Plasma treatment device and plasma treatment method

Abstract

Provided is a plasma treatment device comprising: a first high-frequency power source that supplies a first high-frequency power for generating plasma; a second high-frequency power source that supplies a second high-frequency power to a sample stand on which a sample is placed; and a control unit that controls the first high-frequency power source and the second high-frequency power source. The plasma treatment device is connected to a server. The server is configured so that: upon input of a recipe which is a plasma treatment condition, a sensor value of the plasma treatment device is measured by changing a parameter defined in the recipe; an abnormal value is determined on the basis of the relationship of a time variation between the parameter and the measured sensor value; a model is generated by learning the relationship between the recipe and the abnormal value; upon creation of a recipe, the abnormality of the parameter is inferred using the generated model; and a recipe in which an input range of the parameter is limited is newly created on the basis of the input range of the parameter inferred to be abnormal. The control unit is configured so that the first high-frequency power source and the second high-frequency power source are controlled on the basis of the newly created recipe.

Description

Plasma Processing Apparatus and Plasma Processing Method

[0001] The present disclosure relates to a plasma processing apparatus and a plasma processing method, and more particularly to a plasma processing apparatus and a plasma processing method for processing a substrate-shaped sample such as a semiconductor wafer disposed in a processing chamber in a vacuum chamber using plasma formed in the processing chamber, such as etching.

[0002] For the performance improvement of semiconductor devices, miniaturization of the structure and high integration by adopting a three-dimensional structure are progressing. In a plasma processing apparatus, in order to meet diversified processing needs, various adjustment knobs provided in the apparatus are adjusted to control physical parameters such as ions, radical fluxes, and their energies required for processing.

[0003] Japanese Patent Application Laid-Open No. 2021-141010

[0004] In a plasma processing apparatus, while coping with various processing needs, it is essential to ensure stability during mass production so as not to cause problems in mass production of semiconductor devices. The instability of plasma during processing causes problems related to mass productivity, such as the generation of foreign matter and contamination associated with sudden abnormal discharge at unintended locations and the deterioration of reproducibility during processing. Therefore, in order to avoid these risks, it is necessary to avoid conditions with unstable discharge. As a method for discriminating unstable discharge, for example, Patent Document 1 acquires the amount of change in the reflected wave of the microwave power supplied to the plasma processing apparatus and the peak value of the RF bias power, and determines whether the amount of change satisfies the allowable conditions. However, Patent Document 1 does not assume real-time abnormal determination and does not assume detecting plasma instability in advance, so it is not possible to detect plasma instability in advance.

[0005] In order to meet various processing needs for plasma processing, the number of input parameters of the plasma processing apparatus has become complicated, and it is difficult to comprehensively verify whether the plasma is stable for each condition. In addition to detecting abnormalities, it is required to discriminate plasma instability in advance.

[0006] This disclosure provides a technology that enables plasma processing while avoiding plasma discharge instability conditions. Other challenges and novel features will become apparent from the description herein and the accompanying drawings.

[0007] A brief overview of some of the representative disclosures is as follows:

[0008] According to one embodiment, the plasma processing apparatus comprises a first high-frequency power supply that supplies first high-frequency power for generating plasma, a second high-frequency power supply that supplies second high-frequency power to a sample stage on which a sample is placed, and a control unit that controls the first high-frequency power supply and the second high-frequency power supply, and the plasma processing apparatus is connected to a server. When a recipe, which is the plasma processing condition, is input by the server, sensor values ​​are measured by changing the parameters specified in the recipe, abnormal values ​​are determined based on the relationship between the parameters and the measured sensor values ​​over time, a model is generated by learning the relationship between the recipe and the abnormal values, when the recipe is created, abnormalities in the parameters are inferred using the generated model, a new recipe is created in which the input range of the parameters is restricted based on the input range of the parameters inferred to be abnormal, and the first high-frequency power supply and the second high-frequency power supply are controlled by the control unit based on the newly created recipe.

[0009] Furthermore, according to another embodiment, a plasma treatment method for plasma treatment of a sample includes the steps of: measuring sensor values ​​by changing parameters defined in a recipe when inputting a recipe which is a plasma treatment condition; determining abnormal values ​​based on the relationship between the parameters and the measured sensor values ​​over time; generating a model by learning the relationship between the recipe and the abnormal values; inferring abnormalities in the parameters using the generated model when creating the recipe; creating a new recipe in which the input range of the parameters is restricted based on the input range of the parameters inferred to be abnormal; and plasma treatment of the sample using the newly created recipe.

[0010] According to the above technology, by predicting the plasma discharge stability before plasma processing, plasma processing can be performed while avoiding unstable plasma discharge conditions.

[0011] Cross-sectional view of the etching apparatus according to the first embodiment. Diagram illustrating the microwave supply according to the first embodiment. Diagram illustrating the RF power supply according to the first embodiment. Plasma processing apparatus system according to the first embodiment. Flowchart for learning plasma instability according to the first embodiment. Diagram illustrating the display screen when entering a recipe on the PC according to the first embodiment. Flowchart illustrating plasma processing in the first embodiment.

