Semiconductor process device control method and semiconductor process device
By monitoring the opening of the gas supply valve and the light information generated by the plasma in real time, and combining the machine learning model, the activation time of the execution module is dynamically adjusted, which solves the problem of the asynchronous arrival time of the RF power supply and the gas, improves the stability and controllability of the semiconductor process, and increases the product yield.
Patent Information
- Application Number
- PCT/CN2026/081120
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-16
- Filing Date
- 2026-03-03
- Publication Date
- 2026-11-19
AI Technical Summary
In semiconductor process equipment, the timing of the RF power supply turn-on is out of sync with the time when the gas arrives at the semiconductor process cavity, which leads to a decrease in the stability and controllability of the deposition or etching process and affects product yield.
The detection module monitors the opening time of the gas delivery valve and the light information generated by the plasma in the semiconductor cavity in real time, calculates the delay time of gas transmission and plasma generation, and uses a machine learning model to build a compensation time correction formula to dynamically adjust the activation time of the execution module in order to achieve process synchronization optimization.
It improves the stability and controllability of semiconductor processes, increases product yield, avoids process instability caused by inaccurate delay time calculation, and achieves precise compensation and adaptive adjustment of the execution module activation time.
Smart Images

Figure CN2026081120_19112026_PF_FP_ABST
Abstract
Description
Control methods for semiconductor process equipment and semiconductor process equipment Technical Field
[0001] This application relates to the field of semiconductor technology, and in particular to a control method for semiconductor process equipment and semiconductor process equipment. Background Technology
[0002] Semiconductor process equipment is a key processing unit for front-end processes such as thin film deposition, etching, ion implantation, and cleaning. Plasma etching and deposition processes performed in semiconductor process equipment have become an important technology in the semiconductor manufacturing field. As the requirements for etching and deposition processes become increasingly stringent, cyclic etching and cyclic deposition have become a major trend. For example, in complex through-silicon via (TSV) etching equipment, the etching and deposition gases are switched repeatedly within 1 second. In such a rapid cyclic process, there is a discrepancy between the RF power supply turn-on time and the actual time when the gas arrives at the semiconductor process cavity. This leads to a decrease in the stability and controllability of the deposition or etching process, which affects the overall product yield. Technical issues
[0003] A control method and semiconductor process equipment are provided, which are beneficial to improving the stability and controllability of semiconductor processes and increasing product yield. Technical solutions
[0004] In a first aspect, to achieve the above objective, this application provides a method for controlling the semiconductor process equipment, comprising the following steps:
[0005] The system provides a detection module, a gas supply valve, a processing module connected to the gas supply valve and the detection module respectively, a control module and an execution module connected sequentially to the processing module, wherein the detection module is disposed in a semiconductor cavity, and the gas supply valve is disposed in a gas supply pipeline communicating with the semiconductor cavity;
[0006] The gas delivery valve controls the opening and closing of the gas delivery pipeline and sends valve opening information to the processing module;
[0007] The detection module acquires and sends the optical information generated by the plasma within the semiconductor cavity to the processing module;
[0008] The processing module obtains the first start time of gas flowing through the gas delivery valve based on the valve opening information, and the second start time of plasma generation in the semiconductor cavity based on the optical information, and obtains the delay time of the Nth process based on the first start time and the second start time, where N is an integer greater than or equal to 1;
[0009] The control module obtains the compensation time for the execution module to be pre-activated in the N+1th process based on the delay time of the Nth process, and controls the execution module to operate according to the compensation time in the N+1th process.
[0010] In an embodiment of the first aspect, the step of the control module obtaining the compensation time for the pre-activation of the execution module in the N+1th process based on the delay time of the Nth process includes: the control module correcting the delay time of the Nth process using a compensation time correction formula to obtain the compensation time for the pre-activation of the execution module in the N+1th process.
[0011] In an embodiment of the first aspect, the control method for the semiconductor process equipment further includes the steps of: the control module constructing a compensation time correction model for the compensation time and the delay time of several rounds of process using a machine learning model; and calculating and fitting the compensation time and the delay time of several rounds of process using the compensation time correction model to generate the compensation time correction formula.
[0012] In an embodiment of the first aspect, the compensation time correction formula satisfies the following relationship: the compensation time of the (N+1)th process round is equal to the sum of the delay time parameter and the time correction parameter; the delay time parameter is proportional to the delay time of the Nth process round, and the delay time parameter is proportional to the delay time weight coefficient; the delay time weight coefficient and the time correction parameter are both obtained by fitting the compensation time correction model.
[0013] In the first aspect of the embodiment, the compensation time correction formula is: ,
[0014] Wherein, T is the compensation time of the (N+1)th process cycle, ΔT is the delay time of the Nth process cycle, α is the delay time weighting coefficient, and β is the time correction parameter.
[0015] In an embodiment of the first aspect, the control method of the semiconductor process equipment further includes the step of: the control module inputting the compensation time and the delay time obtained in each round of process into the machine learning model in real time, and calculating and fitting all the compensation time and the delay time in the machine learning model through the compensation time correction model, so as to correct the delay time weight coefficient and the time correction parameter in the compensation time correction formula, and make the delay time weight coefficient greater than 0 and less than 3, and the time correction parameter greater than -1 and less than 1.
[0016] In an embodiment of the first aspect, the step of the control module constructing a compensation time correction model for the compensation time and the delay time of several rounds of process using a machine learning model further includes: the control module also inputting process parameters of several rounds of process into the machine learning model to construct the compensation time correction model; the process parameters include gas type, gas flow rate, pressure in the gas delivery pipeline, pressure in the semiconductor cavity, ambient temperature of the gas delivery pipeline, and usage time of the gas delivery valve.
[0017] In an embodiment of the first aspect, the control method for the semiconductor process equipment further includes the step of: the control module issuing an alarm when the compensation time is greater than a preset time threshold.
