A semiconductor device perturbation prediction and control method, system, device and medium

Abstract

The application provides a semiconductor equipment disturbance prediction and control method, system, device and medium, the method comprising: first, presetting the mapping rule between the process instruction and the disturbance event, analyzing the process instruction in the process recipe during the operation of the semiconductor equipment; screening the disturbance instruction based on the look-ahead window mechanism, converting the disturbance instruction into the disturbance event, and generating the corresponding disturbance prediction curve; generating a plurality of feedforward compensation components based on the disturbance prediction curve, superimposing the feedforward compensation components, obtaining the comprehensive feedforward compensation control quantity, mapping each comprehensive feedforward compensation control quantity to the prediction time axis, and obtaining the feedforward compensation curve; calculating the temperature error based on the current temperature value and the target temperature of the temperature control system, and obtaining the feedback control quantity combined with the PID algorithm; superimposing the comprehensive feedforward compensation control quantity and the feedback control quantity to obtain the final control output quantity, and sending the final control output quantity to the temperature control system to drive the heating or cooling device.

Description

Technical Field

[0001] This application relates to semiconductor technology, and more particularly to a method, system, device, and medium for predicting and controlling disturbances in semiconductor devices. Background Technology

[0002] Semiconductor manufacturing equipment is often subject to frequent disturbances during process execution, such as RF power switching, gas flow adjustment, changes in base temperature control settings, and thermal pulse operation. These disturbances directly affect the stability of the temperature control system and the consistency of process results. Most existing control methods rely on traditional feedback loops, adjusting based on temperature error after a disturbance occurs. However, these methods suffer from problems such as response lag, large transient deviations, and integral saturation. Some methods attempt to introduce feedforward control, but they generally lack systematic prediction of future disturbances, making it difficult to achieve advance compensation under complex process conditions. Therefore, accurately predicting the timing characteristics of disturbances before they actually occur and synergistically integrating the prediction results with feedback control has become a key technical challenge for improving the temperature control accuracy and response speed of semiconductor equipment. Summary of the Invention

[0003] This application provides a method, system, device, and medium for predicting and controlling disturbances in semiconductor devices to solve the problems of the prior art.

[0004] In a first aspect, this application provides a method for predicting and controlling disturbances in semiconductor devices, comprising:

[0005] Process recipe analysis involves pre-setting mapping rules between process instructions and disturbance events before the semiconductor equipment is put into operation, and then analyzing the process instructions in the process recipe during the operation of the semiconductor equipment.

[0006] Disturbance prediction involves filtering disturbance instructions from the process recipe based on a look-ahead window mechanism, converting the disturbance instructions into disturbance events based on the mapping rules, and mapping each disturbance event to a prediction time axis to generate a corresponding disturbance prediction curve.

[0007] Feedforward compensation: Based on the disturbance prediction curve, multiple feedforward compensation components are generated according to the disturbance type, and the feedforward compensation components are superimposed to obtain a comprehensive feedforward compensation control quantity. Each comprehensive feedforward compensation control quantity is mapped to the prediction time axis to obtain a feedforward compensation curve.

[0008] Feedback control calculates the temperature error based on the current temperature value of the temperature control system and the target temperature that is effective at the current moment, and obtains the feedback control quantity by combining it with the PID algorithm;

[0009] The coordinated output combines the integrated feedforward compensation control quantity and the feedback control quantity at the current moment to obtain the final control output quantity, and sends the final control output quantity to the temperature control system to drive the heating or cooling device.

[0010] In one possible design, the feedforward compensation is used to generate a compensation control quantity in advance based on the disturbance prediction curve to counteract the impact of the predicted disturbance on the temperature control system.

[0011] The feedback control is used to generate a correction control quantity based on the deviation between the actual temperature and the target temperature of the temperature control system, and to correct residual disturbances that are not covered by feedforward compensation.

[0012] The feedforward compensation and feedback control constitute a composite control architecture with division of labor and decoupling.

[0013] In one possible design, the disturbance prediction includes:

[0014] A look-ahead window mechanism is used to filter disturbance commands within a future period, and the disturbance commands are converted into disturbance events based on the mapping rules to form a disturbance event sequence.

[0015] Construct a prediction timeline within the prospective window;

[0016] Based on the expansion rule, each perturbation event within the look-ahead window is projected onto the prediction time axis;

[0017] Based on the projection of the disturbance event onto the prediction time axis, the disturbance prediction curve is calculated and generated point by point;

[0018] During the operation of the semiconductor device, when the process recipe is modified, the corresponding disturbance event is updated synchronously in real time.