[0012] The following examples will be described with reference to the drawings. However, in the following description, the same reference numerals will be used for identical components, and repeated explanations may be omitted. In addition, the drawings may be more schematic than the actual embodiments in order to make the explanation clearer, but they are merely examples and do not limit the interpretation of the present invention.

[0013] Figure 1 shows a cross-sectional view of an etching apparatus, which is a plasma processing apparatus 100 according to a first embodiment of the present invention. Microwaves (high-frequency power) are oscillated from a microwave source 1 and transmitted to a circular waveguide 6 via an isolator 2, an automatic matching unit 3, and a rectangular waveguide 4, through a circular-to-rectangular converter 5. The microwave source 1 can be rephrased as a first high-frequency power supply 1 that supplies the first high-frequency power for generating plasma. In this embodiment, a 2.45 GHz microwave, which is commonly used in industry, is used. The isolator 2 is used to protect the microwave source 1 from reflected microwaves, and the automatic matching unit 3 is used to adjust the load impedance and suppress reflected waves to efficiently supply microwaves.

[0014] The microwave can be repeatedly switched between ON and OFF states by pulse driving at a predetermined frequency. The pulse frequency is adjusted within the range of 100 Hz to 10 kHz. The duty cycle, which is the ratio of the ON state time to the pulse period, is also adjusted. By adjusting these parameters (microwave pulse frequency and duty cycle), the degree of plasma dissociation is adjusted.

[0015] Here, in order to facilitate the handling of microwave propagation phenomena, the dimensions of the waveguide cross-sections are defined such that only the fundamental mode TE10 mode propagates in the rectangular waveguide 4, and only the fundamental mode TE11 mode propagates in the circular waveguide 6.

[0016] Microwaves propagate through the cavity 7 and are introduced into the plasma processing chamber 10 through the microwave introduction window 8 and the shower plate 9. The cavity 7 uses a conductor as a material that reflects microwaves. For example, aluminum is used as the material for the cavity 7.

[0017] To form the magnetic field necessary for ECR (Electron Cyclotron Resonance), an upper coil 11, a middle coil 12, and a lower coil 13 are installed outside the plasma processing chamber 10. The plasma generation position can be controlled by adjusting the coil currents of the upper coil 11, middle coil 12, and lower coil 13 using a coil power supply 111, which includes the upper coil power supply UCS, the middle coil power supply MCS, and the lower coil power supply LCS.

[0018] The yoke 14 serves to prevent magnetic field leakage to the outside of the plasma processing chamber 10 and also to adjust the shape of magnetic field lines inside the plasma processing chamber 10. To protect the side walls of the plasma processing chamber 10 from the plasma, an inner cylinder 15 is installed inside the side walls of the plasma processing chamber 10. Quartz is used as a material with high plasma resistance for the inner cylinder 15 located near the plasma. Alternatively, materials with high plasma resistance such as yttria, alumina, yttrium fluoride, aluminum fluoride, or aluminum nitride may be used.

[0019] As the material for the microwave introduction window 8 and shower plate 9, quartz is used as a material that transmits microwaves and is transparent. Alternatively, other dielectric materials that transmit microwaves and are transparent may be used.

[0020] Gas is supplied between the microwave introduction window 8 and the shower plate 9 from a gas supply means 112. The gas supply means 112 includes a function to supply a desired flow rate by a mass flow controller. The type of gas used is appropriately selected according to the film to be processed, and multiple types of gas are supplied in combination at a predetermined flow rate. Multiple gas supply holes 16 are provided in the shower plate 9, and the gas is supplied to the plasma processing chamber 10 through the gas supply holes 16. The supplied gas is evacuated by a turbomolecular pump 18 via a conductance control valve 17.

[0021] At the bottom of the plasma processing chamber 10 is a substrate stage / high-frequency electrode 20 on which the substrate 19 to be processed as a sample is placed, and below it is an insulating plate 21. The substrate stage / high-frequency electrode 20 is supplied with bias power (RF power) from an RF bias power supply (RF power supply) 22 via an automatic matching unit 23. The substrate stage / high-frequency electrode 20 can be simply referred to as a sample stand 20. The RF bias power supply 22 can also be referred to as a second high-frequency power supply that supplies second high-frequency power to the sample stand 20 on which the sample 19 is placed. The RF bias power supply 22 is adjusted so that the desired etching can be performed, and the energy of the ions incident on the substrate 19 to be processed is controlled by adjusting the bias power from the RF bias power supply 22. The bias power is pulse-modulated, similar to microwave power, and the ion energy is controlled by adjusting the pulse frequency and duty cycle. In addition, a grounding electrode 24 is provided at the bottom of the plasma processing wall to provide grounding (ground potential) for the RF power supplied to the substrate stage / high-frequency electrode 20.