[0018] Secondly, this application provides a semiconductor process apparatus, including a gas supply valve, a detection module, a processing module, a control module, and an execution module. The gas supply valve is disposed in a gas supply pipeline communicating with a semiconductor cavity, and is used to control the opening and closing of the gas supply pipeline and send valve opening information to the processing module. The detection module is disposed in the semiconductor cavity, and is used to acquire and send optical information generated by plasma within the semiconductor cavity to the processing module. The processing module is connected to the gas supply valve and the detection module, and is used to obtain a first start time for gas to flow through the gas supply valve based on the valve opening information, and a second start time for plasma generation within the semiconductor cavity based on the optical information, and to obtain a delay time for the Nth process cycle based on the first start time and the second start time, where N is an integer greater than or equal to 1. The control module is connected to the processing module and the execution module, and is used to obtain a compensation time for the execution module to be pre-activated in the N+1th process cycle based on the delay time of the Nth process cycle, and to control the execution module to operate according to the compensation time in the N+1th process cycle.
[0019] In a second embodiment, the detection module includes a photodetector and a clamping member. The clamping member includes a light-blocking part, a fixing base, and a mounting part. The light-blocking part has a receiving cavity formed on its working surface facing the observation window, and the projected area of the light-blocking part on the observation window is greater than or equal to the area of the observation window. The fixing base is disposed in the receiving cavity, and the photodetector is fixed to the fixing base. The mounting part includes an outer edge groove that matches the outer edge protrusion of the observation window. The outer edge groove is arranged around the edge of the light-blocking part, and the detection module is detachably disposed on the observation window by engaging with the outer edge protrusion through the outer edge groove. Beneficial effects
[0020] The control method for the semiconductor process equipment described in this application and the semiconductor process equipment may have, or at least have, the following advantages:
[0021] (1) This application achieves accurate monitoring of the total delay time of gas transmission and plasma generation in the gas transmission pipeline from the gas valve to the generation of plasma in the semiconductor cavity by real-time monitoring of the first start time when the gas valve opens and the second start time when the execution module is activated to generate plasma in the semiconductor cavity. This avoids the situation where the compensation time is inaccurate due to the omission of the delay time of plasma generation from the gas entering the semiconductor cavity to the generation of plasma in the semiconductor cavity.
[0022] (2) This application adopts dual-signal synchronous detection, namely valve opening signal and optical signal, to realize real-time monitoring of gas transmission and plasma generation. The obtained delay time is more accurate, avoiding errors caused by indirect parameters, which is more conducive to improving the stability and controllability of semiconductor process.
[0023] (3) The present application uses the detection module to detect the light information generated by the plasma in the semiconductor cavity, which is more accurate than the detection of gas concentration in the semiconductor cavity in related technologies.
[0024] (4) This application obtains the compensation time for the pre-activation of the execution module in the N+1th process based on the delay time of the Nth process, so that the activation time of the execution module in the N+1th process can be adaptively adjusted and controlled according to the delay time of the Nth process. Compared with some related technologies that rely on experience to set fixed delay time parameters to adjust the activation time of the execution module, such as the RF power supply, the compensation time for the execution module to be activated in the N+1th process obtained by the delay time of the Nth process is more accurate and precise, and real-time compensation is achieved for the activation time of the execution module.
[0025] (5) The compensation time of the execution module in this application is obtained based on the delay time of the previous cycle, realizing adaptive adjustment. That is, the compensation time of this application can dynamically adapt to changes in the external environment, the usage of the gas pipeline, the gas type, etc. in real time. The compensation time has high accuracy, which is more conducive to improving the stability and controllability of the deposition process or etching process and improving the yield of the product.
[0026] In summary, through the single or synergistic control of the above-mentioned multiple factors, this application obtains precise gas transmission and plasma generation delay time by real-time detection of the time difference between plasma optical information and valve opening information, and achieves process synchronous optimization by dynamically compensating for time adaptively adjusting the activation of the execution module.
[0027] Other advantages of this application will be described in the following detailed description. Attached Figure Description
[0028] Figure 1 is a schematic flowchart of a control method for semiconductor process equipment according to some embodiments of this application.
[0029] Figure 2 is a schematic diagram of the structure of a semiconductor process apparatus according to some embodiments of this application.
[0030] Figure 3 is a structural block diagram of a semiconductor process apparatus according to some embodiments of this application.
[0031] Figure 4 is a structural block diagram of a semiconductor process apparatus according to some other embodiments of this application.
[0032] Figure 5 is a schematic diagram of the structure of the detection module in the semiconductor process equipment of some embodiments of this application.
[0033] Explanation of reference numerals in the attached figures:
[0034] 100. Detection module; 110. Photodetector; 111. Detection section; 112. Lead wire section; 120. Clamping component; 121. Light blocking section; 1211. Receiving cavity; 122. Fixing base; 123. Outer edge groove; 200. Gas supply valve; 300. Processing module; 310. Amplifier; 320. Bandpass filter; 330. Noise suppression unit; 340. Processing unit; 400. Control module; 410. Model building unit; 420. Analysis unit; 430. Storage unit; 440. Control unit; 500. Execution module; 600. Semiconductor cavity; 610. Gas supply line; 620. Observation window. Embodiments of the present invention
[0035] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art to which this application pertains. The terms "comprising" and similar expressions used herein mean that the element or object preceding the word covers the element or object listed following the word and its equivalents, but does not exclude other elements or objects.
[0036] TSV etching machines generally use Bosch technology. Bosch technology generates plasma by injecting specific gases (such as CF4, CHF3, Ar, etc.) into a closed chamber and applying radio frequency power. Plasma is a mixture of charged particles (ions, electrons, etc.) formed by the ionization of gas molecules under the action of an electric field, and has high reactivity. In related technologies, due to the time delay between the opening of the solenoid valve and the actual entry of the gas into the chamber, the delay time can usually reach 50ms~500ms, which will lead to the following problems: (1) The time of the radio frequency power supply opening is not synchronized with the time of the gas arriving at the semiconductor cavity, resulting in uneven etching or uneven deposition; (2) Poor process repeatability, affecting aspect ratio and sidewall morphology control.
