1101 upstream of each station.

[0121] The RF regulator 1123 is controlled by the RF controller 1143. The RF controller 1143 can change the RF power of each station by a determined amount based on the measurement results of the sensor 1133 from each station 1151. In some implementations, the RF regulator 1123 can be a variable capacitor. The RF controller 1143 can control a stepper motor (not shown) that can change the capacitance of the variable capacitor. Other ways can be used to change the capacitance. For example, the RF regulator 1123 can also be a group of capacitors with respective switches. The RF power can be controlled by activating (turning on) multiple capacitors with specified values. For example, capacitors can be selected to add 1pF, 2pF, 4pF, 8pF and 16pF parallel capacitance to the station. In this example, all combinations of active (open) and inactive (closed) capacitors cover the range of 0pF to 31pF with a resolution of 1pF. By selecting the capacitor to be activated, the controller can change the RF power to the station. This digital control will be faster than using a stepper motor to control the variable capacitor, especially when a large range of capacitance needs to be covered. Depending on the space available and the amount of control desired, one skilled in the art would be able to design an RF regulator using one or more capacitors to vary the RF power by a certain amount.

[0122] In other implementations, the RF regulator 1123 may be a variable coil inductor. The RF controller 1143 may control the variable coil inductor to affect the RF power delivered to the station. In some implementations, the RF regulator is not limited to capacitors and inductors. In some implementations, other RF regulators 1123 may utilize different mechanisms (e.g., resonator circuits or resistor circuits) to vary the RF power.

[0123] The sensor 1133 measures at least one RF power parameter. The measured RF power parameter may be voltage, current, impedance, phase, or load power. Commercially available probes may be used to measure the RF power parameters and provide the measurements to the RF controller 1143. Non-RF parameters may also be measured and used as source signals for the RF controller 1143. For example, light emission from a station plasma or substrate temperature sensor may measure station characteristics and feed them to the RF controller 1143. A light emission system may be installed near each station to collect light emitted by the station plasma. The substrate temperature sensor may use a remote infrared detection system built below the substrate. The sensor 1133 may also measure multiple RF power parameters, or in some implementations, multiple sensors may be used to measure multiple RF power parameters.

[0124] In some implementations, the RF regulator may be set to a fixed value or range of values ​​in a multi-step process such as a multi-cycle ALD process, In such implementations, little or no real-time sensing of RF power parameters and adjustment of the distribution of RF power from station to station is required.

[0125] Each station 1151 includes a showerhead 1153 that works in conjunction with a grounded pedestal 1157. The power and frequency provided are sufficient to generate a plasma from the process gas, for example, approximately in the range of 50W-6,000W per station. The power level may vary depending on the implementation. RF power is connected to the station processing area via the showerhead 1153, and a plasma is generated or maintained when the RF power is applied. The plasma deposits material onto the substrate through a variety of mechanisms. For example, the plasma can cause the process gas to decompose and react on the substrate surface. In the illustrated implementation, the RF current is grounded on the pedestal 1157, which is connected to the ground 1131. In certain other implementations, the RF current can be grounded at different locations in the chamber, such as at the showerhead.

[0126] The plasma activation step of a multi-step deposition process with a short cycle duration may be brief. The duration of the plasma activation step may be about 150 milliseconds or less (e.g., about 50 milliseconds). Due to the short duration, control of the plasma concentration may affect the uniformity of the process. Plasma balancing may be used to control the plasma concentration.

[0127] Although in Figure 8-11 900, but any tool in these figures may include any features of any other tool, such as controller 950, and the controller may be configured to execute any instructions described herein for reactor 800. In addition, Figure 8-10 The chambers presented in the embodiments 800, 900, and 1000 may be considered Figure 1 Room 102 and Fig.6D and 6E In addition, RF generators 110, 610A, and 610B may also be considered Figure 8-11 Either or both of the HF RF and LF RF generators.

[0128] Unless the context of the present disclosure clearly requires otherwise, throughout the claims, the terms "comprising," "including," and the like are to be interpreted in an inclusive sense rather than an exclusive or exhaustive sense; that is, in the sense of "including, but not limited to, this." The use of singular or plural terms also generally includes the plural or singular, respectively. In addition, the terms "herein," "after," "above," "below," and similar terms refer to the application as a whole rather than to any particular part of the application. When the term "or" is used with reference to a list of two or more items, the term encompasses all of the following interpretations of the term: any one of the items in the list, all of the items in the list, and any combination of the items in the list. The term "implementation" means an implementation of the techniques and methods described herein, as well as a physical object that implements the structure and includes the techniques and / or methods described herein. The terms "substantially," "approximately," and "about" are understood to include a difference of ±10% from the indicated amount.

Claims

1. A method for detecting an abnormal plasma event, comprising: forming a plasma using a signal from an RF generator; detecting a fluctuating optical signal generated by the plasma; calculating a spectral density of the fluctuating optical signal; and A spectral density of the fluctuating light signal is determined to differ by a threshold amount from one or more reference spectral densities of fluctuating light signals generated by the plasma.