[0022] The substrate stage and high-frequency electrode 20 is equipped with an adsorption mechanism for the substrate 19 to be processed (not shown) and a temperature control means, and the temperature of the substrate 19 to be processed is adjusted as needed. A susceptor 25 and a stage cover 26 are installed to protect the outer periphery of the substrate stage and high-frequency electrode 20 from plasma.

[0023] A camera 27 is provided in the plasma processing apparatus 100. In this example, the camera 27 is installed on the ceiling above the cavity 7 and is provided to observe the time variation and distribution of the brightness of the plasma emission generated in the plasma processing chamber 10. The camera 27 is an emission measuring device that measures the emission of plasma, and the time variation and distribution of the brightness of the plasma emission measured by the plasma emission measuring device are used as brightness information (emission intensity information), and the brightness information is represented as a single sensor value.

[0024] The plasma processing apparatus 100 is provided with a device control unit 115. The device control unit 115 can simply be referred to as the control unit 115. The device control unit 115 performs the following controls, for example.

[0025] 1) Control the on / off state of microwave source 1 and the power value of high-frequency power.

[0026] 2) The coil currents of the upper coil 11, middle coil 12, and lower coil 13 are adjusted by controlling the upper coil power supply UCS, middle coil power supply MCS, and lower coil power supply LCS of the coil power supply 111.

[0027] 3) The mass flow controller of the gas supply means 112 is controlled to control the type of gas to be processed and its flow rate.

[0028] 4) Control the on / off state of the RF bias power supply 22 and the power value of the bias power (RF power: high-frequency power).

[0029] 5) The temperature of the substrate to be processed 19 is adjusted by controlling the adsorption mechanism of the substrate to be processed 19 and by controlling the temperature control means.

[0030] 6) The exhaust speed is controlled by the conductance control valve 17.

[0031] 7) The vacuum level is controlled by the turbomolecular pump 18.

[0032] Furthermore, the device control unit 115 acquires the load gamma (reflection coefficient) and reflected power calculated by the automatic matching unit 3, acquires the time variation of the voltage amplitude (voltage amplitude value: peak-to-peak voltage) detected by the automatic matching unit 23 when RF bias power is applied and the reflected power, acquires observation images and brightness information of plasma emission obtained by the camera 27, acquires sensor values ​​from various sensors (not shown), and transmits log data LGD based on the acquired data and various sensor values. In this specification, the various data and various sensor values ​​included in the log data LGD are sometimes collectively referred to as sensor values.

[0033] Furthermore, as will be described later, the device control unit 115 receives a plasma processing instruction 58 which includes a recipe instruction 56 and a processing device instruction 57, which are recipes created from the recipe management server 50.

[0034] The device control unit 115 performs the control described in 1)-7) above based on the received recipe, so that the plasma processing apparatus 100 performs plasma processing according to the received recipe.

[0035] In etching, an example of plasma processing, a plasma 40 is generated by ECR on an ECR surface formed by microwaves supplied from a microwave source 1 and the upper coil 11, middle coil 12, and lower coil 13. The ions and radicals generated therein are then irradiated onto the substrate 19 to be processed to perform etching. In this embodiment, the plasma generation method using ECR has been described, but other plasma generation methods using high-frequency power, such as ICP (inductively coupled plasma) or CCP (capacitively coupled plasma), may also be used. In the case of ICP and CCP, high-frequency power of several MHz to tens of MHz is used as the high-frequency source for plasma generation, rather than microwaves.

[0036] Next, using Figures 2A and 2B, we will explain examples of configurations for microwave (high-frequency power) pulse waveforms and RF bias power (RF power: high-frequency power) pulse waveforms. Figure 2A is a diagram illustrating the microwave supply according to the first embodiment. Figure 2B is a diagram illustrating the RF power supply according to the first embodiment.

[0037] As shown in Figure 2A, the microwave can be set to 1st (first state), 2nd (second state), and OFF state (off state) in the microwave power supply 1 and supplied in pulse modulated form. In other words, the microwave power supply 1 is a pulse-driven high-frequency power supply that can output multiple stages of power (1st, 2nd, OFF state). By setting the frequencies f1 (= 1 / period (T1)) of the 1st and 2nd states, the duty cycle of the 1st and 2nd states, and the delay time td1 representing the start time of the 2nd state (the difference time between the start time of the 1st state and the start time of the 2nd state), a pulse waveform like that shown in Figure 2A can be represented. The 1st (first state) of the microwave is a first power pulse P11 having a first microwave power MWP1. The 2nd (second state) of the microwave is a second power pulse P12 having a second microwave power MWP2 (MWP2 < MWP1) which is smaller than the first microwave power MWP1. The duty cycle of the first power pulse P11 can be calculated using the width W11 of the first power pulse P11, the width W12 of the OFF state following the first power pulse P11, the width W13 of the second power pulse P12, and the width W14 of the OFF state following the second power pulse P12 (duty cycle DR11 of the first power pulse P11 = W11 / (W11 + W12 + W13 + W14)). The duty cycle of the second power pulse P12 can be calculated similarly (duty cycle DR12 of the second power pulse P12 = W13 / (W11 + W12 + W13 + W14)).