[0037] To overcome the problems existing in related technologies, this application provides a control method and semiconductor process equipment for semiconductor process equipment, which is beneficial to improving the stability and controllability of semiconductor processes and increasing product yield.
[0038] In some embodiments, referring to FIG1, the control method of the semiconductor process equipment includes the following steps:
[0039] S0. A detection module, a gas supply valve, a processing module connected to the gas supply valve and the detection module respectively, a control module and an execution module connected sequentially to the processing module are provided, wherein the detection module is disposed in a semiconductor cavity, and the gas supply valve is disposed in a gas supply pipeline communicating with the semiconductor cavity;
[0040] S1. The gas supply valve controls the opening and closing of the gas supply pipeline and sends valve opening information to the processing module; the detection module acquires and sends light information generated by the plasma in the semiconductor cavity to the processing module.
[0041] S2. The processing module obtains the first start time of gas flowing through the gas supply valve based on the valve opening information, and obtains the second start time of plasma generation in the semiconductor cavity based on the optical information, and obtains the delay time of the Nth process based on the first start time and the second start time, where N is an integer greater than or equal to 1.
[0042] S3. The control module obtains the compensation time for the pre-activation of the execution module in the N+1th process based on the delay time of the Nth process, and controls the execution module to operate according to the compensation time in the N+1th process.
[0043] In this application, S1, the gas supply valve controls the opening and closing of the gas supply pipeline and sends valve opening information to the processing module; the detection module acquires and sends light information generated by plasma in the semiconductor cavity to the processing module; S2, the processing module obtains the first start time of gas flowing through the gas supply valve based on the valve opening information, and obtains the second start time of plasma generation in the semiconductor cavity based on the light information, and obtains the delay time of the Nth process step based on the first start time and the second start time. That is, this application can not only accurately monitor the delay time of gas transmission from the gas supply valve to the semiconductor cavity, i.e., the time of gas transmission in the gas supply pipeline, but also, compared with related technologies, only monitors the delay time of gas transmission in the gas supply pipeline from the gas supply valve to the semiconductor cavity. This application also accurately monitors the delay time from when the gas enters the semiconductor cavity to when plasma is generated within the semiconductor cavity. This delay time includes all the delays in the execution module's processes, from receiving the activation command to actual activation and then to plasma generation within the semiconductor cavity. By real-time monitoring the first start time of gas flow through the gas valve (the opening time) and the second start time of the execution module's activation to generate plasma within the semiconductor cavity, this application can accurately monitor the total delay time of gas transmission and plasma generation in the gas pipeline, avoiding inaccuracies in compensation time caused by overlooking the delay time between gas entering the semiconductor cavity and plasma generation. Furthermore, this application employs dual-signal synchronous detection—a valve opening signal and a light signal—to achieve real-time monitoring of gas transmission and plasma generation, resulting in more accurate delay times and avoiding errors caused by indirect parameters. This is more conducive to improving the stability and controllability of the semiconductor process. Moreover, this application uses the detection module to detect the light information generated by the plasma within the semiconductor cavity, which is more accurate than detecting the gas concentration within the semiconductor cavity in related technologies.
[0044] Furthermore, by obtaining the compensation time for the pre-activation of the execution module in the (N+1)th process stage based on the delay time of the Nth process stage, and controlling the execution module to operate according to the compensation time in the (N+1)th process stage, the activation time of the execution module in the (N+1)th process stage can be adaptively adjusted based on the delay time of the Nth process stage. That is, the execution module outputs power by advancing or delaying the RF power supply by the compensation time. Compared to some related technologies that rely on experience to set a fixed delay time parameter to adjust the activation time of the execution module, this application achieves this through the Nth process stage... The compensation time obtained from the delay time of the process allows the execution module to be activated in the N+1th process cycle with higher accuracy, thus achieving real-time compensation for the activation time of the execution module. Simultaneously, the compensation time activated by the execution module in each process cycle is based on the delay time of the previous process cycle, achieving adaptive adjustment. That is, the compensation time of this application can dynamically adapt in real time to changes in the external environment, such as temperature variations, gas pipeline usage, gas type, and rate. The high accuracy of the compensation time is more conducive to improving the stability and controllability of the deposition or etching process, thereby increasing product yield.
[0045] In this embodiment, the Nth and N+1th process cycles are the same processing technology, including but not limited to etching or deposition processes. This application has a wide range of applications and can be used in various semiconductor processes. For example, in some embodiments, in the Bosch process, the control module obtains the compensation time for the pre-activation of the execution module in the N+1th etching cycle based on the delay time of the Nth etching cycle, and obtains the compensation time for the pre-activation of the execution module in the N+1th deposition cycle based on the delay time of the Nth deposition cycle. The compensation time obtained from the delay time in the etching process cannot be applied to the deposition process to control the pre-activation of the execution module, as these are two different process lines that do not interfere with each other.
[0046] In this embodiment, the execution module is a radio frequency (RF) power supply. The RF power supply generates a high-frequency electromagnetic field, which ionizes gas at low or normal pressure to form plasma. Specifically, the control module controls the RF power supply to delay or advance the compensation time in the (N+1)th process stage based on the compensation time obtained in the Nth process stage. The synchronization range of the RF power supply can reach ±500ms, with a resolution of 1μs.
[0047] In some embodiments, the step of the control module obtaining the compensation time for the pre-activation of the execution module in the (N+1)th process stage based on the delay time of the Nth process stage includes: the control module correcting the delay time of the Nth process stage using a compensation time correction formula to obtain the compensation time for the pre-activation of the execution module in the (N+1)th process stage. By correcting the delay time, the accuracy of the compensation time is improved, avoiding errors caused by indirect parameters, enabling real-time dynamic adjustment of the execution module's activation time in the next process stage, and improving process stability.