2. The method of claim 1, wherein calculating the spectral density of the fluctuating optical signal comprises: utilizing a Fast Fourier Transform.

3. The method according to claim 1, wherein: The one or more reference spectral densities of the fluctuating optical signal correspond to spectral densities of the fluctuating optical signal produced by the plasma maintained under nominal conditions.

4. The method of claim 3, wherein the threshold amount corresponds to at least one standard deviation relative to the spectral density of the fluctuating optical signal from the plasma maintained under nominal conditions.

5. The method of claim 1, wherein calculating the spectral density of the fluctuating light signal comprises determining the spectral density of the fluctuating light signal at a frequency of approximately 400 kHz or at a frequency between approximately 400 kHz and approximately 4 MHz.

6. A device comprising: a signal processor configured to analyze an output signal from a light detector that receives light signals from the multi-station processing tool, the signal processor further configured to: detecting a fluctuating optical signal generated by a plasma formed in a semiconductor processing chamber; calculating a spectral density of the fluctuating optical signal; and A spectral density of the fluctuating light signal is determined to differ by a threshold amount from one or more reference spectral densities of the fluctuating light signal generated by the plasma.

7. The apparatus of claim 6, wherein the spectral density of the fluctuating optical signal is calculated when an RF signal having a frequency of approximately 400 kHz is coupled to the multi-station processing tool.

8. The apparatus of claim 6, wherein the spectral density of the fluctuating optical signal is calculated to determine the spectral density of the fluctuating optical signal at a frequency between 400 kHz and 4 MHz.

9. The apparatus of claim 6, further comprising an RF generator configured to generate the plasma in the semiconductor processing chamber.

10. The apparatus of claim 6, wherein the one or more baseline spectral densities of the fluctuating optical signal are equivalent to spectral densities calculated when the plasma is maintained at nominal conditions.

11. The apparatus of claim 10, wherein the threshold amount is equivalent to one standard deviation relative to the spectral density of the fluctuating optical signal from the plasma maintained under nominal conditions.

12. The apparatus of claim 11, wherein the threshold amount is equivalent to two standard deviations relative to the spectral density of the fluctuating optical signal from the plasma maintained under nominal conditions.

13. The apparatus of claim 12, wherein the spectral density of the fluctuating optical signal is calculated at a frequency of an RF signal coupled to the multi-station processing tool.

14. The apparatus according to claim 6, wherein: The spectral density of the fluctuating optical signal is calculated using a Fast Fourier Transform.

15. A semiconductor processing tool comprising: Semiconductor processing chambers; a light detector configured to detect a light signal indicative of plasma luminescence in the semiconductor processing chamber, the light detector configured to convert the light signal into an analog voltage signal; and A slope change detection unit configured to: receiving the analog voltage signal, filtering the analog voltage signal to form a low-pass filtered voltage signal, and The low pass filtered voltage signal is compared to the analog voltage signal to determine whether a change associated with the low pass filtered voltage signal exceeds a threshold.

16. The semiconductor processing tool of claim 15, further comprising: a lens on the semiconductor processing chamber configured to allow light within the semiconductor processing chamber to pass through the lens, and A fiber optic cable is interposed between the lens and the light detector and is configured to transmit at least a portion of the light passing through the lens to the light detector.

17. The semiconductor processing tool of claim 15, wherein the slope change detection unit further comprises a comparator for comparing a change between the low pass filtered voltage signal and the analog voltage signal.

18. The semiconductor processing tool of claim 15, wherein the slope change detection unit further comprises: a differentiator configured to obtain a derivative of the analog voltage signal to convert the analog voltage signal into an adjusted voltage signal, and A comparator is configured to determine whether a change in the regulated voltage signal exceeds the threshold.

19. The semiconductor processing tool of claim 15, further comprising: A fiber optic cable between the semiconductor processing chamber and the light detector is configured to transmit light generated by a plasma in the semiconductor processing chamber to the light detector.

20. The semiconductor processing tool of claim 15, further comprising an RF generator configured to provide RF power to the semiconductor processing chamber to generate and sustain a plasma.

21. The semiconductor processing tool of claim 20, wherein: The slope change detection unit is further configured to cause a signal to be sent to the RF generator in response to the determination, wherein the signal is configured to cause an adjustment to the RF generator.

22. The semiconductor processing tool of claim 21, wherein the RF generator is configured to reduce RF power from a first power level to a second power level in response to receiving the signal from the slope change detection unit.

23. The semiconductor processing tool of claim 20, wherein: The photodetector is electrically connected to the RF generator, and The optical signal from the semiconductor processing chamber is fed from the semiconductor processing chamber to the RF generator.

24. The semiconductor processing tool according to claim 20, wherein the slope change detection unit further comprises a comparator, wherein: is configured to determine whether a change in the low-pass filtered analog voltage signal exceeds a threshold, and A transistor-transistor logic (TTL) is included, which is configured to send a TTL signal to the RF generator.

25. The semiconductor processing tool of claim 20, wherein the light detector and the slope change detection unit are disposed between the semiconductor processing chamber and the RF generator.

Images