[0038] In other words, the first high-frequency power supply, the microwave power supply 1, is configured to supply high-frequency power modulated by pulses (P11, P12) having multiple periods with amplitudes greater than zero (0). The microwave power supply 1 can be controlled by parameters including the pulse-modulated high-frequency power value for each period with amplitudes greater than zero (0), the pulse frequency, and the pulse duty cycle.

[0039] As shown in Figure 2B, similar to the microwave in Figure 2A, the RF bias power can also be pulse-modulated in 1st, 2nd, and OFF states, allowing for waveform settings different from those of the microwave. The RF bias power can be set to 1st (first state), 2nd (second state), and OFF state (off state) in the RF bias power supply 22 and supplied in pulse-modulated form. In other words, the RF bias power supply 22 is a pulse-driven RF bias power supply that can output multiple power levels (1st, 2nd, and OFF states). By setting the frequencies f2 (= 1 / period (T2)) of the 1st and 2nd states, the duty cycle of the 1st and 2nd states, and the delay time td2 representing the start time of the 2nd state (the difference between the start time of the 1st state and the start time of the 2nd state), a pulse waveform like that shown in Figure 2B can be represented. The 1st (first state) of the RF bias power is a first RF power pulse P21 having a first RF bias power RFP1. The second state of the RF bias power is the second RF power pulse P22, which has a second RF bias power RFP2 (RFP2 > RFP1) greater than the first RF bias power RFP1. The duty cycle of the first RF power pulse P21 can be calculated using the width W21 of the first RF power pulse P21, the width W22 of the OFF state following the first RF power pulse P21, the width W23 of the second RF power pulse P22, and the width W24 of the OFF state following the second RF power pulse P22 (duty cycle DR21 of the first RF power pulse P21 = W21 / (W21 + W22 + W23 + W24)). The duty cycle of the second RF power pulse P22 can be calculated similarly (duty cycle DR22 of the second RF power pulse P22 = W23 / (W21 + W22 + W23 + W24)).

[0040] In other words, the second high-frequency power supply, the RF bias power supply 22, is configured to supply high-frequency power modulated by pulses (P21, P22) having multiple periods with amplitudes greater than zero (0). The RF bias power supply 22 can be controlled by parameters including the pulse-modulated high-frequency power value for each period with amplitudes greater than zero (0), the pulse frequency, and the pulse duty cycle.

[0041] The microwave output contributes to the increase and decrease of the plasma density, and the RF bias power contributes to the increase and decrease of the ion energy incident on the substrate. Also, during the OFF time of the microwave, the plasma density decays, resulting in a low electron temperature and a state with many negative ions and radicals.

[0042] Thus, by combining various microwave power waveforms and RF bias waveforms, the ratios of radicals and ions in the gas phase, the fluxes of various radicals incident on the processing substrate 19, the ion flux, the ion energy, etc. can be changed, and the desired etching process in a fine pattern can be realized.

[0043] However, depending on the combination of pulse waveforms, the plasma density may become unstable during microwave power modulation. Also, the RF bias power also has a significant impact on the plasma density. To detect this plasma instability, the state of the plasma load is measured and observed through the time variation of the measured value of the load gamma (reflection coefficient) calculated in the microwave automatic matcher 3, the reflection power, the time variation of the amplitude of the voltage (voltage amplitude value: peak-to-peak voltage) when the RF bias power is applied detected in the automatic matcher 23, the reflection power, and the time variation and distribution (luminance information) of the luminance of the plasma emission obtained by the camera 27. That is, these values (the reflection coefficient and reflection power of the microwave, the voltage amplitude value and reflection power of the RF bias power, the luminance information of the plasma emission) are acquired as sensor values provided in the plasma processing apparatus 100 and used for detecting and determining the plasma instability. The acquired sensor values are transmitted to the recipe management server 50 described later as log data LGD.

[0044] In this embodiment, the pulse waveform is expressed in a form defined by three states: 1st, 2nd, and OFF state. However, this is merely an example, and as the method of applying the pulse waveform, the states after the 3rd and the waveform of the pulse may not be rectangular but sine waves or triangular waves, and a configuration in which the rise and fall times of the pulse can be set may also be used.

[0045] Next, the system components of the plasma processing apparatus 100 will be explained using Figure 3. The plasma processing apparatuses 41 to 4N are connected to the recipe management server 50. In addition, the recipe input PCs 61 to 6n, which are information terminals, are connected to the recipe management server 50. When performing plasma processing, users operate the recipe input PCs (61 to 6n) to create a recipe in advance, specifying which plasma processing apparatus (41 to 4N) to use and under what plasma processing conditions to perform the plasma processing.