[0048] In some embodiments, the control method for the semiconductor process equipment further includes the step of: the control module constructing a compensation time correction model for the compensation time and the delay time of several rounds of process using a machine learning model; and calculating and fitting the compensation time and the delay time of several rounds of process using the compensation time correction model to generate the compensation time correction formula. Constructing the compensation time correction model using the machine learning model enables dynamic prediction of the trend of time delay increase or decrease, and dynamic adaptation to changes in gas path conditions (such as pipeline aging, gas type switching) and changes in the external environment. This helps improve the accuracy of the compensation time, achieves real-time dynamic adaptive adjustment of the execution module's activation time, realizes process synchronization optimization, and improves process stability.
[0049] In this embodiment, the machine learning model analyzes a large amount of data to identify patterns and correlations, thereby enabling effective prediction of new and unseen data. In some embodiments, the machine learning model includes, but is not limited to, LSTM temporal networks.
[0050] In some embodiments, the step of the control module constructing a compensation time correction model for the compensation time and delay time of several rounds of process using a machine learning model further includes: the control module inputting process parameters of several rounds of process into the machine learning model to construct the compensation time correction model; the process parameters include gas type, gas flow rate, pressure in the gas delivery pipeline, pressure in the semiconductor cavity, ambient temperature of the gas delivery pipeline, and usage time of the gas delivery valve. These process parameters help to construct a more accurate compensation time correction model, enabling the compensation time to dynamically adapt to changes in gas path conditions (such as pipeline aging, gas type switching) and changes in the external environment, thus improving the accuracy of the compensation time and achieving real-time dynamic adjustment of the execution module's activation time in the next round of process, thereby enhancing process stability.
[0051] In some embodiments, the compensation time correction formula satisfies the following relationship: the compensation time T of the (N+1)th process round is equal to the sum of the delay time parameter A and the time correction parameter β; the delay time parameter A is proportional to the delay time ΔT of the Nth process round, and the delay time parameter A is proportional to the delay time weighting coefficient α; both the delay time weighting coefficient α and the time correction parameter β are obtained by fitting the compensation time correction model. This enables dynamic adaptation to changes in gas path conditions (such as pipeline aging, gas type switching) and changes in the external environment, helping to improve the accuracy of the compensation time, realize real-time dynamic adjustment of the activation time of the execution module in the next process round, and improve the stability of the process.
[0052] In some specific embodiments, the compensation time correction formula is:
[0053] In some specific embodiments, the delay time parameter satisfies the following formula:
[0054] In some embodiments, the compensation time correction formula is: ,
[0055] Wherein, T is the compensation time of the (N+1)th process cycle, ΔT is the delay time of the Nth process cycle, that is, ΔT is equal to the difference between the second start time and the first start time; α is the delay time weighting coefficient, and β is the time correction parameter.
[0056] In some embodiments, the control method of the semiconductor process equipment further includes the step of: the control module inputting the compensation time and the delay time obtained in each round of process into the machine learning model in real time, and calculating and fitting all the compensation time and the delay time in the machine learning model through the compensation time correction model, so as to correct the delay time weight coefficient and the time correction parameter in the compensation time correction formula, and make the delay time weight coefficient greater than 0 and less than 3, and the time correction parameter greater than -1 and less than 1. In this embodiment, when the number of process cycles is small, the machine learning model can analyze less data. For example, when only one process cycle is performed, the compensation time of the second process cycle may be equal to the delay time of the first process cycle. As the number of process cycles increases, the machine learning model updates and supplements the data of the compensation time correction model in real time, thereby dynamically generating a new compensation time correction formula. This allows for real-time correction of the delay time weight coefficient and the time correction parameter in the compensation time correction formula, which helps improve the accuracy of the compensation time and effectively avoids the influence of factors such as aging of the gas pipeline, ambient temperature of the gas pipeline, and gas liquefaction in the gas pipeline on the delay time weight coefficient and the time correction parameter.
[0057] In this embodiment, the delay time weighting coefficient α may be the same or different in several process rounds throughout the entire process; the time correction parameter β may also be the same or different in several process rounds. The delay time weighting coefficient α and the time correction parameter β are updated through fitting the compensation time correction model, which helps to improve the accuracy of the compensation time and avoid errors caused by indirect parameters. For example, some components in semiconductor process equipment will dissipate heat, causing the ambient temperature of the gas pipeline to rise, which will lead to a decrease in the delay time weighting coefficient α and the time correction parameter β, thus reducing the obtained compensation time. As the service life of the gas pipeline increases, the gas pipeline will gradually age, which will lead to an increase in the delay time weighting coefficient α and the time correction parameter β, thus increasing the obtained compensation time.
[0058] In some specific embodiments, the delay time weighting coefficient α is 0.01, 0.1, 0.5, 1, 1.2, 1.5, 1.8, 2, 2.3, 2.7 or 2.9, etc.
[0059] In some specific embodiments, the time correction parameter β is -0.9, -0.5, -0.2, -0.1, -0.05, 0, 0.07, 0.1, 0.3, 0.6, 0.75, or 0.95, etc.
[0060] In some embodiments, when the time correction parameter β is less than 0, it means that the compensation time T is reduced by the corresponding number of seconds based on the delay time parameter A. For example, when the time correction parameter β is -0.5, it means that the compensation time T is reduced by 0.5 seconds based on the delay time parameter A.
[0061] In some embodiments, when the time correction parameter β is greater than 0, it means that the compensation time T is increased by the corresponding number of seconds based on the delay time parameter A. For example, when the time correction parameter β is 0.6, it means that the compensation time T is increased by 0.6 seconds based on the delay time parameter A.
[0062] In some embodiments, the control method for the semiconductor process equipment further includes the step of: the control module issuing an alarm when the compensation time exceeds a preset time threshold to remind personnel to check for equipment damage. This helps avoid inaccurate compensation time due to equipment damage, which could lead to process instability and affect product yield. Examples of inaccurate compensation time include damage to the gas valve or detection module causing inaccurate data, damage to the execution module causing deviations in the current cycle output power or start-up time, increased delay time due to blocked gas pipelines, or software vulnerabilities in the processing or control modules.