[0019] In one possible design, the feedforward compensation includes:

[0020] Based on the disturbance type, the disturbance prediction curve is split into multiple disturbance component curves;

[0021] Multiple feedforward compensation components are calculated based on the disturbance component curves.

[0022] At the same prediction time point on the prediction time axis, each feedforward compensation component is superimposed to generate a comprehensive feedforward compensation control quantity.

[0023] The combined feedforward compensation amount at each of the predicted time points is spliced ​​together to obtain the feedforward compensation curve.

[0024] In one possible design, the unfolding rules include:

[0025] When the perturbation shape is step type, the corresponding perturbation amplitude is applied at the prediction time point corresponding to the event start time, and the perturbation amplitude remains unchanged at all subsequent prediction time points within the look-ahead window.

[0026] When the disturbance shape is a ramp, it is expanded point by point according to a linear rule between the start time and the end time of the event, so that the disturbance amplitude gradually transitions from the initial value to the target disturbance amplitude.

[0027] When the disturbance shape is pulse-like, a fixed disturbance amplitude is applied between the start time and the end time of the event, and the baseline value is restored immediately after the event ends.

[0028] In one possible design, the prediction timeline is constructed by dividing the look-ahead window into several consecutive prediction time points with a fixed update cycle, and the prediction time points are arranged sequentially to form the prediction timeline.

[0029] Each disturbance event within the look-ahead window is projected onto the prediction time axis to form a disturbance prediction curve.

[0030] In one possible design, at each prediction time point on the prediction time axis, it is checked whether a disturbance event has occurred at that prediction time point;

[0031] If only one disturbance event occurs at the predicted time, the disturbance amplitude of the disturbance event is directly used as the predicted value at that predicted time.

[0032] If multiple disturbance events occur simultaneously at the predicted time point, the disturbance amplitudes of these disturbance events are added together to obtain the predicted value for that predicted time point.

[0033] Secondly, this application provides a semiconductor device disturbance prediction and control system, comprising:

[0034] The process recipe analysis module pre-sets the mapping rules between process instructions and disturbance events before the semiconductor equipment is run, and analyzes the process instructions in the process recipe during the operation of the semiconductor equipment.

[0035] The disturbance prediction module filters disturbance instructions from the process recipe based on the look-ahead window mechanism, converts the disturbance instructions into disturbance events based on the mapping rules, and maps each disturbance event to the prediction time axis to generate a corresponding disturbance prediction curve.

[0036] The feedforward compensation module generates multiple feedforward compensation components according to the type of disturbance based on the disturbance prediction curve, and superimposes the feedforward compensation components to obtain a comprehensive feedforward compensation control quantity. Each comprehensive feedforward compensation control quantity is mapped to the prediction time axis to obtain the feedforward compensation curve.

[0037] The feedback control module calculates the temperature error based on the current temperature value of the temperature control system and the target temperature that is currently in effect, and obtains the feedback control quantity by combining it with the PID algorithm.

[0038] The collaborative output module superimposes the comprehensive feedforward compensation control quantity and the feedback control quantity at the current moment to obtain the final control output quantity, and sends the final control output quantity to the temperature control system to drive the heating or cooling device.

[0039] Thirdly, this application provides an electronic device, comprising:

[0040] Processor; and,

[0041] Memory for storing the executable instructions of the processor;

[0042] The processor is configured to perform any of the possible methods described in the first aspect by executing the executable instructions.

[0043] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement any of the possible methods described in the first aspect.

[0044] This application provides a semiconductor device disturbance prediction and control method, system, device, and medium. The method constructs an integrated control approach from prediction to compensation to correction through a complete process flow including process recipe analysis, disturbance prediction, feedforward compensation, feedback control, and coordinated output. This method can generate a disturbance prediction curve before the disturbance occurs and form a feedforward compensation quantity, which is superimposed on the real-time feedback control quantity within a unified framework, thus balancing rapid response and precise control, effectively improving the stability and process consistency of the temperature control system.