[0046] In the following section, as an example, we will explain the case in which a user creates a new plasma processing recipe for the plasma processing device 4N using the recipe input PC 6n. The recipe management server 50 explicitly displays a recommended range 55 of parameters that can be entered into the recipe to the recipe input PC 6n. This recommended range 55 is derived through learning and inference of log data LGD related to discharge instability, which will be described later. The user creates a new recipe on the recipe input creation screen displayed on the display screen of the recipe input PC 6n, referring to the recommended range 55 of inputtable parameters displayed on the recipe input creation screen. The newly created recipe is sent from the recipe input PC 6n to the recipe management server 50 as a recipe instruction 56 and stored in the storage 51 of the recipe management server 50. The recipe input PC 6n also sends a processing instruction 57 to the recipe management server 50 specifying the plasma processing device 4N that will process the newly created recipe. The recipe management server 50 then sends a plasma processing instruction 58, including the newly created recipe, the recipe instruction 56, and the processing instruction 57, to the specified plasma processing device 4N from the storage 51. This allows the designated plasma processing apparatus 4N to perform plasma processing based on the created recipe.

[0047] Furthermore, the recipe input PCs (61-6n) can issue learning instructions 59 to learn about the discharge instability described later. The recipe management server 50 is equipped with a computing unit 53 for creating a learning model. In addition, log data LGD of various sensor values ​​during plasma processing is transmitted from the plasma processing equipment (41-4N) to the recipe management server 50 and stored, for example, in the storage 51 of the recipe management server 50.

[0048] Next, the flow of learning discharge instability will be described with reference to FIG. 4. First, as shown in FIG. 3, the user gives a learning instruction (59) for learning discharge instability from the recipe input PC (6n) to the recipe management server 50. Thereby, the learning flow in FIG. 4 is started.

[0049] First, in step A28, the user inputs the central conditions as the reference for the etching processing conditions into the recipe input PC (6n).

[0050] Next, in step B29, the user designates, using the recipe input PC (6n), parameters related to plasma impedance fluctuations, that is, discharge instability. Here, as an example of the parameters related to discharge instability, the set value of the pulse waveform is used. Therefore, as parameters, the user designates the power values (MWP1, MWP2) of each pulse of the microwave (high-frequency power), the frequency of the pulse (f1 = 1 / T1), the duty ratio (DR11, DR12), and the phase delay time td1, and the RF power values (RFW1, RFW2) of each pulse of the RF bias power, the frequency of the pulse (f2 = 1 / T2), the duty ratio (DR21, DR22), and the phase delay time td2, etc. The user also designates the change range of the parameters to be changed during learning. These input data are transmitted from the recipe input PC (6n) to the plasma processing apparatus (4N) via the recipe management server 50.

[0051] Next, in step C30, the plasma processing apparatus (4N) performs plasma processing while changing the designated parameters within the designated range according to the instruction from the recipe input PC (6n). At this time, during plasma processing, a dummy wafer simulating actual mass production is placed on the substrate stage and high-frequency electrode 20 of the plasma processing apparatus (4N). For example, when the film to be processed during device manufacturing is silicon (Si), a Si wafer (a semiconductor substrate made of silicon) is placed on the substrate stage and high-frequency electrode 20 of the plasma processing apparatus (4N). Also, assuming the aperture ratio of the mask during device mass production, a wafer with a partial area of a wafer fragment masked may be processed.

[0052] In process D31, various sensor values ​​from the plasma processing apparatus (4N) during plasma processing are transmitted from the plasma processing apparatus (4N) to the recipe management server 50 and stored in the recipe management server 50. After plasma processing, data related to the amount of variation in sensor values ​​during plasma processing is also transmitted from the plasma processing apparatus (4N) to the recipe management server 50. For example, the standard deviation of the load gamma during the discharge time is transmitted as an indicator of the amount of time variation of the load gamma calculated by the automatic matching unit 3.

[0053] In step E32, the computing unit 53 of the recipe management server 50 determines the instability of the discharge from the obtained measurement data. The determination of the instability of the discharge is performed by performing predetermined calculations and analyses on the sensor values ​​provided in the plasma processing device (4N). For example, it can be determined that the discharge is abnormal when the amount of temporal variation of the acquired sensor value exceeds a threshold. Step E32 can be considered a determination step in which abnormal values ​​are determined from the relationship between the parameters and the temporal variation of the sensor values.

[0054] It can be difficult to uniquely determine a threshold for discharge instability. For example, a method that uses a threshold for the time variation of load gamma is suitable for extracting macroscopic fluctuations of the plasma within a plasma processing device (4N). On the other hand, if the discharge is stable macroscopically within the plasma processing device (4N), but there is a localized abnormal discharge, the time variation of load gamma will be calculated as small, which may cause the localized abnormal discharge to be overlooked.