[0063] The preset time threshold in this embodiment is specifically set according to the type of machine, gas type, process, etc. in the actual application. For example, in some specific embodiments, in certain etching processes, the compensation time is generally 0~3s, and the preset time threshold is 5s. When the control module determines that the compensation time is greater than 5s, it will issue an alarm to remind the staff to check whether the equipment is damaged.
[0064] In some embodiments, when the delay time weighting coefficient α is less than or equal to 0, or greater than or equal to 3, the resulting compensation time will be relatively large, and the control module will issue an alarm in this case.
[0065] In some other embodiments, when the time correction parameter β is less than or equal to -1 and greater than or equal to 1, the resulting compensation time will be relatively large, and the control module will issue an alarm.
[0066] In other embodiments, when the delay time weighting coefficient α is less than or equal to 0 and greater than or equal to 3, and the time correction parameter β is less than or equal to -1 and greater than or equal to 1, the obtained compensation time will be relatively large, and the control module will control the issuance of an alarm.
[0067] In some specific embodiments, the control method for the semiconductor process equipment further includes the following steps:
[0068] The gas supply valve opens and sends valve opening information to the processing module. The processing module obtains the first start time T1 of gas flowing through the gas supply valve based on the valve opening information.
[0069] Gas enters the semiconductor cavity and generates plasma. The detection module acquires and sends the light information generated by the plasma in the semiconductor cavity to the processing module. The processing module obtains the second start time T2 of plasma generation in the semiconductor cavity based on the light information.
[0070] The processing module obtains the delay time ΔT of the Nth process based on the first start time T1 and the second start time T2 and sends it to the control module, where ΔT = T2 - T1;
[0071] The control module inputs the delay time ΔT into the machine learning model to correct the compensation time correction formula through the compensation time correction model and obtain the compensation time T for the N+1th process. The control module controls the activation of the execution module in the N+1th process according to the compensation time T for the N+1th process.
[0072] In the N+1th round of process, the execution module is activated after a delay of T hours to ensure synchronization with plasma generation.
[0073] In some embodiments, referring to Figures 2 to 5, the semiconductor process equipment includes a detection module 100, a gas supply valve 200, a processing module 300, a control module 400, and an execution module 500. The gas supply valve 200 is disposed in a gas supply pipeline 610 communicating with a semiconductor cavity 600. The gas supply valve 200 controls the opening and closing of the gas supply pipeline 610 and sends valve opening information to the processing module 300. The detection module 100 is disposed in the semiconductor cavity 600. The detection module 100 acquires and sends optical information generated by the plasma within the semiconductor cavity 600 to the processing module 300. The processing module 300 is connected to the gas supply valve 200 and the detection module 100. Block 300 is used to obtain the first start time of gas flowing through the gas supply valve 200 based on the valve opening information, and the second start time of plasma generation in the semiconductor cavity 600 based on the light information, and to obtain the delay time of the Nth process based on the first start time and the second start time, where N is an integer greater than or equal to 1; the control module 400 is connected to the processing module 300 and the execution module 500, and the control module 400 is used to obtain the compensation time for the execution module 500 to be pre-activated in the N+1th process based on the delay time of the Nth process, and to control the execution module 500 to operate according to the compensation time in the N+1th process.
[0074] In this embodiment, the gas supply valve 200 is disposed in the gas supply pipeline 610 connecting the semiconductor cavity 600. The gas supply valve 200 is used to control the opening and closing of the gas supply pipeline 610 and send valve opening information to the processing module 300. The detection module 100 is disposed in the semiconductor cavity 600. The detection module 100 is used to acquire and send light information generated by plasma in the semiconductor cavity 600 to the processing module 300. The processing module 300 is connected to the gas supply valve 200 and the detection module 100. The processing module 300 is used to obtain the first start time of gas flowing through the gas supply valve 200 according to the valve opening information, and to obtain the second start time of plasma generation in the semiconductor cavity 600 according to the light information. It also obtains the delay time of the Nth process step according to the first start time and the second start time. That is, this application can not only accurately monitor the delay time of gas transmission from the gas supply valve 200 to the semiconductor cavity 600, i.e., the time of gas transmission in the gas supply pipeline 610, but also compared to In related technologies, only the gas transmission delay time in the gas delivery pipeline 610 from the gas delivery valve 200 to the semiconductor cavity 600 is monitored. This application also accurately monitors the plasma generation delay time from the gas entering the semiconductor cavity 600 to the generation of plasma within the semiconductor cavity 600. That is, the delay time includes all the delay times existing in each process of the execution module 500 receiving the start command, the actual start, and the generation of plasma within the semiconductor cavity 600. This application can achieve accurate monitoring of the total delay time of gas transmission and plasma generation in the gas delivery pipeline from the gas delivery valve 200 to the generation of plasma within the semiconductor cavity 600 by real-time monitoring of the first start time of the gas delivery valve 200 opening (i.e., the gas flowing through the gas delivery valve 200) and the second start time of the execution module 500 starting to generate plasma within the semiconductor cavity. This avoids the situation where the compensation time is inaccurate due to the omission of the plasma generation delay time from the gas entering the semiconductor cavity to the generation of plasma within the semiconductor cavity 600. Simultaneously, this application employs dual-signal synchronous detection, namely a valve opening signal and a light signal, to achieve real-time monitoring of gas transmission and plasma generation. This results in more accurate delay times, avoiding errors caused by indirect parameters and further improving the stability and controllability of the semiconductor process. Furthermore, this application uses the detection module 100 to detect the light information generated by the plasma within the semiconductor cavity 600, which is more accurate than the detection of gas concentration within the semiconductor cavity 600 in related technologies.