[0045] Furthermore, this application achieves refined prediction and compensation for disturbances in different physical channels through disturbance type decomposition, multiple feedforward compensation components, and a look-ahead window mechanism, ensuring the continuity and superpositionability of the disturbance prediction curve. The feedforward and feedback mechanisms are decoupled in the architecture, allowing the feedback loop to only correct residual disturbances and unknown factors, avoiding integral saturation and frequent parameter tuning, thereby significantly improving the system's robustness and practicality. Attached Figure Description

[0046] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0047] Figure 1 This is a schematic flowchart illustrating a semiconductor device disturbance prediction and control method according to an example embodiment of this application;

[0048] Figure 2 This is a schematic diagram of the disturbance prediction process according to an example embodiment of this application;

[0049] Figure 3 This is a schematic diagram of the feedforward compensation process according to an example embodiment of this application;

[0050] Figure 4 This is a schematic diagram of the structure of a semiconductor device disturbance prediction and control system according to an example embodiment of this application;

[0051] Figure 5 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application.

[0052] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0053] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0054] This embodiment proposes a method, system, device, and medium for predicting and controlling disturbances in semiconductor devices, which adopts the overall collaborative control concept of "feedforward prediction + feedback correction".

[0055] The feedforward compensation part is based on the analysis results of the process recipe. Within the look-ahead window, it identifies disturbance events such as changes in RF power, gas switching, and adjustment of the base temperature control setpoint, establishes a disturbance prediction curve, and converts it into a feedforward compensation amount, so as to achieve compensation in advance before the disturbance actually occurs.

[0056] Feedback control is based on the real-time temperature acquisition value of the temperature control system, which is compared with the target temperature set by the process formula to obtain the temperature error, and the feedback control quantity is calculated to eliminate unpredictable disturbances and steady-state deviations.

[0057] The combination of predictive compensation with feedforward compensation and real-time correction with feedback control enables this embodiment to balance the speed of response and the accuracy of control when dealing with frequent disturbances in the process steps of semiconductor equipment, thus significantly improving the stability and process consistency of the temperature control system.

[0058] Figure 1This is a schematic flowchart illustrating a semiconductor device disturbance prediction and control method according to an example embodiment of this application. Figure 1 As shown, the semiconductor device disturbance prediction and control method provided in this embodiment includes:

[0059] Step S101: Process recipe analysis. Before the semiconductor equipment is running, a mapping rule between process instructions and disturbance events is preset. During the operation of the semiconductor equipment, the process instructions in the process recipe are analyzed.

[0060] In this step, the process recipe is a set of sequentially ordered process instructions issued by the semiconductor equipment control system, serving as the overall set of process operations. Only a portion of these process instructions affect the temperature control system, such as the RF power setting instruction (RF_SET), RF power on instruction (RF_ON), ​​RF power off instruction (RF_OFF), gas switching instruction (GAS_SWITCH), base temperature control setting adjustment instruction (CHUCK_SETPOINT), and heat pulse instruction (HEAT_PULSE). These instructions are collectively referred to as disturbance instructions in this embodiment. Disturbance instructions are selected from the process recipe and converted into disturbance events based on preset mapping rules.

[0061] Each process instruction includes:

[0062] Instruction ID (InstrID) uniquely identifies a process instruction and is used to index and retrieve the start and end times of the instruction. It also serves as the basis for binding and tracing disturbance events with process instructions.

[0063] The Instruction Type (InstrType) describes the category of the process instruction, such as RF power setting instruction (RF_SET), RF power on instruction (RF_ON), ​​RF power off instruction (RF_OFF), gas switching instruction (GAS_SWITCH), base temperature control setting adjustment instruction (CHUCK_SETPOINT), heat pulse instruction (HEAT_PULSE), process chamber pressure setting instruction (PRESSURE_SET), and plasma switching instruction (PLASMA_ON / PLASMA_OFF), etc.

[0064] Instruction start time (InstrStartTime) is the timestamp at which the process instruction begins execution.

[0065] Instruction End Time (InstrEndTime) is the timestamp indicating when the execution of this process instruction has ended.

[0066] Additional information about the instruction (InstrMeta) is a list of parameters carried by the process instruction, such as power P, gas type type, flow rate F, target temperature T_target, etc.

[0067] It should be noted that the instruction start time and instruction end time are both generated by the timing scheduling module of the semiconductor equipment control system. Upon entering the process recipe-based operating state, the semiconductor equipment control system first generates an execution plan, which includes: instruction identifier (InstrID), instruction type (InstrType), instruction start time (InstrStartTime), instruction end time (InstrEndTime), and related parameters. In this embodiment, the process instructions are assigned values ​​based on the execution plan, thereby ensuring that the timing of all instructions is uniformly determined by the timing scheduling module of the semiconductor control system.