[0055] Therefore, it is necessary to observe the discharge using not only load gamma but also other methods such as optical means to determine discharge instability. As a means of identifying abnormal values ​​from such a large number of sensor values, a pattern recognition model using unsupervised learning, such as a support vector machine, may be used. By accumulating sensor data, it is possible to distinguish between stable and unstable discharge conditions from various sensor values. Here, the arithmetic unit 53 of the recipe management server 50 stores the relationship between the obtained recipe and the abnormal value in the storage 51 of the recipe management server 50 (storage step).

[0056] Next, in step F33, the computing unit 53 of the recipe management server 50 constructs a learning model that infers plasma instability for unknown recipes using the relationship between the obtained recipes and plasma instability. Step F33 can be considered a learning step. As a method of inference, for example, a deep learning model is used. The recipe is used as the input layer, the plasma instability determination result as the output layer, and a neural network is constructed considering the hidden layer. The created learning model is centrally managed by the recipe management server 50. The above is the flow of learning discharge instability. In this embodiment, a neural network model was used as one example, but other machine learning models may also be used.

[0057] Finally, we will describe the plasma treatment using Figures 5 and 6. Figure 5 is a diagram illustrating the display screen when entering recipes on the recipe input PC. Figure 6 is a flowchart illustrating the plasma treatment in the first embodiment.

[0058] When a user enters a recipe into the recipe input PC, the input parameter range in which plasma instability is predicted, that is, the parameter input range that is likely to be abnormal, is clearly displayed on the DISP screen of the recipe input PC, based on the discharge instability determination inferred by the aforementioned learning model.

[0059] As shown in Figure 5, the display screen DISP of the recipe input PC 6i (I=1 to n) has, for example, a recipe input area IND and a prohibited parameter range display area PPRD displayed to the right of the input area IND. The user inputs the recipe in the input area IND. The prohibited parameter range display area PPRD displays parameter input ranges that are likely to result in errors.

[0060] In this example, the display area PPRD shows that the input range most likely to result in an abnormality for parameter A is a1-a2, and the input range most likely to result in an abnormality for parameter B is b1-b2. Therefore, when the user inputs the recipe RCP into the input area IND, they should input the range a3-a4 for parameter A, excluding the input range a1-a2 which is most likely to result in an abnormality. Similarly, they should input the range b3-b4 for parameter B, excluding the input range b1-b2 which is most likely to result in an abnormality. Parameter A schematically represents, for example, the power value of each microwave pulse, pulse frequency, duty cycle, and phase delay time, while parameter B schematically represents, for example, the power value of each RF bias power pulse, pulse frequency, duty cycle, and phase delay time.

[0061] In other words, by inputting and setting a parameter range that avoids these conditions (parameter input ranges that are likely to result in abnormalities) into the recipe, users can create a new RCPN recipe that allows plasma processing to be performed while avoiding conditions that would cause plasma instability. Specifically, the discharge instability is inferred using an inference model (learning model) generated using the correlation between the parameters input into the recipe and discharge instability, and the input range of parameters when creating the recipe is restricted based on the inferred discharge instability (parameter input ranges that are likely to result in abnormalities).

[0062] This RegiP RCPN is supplied to the designated plasma processing apparatus (100 in Figure 1, 41-4N in Figure 4), and the substrate 19 to be processed is subjected to plasma processing. If a parameter input range that matches a parameter input range that is likely to be abnormal, as displayed in the prohibited parameter range display area PPRD, or a parameter input range that includes a parameter input range that is likely to be abnormal, exists in the recipe's input area IND, it is preferable to display a warning on the display screen DISP to alert the user and prompt them to re-enter the data. Alternatively, if a parameter input range that matches a parameter input range that is likely to be abnormal, as displayed in the prohibited parameter range display area PPRD, or a parameter input range that includes a parameter input range that is likely to be abnormal, exists in the recipe's input area IND, it is also possible to generate only a warning and allow the user's input to remain as is.

[0063] Therefore, the plasma treatment according to Recipe RCPN is a plasma treatment that is prepared in advance to avoid conditions that would cause plasma instability, and a plasma treatment method is provided that can perform plasma treatment while avoiding discharge instability conditions.

[0064] The flowchart in Figure 6 shows the process of inputting a recipe RCPN created by the method described in Figures 4 and 5 from the recipe management server 50 to the target plasma processing apparatus 100, and then performing plasma processing on the sample using the plasma processing apparatus 100 controlled by the recipe RCPN. The plasma processing method will now be explained using Figure 6, based on each step (process) 34-38.

[0065] Step 34: The user enters the recipe processing conditions into the recipe input area IND on the DISP display screen of the recipe input PC6i (I=1 to n). The processing conditions include input of various parameters.

[0066] Step 35: When a user inputs a recipe into the recipe input PC 6i (I=1 to n), the recipe management server 50, based on inference from the learned data, clearly indicates the range of inputs for the specified parameters in the prohibited parameter range display area PPRD on the DISP display screen of the recipe input PC. In other words, the recipe management server 50 displays the input parameter range in which plasma instability is predicted, that is, the parameter input range that is likely to be abnormal, in the display area PPRD, based on the discharge instability determination inferred by the learning model as learning data.