[0075] In addition, in this embodiment, the control module 400 is connected to the processing module 300 and the execution module 500. The control module 400 is used to obtain the compensation time for the pre-activation of the execution module 500 in the (N+1)th process stage based on the delay time of the Nth process stage, and controls the execution module 500 to operate according to the compensation time in the (N+1)th process stage. This allows the semiconductor process equipment to obtain the compensation time for the (N+1)th process stage based on the delay time of the Nth process stage, thereby adaptively adjusting the activation time of the execution module 500 in the (N+1)th process stage. This is a significant improvement over some related technologies that rely on experience to set a fixed delay time parameter to adjust the execution module. Regarding the RF power-on time, the compensation time for the execution module 500 in the N+1th process cycle, obtained through the delay time of the Nth process cycle, is more accurate and precise, achieving real-time compensation for the activation time of the execution module 500. Furthermore, the compensation time activated by the execution module 500 in each process cycle is based on the delay time of the previous process cycle, achieving adaptive adjustment. This means that the compensation time in this application can dynamically adapt to changes in the external environment, the usage of the gas pipeline 610, and the gas type, resulting in high compensation time accuracy. This is more conducive to improving the stability and controllability of the deposition or etching process, and increasing product yield.
[0076] In this embodiment, a gas delivery valve 200 can be installed on the corresponding gas delivery pipeline 610, or the gas delivery valve 200 that already exists in the original semiconductor process equipment can be used to control the opening and closing of the gas delivery pipeline 610 and send valve opening information to the processing module 300. This requires minimal modification to the original equipment and has low investment costs.
[0077] In some embodiments, the gas supply valve 200 is positioned close to the semiconductor cavity 600 to avoid or reduce indirect errors caused by the excessive length of the gas supply pipeline 610 between the gas supply valve 200 and the semiconductor cavity 600, which would be significantly affected by the external environment. In some specific embodiments, the gas supply valve 200 is a solenoid valve with a response time of less than 10ms, resulting in a fast valve response and further avoiding errors caused by indirect parameters.
[0078] In some embodiments, referring to Figures 2 and 5, the detection module 100 includes a photodetector 110, which is detachably disposed outside the observation window 620 of the semiconductor cavity 600, and the detection part 111 of the photodetector 110 is aligned with the region in the semiconductor cavity 600 where plasma is generated, so that the photodetector 110 can detect the light information generated by the plasma through the observation window 620. Based on the characteristic that atoms or molecules in the plasma generated within the semiconductor cavity 600 emit light upon returning to another energy state after being excited to an excited state by electrons, this application utilizes a photodetector 110 to detect the plasma generated within the semiconductor cavity 600 in a timely and accurate manner. This allows for precise detection of the second start time of the plasma generated within the semiconductor cavity 600, thereby achieving accurate monitoring of the delay time from gas flowing through the gas delivery valve 200 to the generation of plasma within the semiconductor cavity 600. This avoids the situation where the compensation time is inaccurate due to the omission of the delay time from gas entering the semiconductor cavity 600 to the generation of plasma within the semiconductor cavity 600. The delay time obtained by this application is more accurate, avoiding errors caused by indirect parameters, thus further improving the stability and controllability of the semiconductor process.
[0079] In some embodiments, the photodetector includes, but is not limited to, a photodiode. Photodiodes are inexpensive and readily available, which helps reduce investment costs. The photodiode is a photodiode with a wavelength range of 200nm to 1000nm.
[0080] In some embodiments, referring to Figures 2 and 5, the detection module 100 further includes a clamping member 120, which includes a light-blocking part 121 and a fixing base 122. The light-blocking part 121 has a receiving cavity 1211 formed on its working surface facing the observation window 620. The fixing base 122 is disposed in the receiving cavity 1211, and the photodetector 110 is fixed to the fixing base 122. The projected area of the light-blocking part 121 on the observation window 620 is greater than or equal to the area of the observation window 620, so that the light-blocking part 121 can form a shielding cover for the observation window 620 to block external light sources and prevent external light sources from affecting the photodetector 110 in detecting the light information generated by the plasma in the semiconductor cavity 600. This avoids or reduces detection errors, making the delay time obtained in this application more accurate and enabling more precise control of the execution module activation, thereby improving the stability and controllability of the semiconductor process.
[0081] In this embodiment, the projected area of the light-blocking part 121 on the observation window 620 is: the projected area of the light-blocking part 121 on the observation window 620 when projected along a direction perpendicular to the surface of the observation window 620.
[0082] In some embodiments, the light-blocking portion 121 includes, but is not limited to, a cylindrical structure, a conical structure, a frustum-shaped structure (as shown in Figure 5), a hemispherical structure, or a semi-frame structure. Of course, in some embodiments, the light-blocking portion 121 can also be a circular plate-like structure or a square plate-like structure.
[0083] In some embodiments, referring to FIG5, the lead portion 112 of the photodetector 110 passes through the light-blocking portion 121 to connect with the processing module 300.
[0084] In some embodiments, the clamping member further includes a mounting part, through which the detection module 100 is detachably mounted on the observation window 620. This allows for convenient and quick disassembly and installation, thus achieving the installation and fixation of the detection module 100 without affecting the original use of the observation window 620. It also facilitates periodic disassembly and replacement of the photodetector 110 in the detection module 100, ensuring the accuracy of the light information collected by the photodetector 110.
[0085] In some embodiments, the connection method between the mounting part and the observation window 620 includes, but is not limited to, adhesive bonding, snap-fit bonding, threaded connection, etc.
[0086] In some embodiments, referring to FIG5, the mounting portion includes an outer edge groove 123, which is arranged around the edge of the light-blocking portion 121, and the outer edge groove 123 is adapted to the outer edge protrusion of the observation window 620. The detection module is detachably mounted on the observation window by engaging the outer edge groove 123 with the outer edge protrusion, thus realizing the detachable mounting of the detection module 100 on the observation window 620. In this embodiment, the detection module 100 can be installed using the outer edge protrusion on the original observation window 620, requiring minimal modification to the original equipment and resulting in low investment costs. Furthermore, the clamping member 120 can be detachably mounted on the observation window 620, allowing the detection module 100 to be removed when it is necessary to observe the situation inside the semiconductor cavity 600 through the observation window 620. This convenient and quick disassembly and installation not only achieves the installation and fixation of the detection module 100 but also does not affect the original use of the observation window 620. It also facilitates the periodic replacement of the photodetector 110 in the detection module 100, ensuring the accuracy of the light information collected by the photodetector 110.