[0067] Step 36: The user modifies or changes each parameter range in the recipe while referring to the parameter input ranges that are likely to cause abnormalities, as displayed in the PPRD display area. At this point, the recipe management server 50 checks whether all parameters are within the input range. In other words, it checks whether the input range for each parameter is not within the range of parameters for which plasma instability is predicted. If the input range for each parameter is not within the range of parameters for which plasma instability is predicted (YES), the system proceeds to step 37, assuming that a new recipe RCPN has been created by the user. If the input range for a parameter is within the range of parameters for which plasma instability is predicted (NO), a warning is displayed on the DISP display screen, and the system proceeds to step 38.

[0068] Step 37: The new recipe RCPN created by the user is sent to the recipe management server 50, and from the recipe management server 50 it is sent to the device control unit 115 of the target plasma processing apparatus 100. The device control unit 115 then controls the operation of the microwave source 1 and bias power supply 22 of the plasma processing apparatus 100 based on the new recipe RCPN to stably generate plasma in the processing chamber 10 and perform plasma processing on the substrate 19, which is the sample placed on the substrate stage / high-frequency electrode 20.

[0069] Step 38: Here, a final confirmation is made as to whether or not to perform plasma processing using a new recipe RCPN2 where the input range of certain parameters is within the range of parameters for which plasma instability is predicted.

[0070] If the final confirmation is to perform the plasma treatment based on recipe RCPN2 (YES), proceed to step 37, and the plasma treatment using recipe RCPN2 will be performed. In this case, although the plasma treatment using recipe RCPN2 will partially include a state of plasma instability, it is recognized that this recipe will not cause significant problems in terms of plasma treatment quality. If problems occur in the plasma treatment using recipe RCPN2, it is advisable to repeat steps 34-36.

[0071] If the final confirmation is that the plasma processing based on recipe RCPN2 will not be performed (NO), the process proceeds to step 34, where the processing conditions for the recipe are entered again.

[0072] The above description of the embodiment can be summarized as follows for the plasma processing apparatus 100.

[0073] The plasma processing apparatus 100 comprises a first high-frequency power supply 1 that supplies first high-frequency power for generating plasma, a second high-frequency power supply 22 that supplies second high-frequency power to a sample stage 20 on which a sample 19 is placed, and a device control unit 115 that controls the first high-frequency power supply 1 and the second high-frequency power supply 22, and is connected to a recipe management server 50.

[0074] The following steps 1)-5) are performed by the recipe management server 50. 1) When a recipe, which is the plasma processing condition, is input, the sensor values ​​of the plasma processing device 100 are measured by changing the parameters specified in the recipe. 2) Anomalies are determined based on the relationship between the parameters and the measured sensor values ​​over time. 3) A learning model is generated by learning the relationship between the recipe and the anomalies. 4) When a new recipe RCPM is created, anomalies in the parameters are inferred using the generated learning model. 5) A new recipe RCPM is created with a restricted input range for the parameters based on the input range of the parameters inferred to be anomalies.

[0075] Then, the device control unit 115 performs the following 6). 6) The first high-frequency power supply 1 and the second high-frequency power supply 22 are controlled based on the newly created recipe RCPM.

[0076] Furthermore, the above description can be summarized as follows as a plasma processing method: 1) A step (28-31) in which, when inputting a recipe which is the plasma processing condition, the sensor values ​​of the plasma processing apparatus 100 are measured by changing the parameters specified in the recipe; 2) A step (32) in which abnormal values ​​are determined based on the relationship between the parameters and the measured sensor values ​​over time; 3) A step (33) in which a learning model is generated by learning the relationship between the recipe and the abnormal values; 4) A step (34) in which abnormalities in the parameters are inferred using the generated model when creating a recipe; 5) A step (35, 36) in which a new recipe RCPN is created in which the input range of the parameters is restricted based on the input range of the parameters inferred to be abnormal; and 6) A step (37) in which the microwave source 1 and bias power supply 22 of the plasma processing apparatus 100 are controlled using the newly created recipe RCPN to perform plasma processing on the sample placed on the substrate stage and high-frequency electrode 20.

[0077] Here, the newly created recipe RCPM is set to a range of values ​​for the parameter excluding the input range of the parameter that was inferred to be abnormal. The sensor value includes the reflection coefficient or reflected power of the first high-frequency power supplied by the first high-frequency power supply 1 to generate the plasma, or the peak-to-peak voltage of the second high-frequency power supplied by the second high-frequency power supply 22 to the sample stage 21 on which the sample 19 is placed. The parameter includes the pulse-modulated high-frequency power value for each of the periods, the frequency of the pulse, and the duty cycle of the pulse.

[0078] According to the plasma processing apparatus and plasma processing method described above, by predicting the plasma discharge stability before plasma processing, it is possible to perform plasma processing on a sample while avoiding unstable plasma discharge conditions.