[0087] In some specific embodiments, the inner wall of the outer edge groove 123 is provided with an internal thread structure, and the internal thread structure is adapted to the external thread structure on the outer edge boss of the observation window 620, so that the detection module 100 can be detachably installed on the observation window 620 by means of threaded connection.
[0088] In some embodiments, referring to Figures 3 and 4, the processing module 300 includes an amplifier 310, a bandpass filter 320, and a noise suppression unit 330 sequentially connected to the detection module 100, and a processing unit 340 connected to the noise suppression unit 330 and the gas valve 200 respectively. The amplifier 310 is used to convert the optical signal in the optical information into an electrical signal and amplify it; the bandpass filter 320 is used to filter the amplified electrical signal; the noise suppression unit 330 is used to reduce noise in the filtered electrical signal; the processing unit is used to obtain the first start time based on the valve opening information, and the second start time based on the optical information, and to obtain the delay time of the Nth process step based on the first start time and the second start time. This reduces the influence of other signal interference on the optical information collected by the detection module 100, which is beneficial to obtaining a more accurate delay time and avoids errors caused by indirect parameters.
[0089] In some embodiments, the center frequency of the bandpass filter is 1 kHz and the bandwidth is 100 Hz.
[0090] In some embodiments, the processing unit includes an analog-to-digital converter (ADC) and a digital signal processor (DSP). The ADC is connected to the noise suppression unit and the DSP, respectively. The DSP is connected to the control module. The ADC has a sampling rate greater than or equal to 1 MHz to ensure a detection accuracy of less than or equal to 1 ms, which is beneficial for obtaining a more accurate delay time.
[0091] In some embodiments, referring to Figures 3 and 4, the control module 400 includes a model building unit 410 and an analysis unit 420. The model building unit 410 is used to construct a compensation time correction model for the compensation time and delay time of several rounds of process using a machine learning model, and to calculate and fit the compensation time and delay time of several rounds of process using the compensation time correction model to generate a compensation time correction formula. The analysis unit 420 is connected to the model building unit 410 and the processing module 300, respectively. The analysis unit 420 is used to correct the delay time of the Nth round of process using the compensation time correction formula to obtain the compensation time pre-activated by the execution module 500 in the N+1th round of process. In some specific embodiments, the analysis unit 420 is connected to the processing unit 340 in the processing module 300. This embodiment constructs a compensation time correction model through the machine learning model, enabling dynamic prediction of the trend of time delay increase or decrease, and dynamic adaptation to changes in gas path status (such as pipeline aging, gas type switching) and external environment changes. This helps improve the accuracy of compensation time, realizes real-time dynamic adaptive adjustment of the activation time of the execution module 500, achieves process synchronization optimization, and improves process stability. Moreover, as the delay time changes in each round of process, the machine learning model can update and supplement the data of the compensation time correction model in real time, thereby dynamically correcting the compensation time correction formula to ensure higher and more accurate compensation time.
[0092] In some embodiments, referring to Figures 3 and 4, the control module 400 further includes a storage unit 430 connected to the processing module 300 and the model building unit 410, respectively. The storage unit 430 stores the compensation time, the delay time, and the process parameters for each round of the process. The model building unit 410 calls the compensation time, the delay time, and the process parameters stored in the storage unit 430 for each round of the process and inputs them into the machine learning model to construct the compensation time correction model. That is, the compensation time, the delay time, and the process parameters for each round of the process are all stored in the storage unit 430 for the model building unit 410 to call when constructing the compensation time correction model. This allows for real-time updates and supplementation of the data in the compensation time correction model, thereby dynamically correcting the compensation time correction formula to ensure higher and more accurate compensation time precision.
[0093] In some embodiments, the control module 400 further includes an input unit connected to the storage unit 430, the input unit being used to input the process parameters.
[0094] In some specific embodiments, the processing module 300 sends the delay time and the compensation time of N rounds of process to the storage unit for storage; the input unit inputs the process parameters of each round of process and sends them to the storage unit for storage.
[0095] In some embodiments, referring to Figures 3 and 4, the control module 400 further includes a control unit 440 connected to the analysis unit 420, the control unit 440 being used to control the execution module to be activated after the compensation time of the N+1th process cycle is delayed or advanced according to the compensation time of the N+1th process cycle.
[0096] In some embodiments, the control module is a host computer, and the host computer software integrates machine learning algorithms written in Python.
[0097] In some embodiments, the communication interface between the control module 400 and the execution module 500 adopts the EtherCAT protocol, with a time delay of less than 1ms, ensuring real-time performance and avoiding errors caused by indirect parameters.
[0098] In some embodiments, the semiconductor process equipment further includes an alarm module connected to the control module. The alarm module is used to issue an alarm when the compensation time is greater than a preset time threshold to remind staff to check whether the equipment is damaged, which helps to avoid inaccurate compensation time due to equipment damage, which could lead to process instability and affect product yield.
[0099] In some embodiments, the semiconductor process equipment includes inductively coupled plasma etching (ICP-E), capacitively coupled plasma etching (CCP-E), through-silicon via (TSV) etching, and vapor deposition (VSD) equipment. Only the existing ICP-E, CCP-E, TSV etching, and VSD equipment need to be equipped with a detection module 100, the processing module 300, and the control module 400. This approach offers strong adaptability, a wide range of applications, simple and convenient modification, and low investment costs.
[0100] For example, in some embodiments, in a TSV etching machine, the Bosch process requires alternating introduction of etching gas (such as SF6) and passivation gas (such as C4F8) during reactive ion etching. The through-silicon via (TSV) etching apparatus originally has solenoid valves on the gas supply lines for both the etching gas and passivation gas. By simply installing the detection module outside the observation window of the TSV etching apparatus, and configuring a processing module connected to the solenoid valve and the detection module, and a control module connected to the processing module, the first start time of the etching gas flowing through the solenoid valve can be obtained by real-time monitoring of the solenoid valve's opening. The second start time of the etching gas entering the semiconductor cavity to generate plasma can be obtained by monitoring the light information of the RF power supply activation to generate plasma within the semiconductor cavity. The processing module analyzes the first and second start times to obtain the delay time of the current etching process. This allows the control module to adjust the activation time of the RF power supply in the next etching process based on the compensation time obtained in real-time from the current etching process. Furthermore, the processing module analyzes the third and fourth start times to obtain the delay time of the current deposition process. This allows the control module to adjust the activation time of the RF power supply in the next deposition process based on the compensation time obtained in real-time from the current deposition process. This enables simultaneous optimization of the etching or deposition processes, thereby improving product stability and yield.
[0101] While the embodiments of this application have been described in detail above, it will be apparent to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations fall within the scope and spirit of this application as set forth in the claims. Furthermore, the application described herein may have other embodiments and can be implemented or carried out in various ways.
Claims
1. A control method of a semiconductor process apparatus, characterized by, Includes the following steps: The system provides a detection module, a gas supply valve, a processing module connected to the gas supply valve and the detection module respectively, a control module and an execution module connected sequentially to the processing module, wherein the gas supply valve is disposed in a gas supply pipeline communicating with a semiconductor cavity, and the detection module is disposed in the semiconductor cavity; The gas delivery valve controls the opening and closing of the gas delivery pipeline and sends valve opening information to the processing module; The detection module acquires and sends the optical information generated by the plasma within the semiconductor cavity to the processing module; The processing module obtains the first start time of gas flowing through the gas delivery valve based on the valve opening information, and the second start time of plasma generation in the semiconductor cavity based on the optical information, and obtains the delay time of the Nth process based on the first start time and the second start time, where N is an integer greater than or equal to 1; The control module obtains the compensation time for the execution module to be pre-activated in the N+1th process based on the delay time of the Nth process, and controls the execution module to operate according to the compensation time in the N+1th process.
2. The control method of a semiconductor process apparatus according to claim 1, characterized by, The step of the control module obtaining the compensation time for the pre-activation of the execution module in the N+1th process based on the delay time of the Nth process includes: The control module corrects the delay time of the Nth process cycle using a compensation time correction formula to obtain the compensation time pre-activated by the execution module in the N+1th process cycle.
3. The control method of a semiconductor process apparatus according to claim 2, characterized by, It also includes the following steps: The control module constructs a compensation time correction model for the compensation time and delay time of several rounds of process using a machine learning model; and calculates and fits the compensation time and delay time of several rounds of process using the compensation time correction model to generate the compensation time correction formula.
4. The control method of a semiconductor process apparatus according to Claim 3, wherein The compensation time correction formula satisfies the following relationship: The compensation time for the (N+1)th process round is equal to the sum of the delay time parameter and the time correction parameter; the delay time parameter is proportional to the delay time of the Nth process round, and the delay time parameter is proportional to the delay time weight coefficient; the delay time weight coefficient and the time correction parameter are both obtained by fitting the compensation time correction model.
5. The control method of a semiconductor process apparatus according to Claim 4, wherein The compensation time correction formula is: , wherein the T is the compensation time of the N+1th process, the ΔT is the delay time of the Nth process, the α is the delay time weight coefficient, and the β is the time correction parameter.
6. The control method of a semiconductor process apparatus according to Claim 5, wherein It also includes the following steps: The control module inputs the compensation time and delay time obtained in each round of process into the machine learning model in real time, and calculates and fits all the compensation time and delay time in the machine learning model through the compensation time correction model, so as to correct the delay time weight coefficient and the time correction parameter in the compensation time correction formula, and make the delay time weight coefficient greater than 0 and less than 3, and the time correction parameter greater than -1 and less than 1.
7. The method of controlling a semiconductor process apparatus according to Claim 3, wherein The step of the control module constructing a compensation time correction model for the compensation time and the delay time of several rounds of process using a machine learning model further includes: The control module also inputs process parameters from several rounds of process manufacturing into the machine learning model to construct the compensation time correction model; The process parameters include gas type, gas flow rate, pressure in the gas delivery pipeline, pressure in the semiconductor cavity, ambient temperature of the gas delivery pipeline, and usage time of the gas delivery valve.
8. The method of controlling a semiconductor process apparatus according to Claim 1, wherein It also includes the following steps: The control module issues an alarm when the compensation time exceeds a preset time threshold.
9. A semiconductor process apparatus, characterized by, It includes a gas delivery valve, a detection module, a processing module, a control module, and an execution module; The gas supply valve is installed in the gas supply pipeline that connects to the semiconductor cavity. The gas supply valve is used to control the opening and closing of the gas supply pipeline and send valve opening information to the processing module. The detection module is disposed in the semiconductor cavity, and the detection module is used to acquire and send the light information generated by the plasma in the semiconductor cavity to the processing module; The processing module is connected to the gas supply valve and the detection module. The processing module is used to obtain the first start time of gas flowing through the gas supply valve according to the valve opening information, and to obtain the second start time of plasma generation in the semiconductor cavity according to the light information. The processing module is also used to obtain the delay time of the Nth process based on the first start time and the second start time, where N is an integer greater than or equal to 1. The control module is connected to the processing module and the execution module. The control module is used to obtain the compensation time for the execution module to be pre-activated in the N+1th process based on the delay time of the Nth process, and to control the execution module to operate according to the compensation time in the N+1th process.
10. The semiconductor process apparatus according to claim 9, wherein The detection module includes a photodetector and a clamping component, the clamping component comprising: The light-blocking part has a receiving cavity formed on its working surface facing the observation window, and the projected area of the light-blocking part on the observation window is greater than or equal to the area of the observation window; A mounting base is disposed within the receiving cavity, and the photodetector is fixed to the mounting base; and The mounting part includes an outer edge groove that matches the outer edge protrusion of the observation window. The outer edge groove is arranged around the edge of the light-blocking part. The detection module is detachably mounted on the observation window by engaging with the outer edge protrusion through the outer edge groove.