[0079] The disclosure made by the Discloser has been described in detail based on the examples above, but it goes without saying that the disclosure is not limited to the above examples and can be modified in various ways.

[0080] This disclosure is applicable to plasma processing equipment used to process samples on substrates such as semiconductor wafers by etching or other methods.

[0081] 1: Microwave source (first high-frequency power supply), 2: Isolator, 3: Automatic matching unit, 4: Rectangular waveguide, 5: Circular-rectangular converter, 6: Circular waveguide, 7: Cavity, 8: Microwave introduction window, 9: Shower plate, 10: Plasma processing chamber, 11: Upper coil, 12: Middle coil, 13: Lower coil, 14: Yoke, 15: Inner cylinder, 16: Gas supply port, 17: Conductance adjustment valve, 18: Turbomolecular pump, 19: Substrate to be processed (sample), 20: Substrate stage and high-frequency electrode (sample stage), 21: Insulating plate, 22: Bias power supply (second high-frequency power supply), 23: Automatic matching unit, 24: Ground electrode, 25: Susceptor, 26: Stage cover, 27: Camera, 28: Process A, 29: Process B, 30: Process C, 31: Process D, 32: Process E, 33: Process F.

Claims

1. A plasma processing apparatus connected to a server, comprising: a first high-frequency power supply that supplies first high-frequency power for generating plasma; a second high-frequency power supply that supplies second high-frequency power to a sample stage on which a sample is placed; and a control unit that controls the first high-frequency power supply and the second high-frequency power supply, wherein when a recipe, which is a plasma processing condition, is input from the server, sensor values ​​are measured by changing the parameters specified in the recipe; abnormal values ​​are determined based on the relationship between the parameters and the measured sensor values ​​over time; a model is generated by learning the relationship between the recipe and the abnormal values; when the recipe is created, abnormalities in the parameters are inferred using the generated model; a new recipe is created with a restricted input range for the parameters based on the input range of the parameters inferred to be abnormal; and the control unit controls the first high-frequency power supply and the second high-frequency power supply based on the newly created recipe.

2. A plasma processing apparatus according to claim 1, characterized in that the newly created recipe is set to a range of values ​​for the parameter excluding the input range of the parameter that was inferred to be abnormal.

3. A plasma processing apparatus according to claim 1, characterized in that the sensor value is the reflection coefficient or reflected power of the first high-frequency power or the peak voltage of the second high-frequency power.

4. A plasma processing apparatus according to claim 1, wherein the first high-frequency power supply provides, as the first high-frequency power, high-frequency power modulated by pulses having a plurality of periods with an amplitude greater than 0, and the parameter includes the pulse-modulated high-frequency power value for each of the periods, the frequency of the pulse, and the duty cycle of the pulse.

5. A plasma processing apparatus according to claim 1 or claim 4, wherein the second high-frequency power supply provides, as the second high-frequency power, high-frequency power modulated by pulses having a plurality of periods with an amplitude greater than 0, and the parameters include the pulse-modulated high-frequency power value for each of the periods, the frequency of the pulses, and the duty cycle of the pulses.

6. A plasma processing apparatus according to claim 1, further comprising a light emission measuring device for measuring the light emission of plasma, wherein the sensor value includes brightness information measured by the light emission measuring device.

7. A plasma treatment method for plasma treatment of a sample, comprising: a step of measuring sensor values ​​by changing parameters specified in a recipe when inputting a recipe which is a plasma treatment condition; a step of determining abnormal values ​​based on the relationship between the parameters and the measured sensor values ​​over time; a step of generating a model by learning the relationship between the recipe and the abnormal values; a step of inferring abnormalities in the parameters using the generated model when creating the recipe; a step of creating a new recipe in which the input range of the parameters is restricted based on the input range of the parameters inferred to be abnormal; and a step of plasma treatment of the sample using the newly created recipe.

8. A plasma processing method according to claim 7, characterized in that the newly created recipe is set to a range of values ​​for the parameter excluding the input range of the parameter that was inferred to be abnormal.

9. A plasma processing method according to claim 7, characterized in that the sensor value is the reflection coefficient or reflected power of high-frequency power for generating plasma, or the peak-to-peak voltage of high-frequency power supplied to a sample stage on which the sample is placed.

10. A plasma processing method according to claim 7, wherein the high-frequency power for generating the plasma is modulated by pulses having a plurality of periods with an amplitude greater than zero, and the parameters include the pulse-modulated high-frequency power value for each of the periods, the frequency of the pulse, and the duty cycle of the pulse.

11. A plasma processing method according to claim 7 or claim 10, wherein the high-frequency power supplied to a sample stage on which the sample is placed is modulated by pulses having a plurality of periods with an amplitude greater than 0, and the parameter includes the pulse-modulated high-frequency power value for each of the periods, the frequency of the pulse, and the duty cycle of the pulse.

12. A plasma processing method according to claim 7, characterized in that the sensor value includes brightness information measured by a light emission measuring device for measuring the emission of plasma.

Images