[0068] Each disturbance event includes:

[0069] EventType is used to characterize the physical channel category to which the disturbance event belongs, including radio frequency (RF), gas (GAS), base temperature control (CHUCK), and thermal pulse (PULSE).

[0070] Event Start Time (EventStartTime) is the timestamp at which the disturbance event begins to take effect, determined by the InstrStartTime of the process instruction.

[0071] EventEndTime is the timestamp at which the disturbance event ceases to function, determined by the InstrEndTime of the process instruction.

[0072] The disturbance magnitude is the equivalent disturbance intensity caused by the disturbance event to the temperature control system during the effective period. It is calculated by combining process instruction parameters with semiconductor equipment calibration coefficients, such as α_rf·P and α_gas(type)·F·Δh_ref.

[0073] The perturbation shape describes how a perturbation event evolves over time, including:

[0074] A step type (ramp) causes a sudden change in the amplitude of the perturbation at EventStartTime and remains constant before EventEndTime;

[0075] The step type perturbs the perturbation amplitude linearly between EventStartTime and EventEndTime until the target value is reached.

[0076] The pulse class applies a fixed perturbation amplitude between EventStartTime and EventEndTime, and immediately restores the baseline when the event ends.

[0077] Event Attachment Information (EventMeta) stores necessary information in addition to the fields mentioned above, including:

[0078] Instruction ID (InstrID) is a process instruction that uniquely identifies the source of a disturbance event, facilitating the binding and tracing of disturbance events with process instructions.

[0079] Target temperature (T_target): When the event type is CHUCK or PULSE, EventMeta contains the target temperature T_target, which is taken from the process recipe configuration. In feedback control, the currently effective T_target is used as a reference and compared with the current temperature T_actual(t) of the temperature control system to calculate the temperature error.

[0080] To facilitate the conversion of disturbance commands into quantifiable disturbance events, this embodiment pre-defines a set of mapping rules to describe the correspondence between different types of process commands and disturbance events. For example:

[0081] RF power setting command (RF_SET):

[0082] RF_SET(P,τ_ramp)={InstrID=Num, InstrType=RF_SET, InstrStartTime=t1,InstrEndTime=t1+τ_ramp, InstrMeta={power=P}},

[0083] The corresponding disturbance event is:

[0084] e_RF={EventType=RF, EventStartTime=InstrStartTime, EventEndTime=InstrEndTime, magnitude=α_rf·P, Shape=ramp, EventMeta={InstrID=Num}},

[0085] in:

[0086] InstrID=Num is used to bind to process instructions;

[0087] α_rf is the absorption coefficient calibrated by the semiconductor device, which is obtained from the semiconductor device calibration.

[0088] RF power-on command (RF_ON):

[0089] RF_ON={InstrID=Num, InstrType=RF_ON, InstrStartTime=t1, InstrMeta={P_set=P,resp_time =Δt_resp}},

[0090] The corresponding disturbance event is:

[0091] e_RF_ON={EventType=RF, EventStartTime=InstrStartTime, EventEndTime=InstrStartTime+Δt_resp, magnitude =α_rf·P,Shape = step, EventMeta={InstrID=Num}},

[0092] in:

[0093] InstrID=Num is used to bind to process instructions;

[0094] Δt_resp is the power response time of the semiconductor device, which is calibrated by the characteristics of the semiconductor device.

[0095] RF power-off command (RF_OFF):

[0096] RF_OFF={InstrID=Num, InstrType = RF_OFF, InstrStartTime = t2,InstrMeta = {resp_time=Δt_resp}},

[0097] The corresponding disturbance event is:

[0098] e_RF_OFF={EventType = RF, EventStartTime =InstrStartTime,EventEndTime=InstrStartTime + Δt_resp, magnitude=0, Shape=step, EventMeta={InstrID=Num}},

[0099] in:

[0100] InstrID=Num is used to bind to process instructions;

[0101] Δt_resp is the power response time of the semiconductor device, which is calibrated by the characteristics of the semiconductor device.

[0102] Gas switching command (GAS_SWITCH):

[0103] GAS_SWITCH(type, F) = {InstrID=Num, InstrType = GAS_SWITCH,InstrStartTime = t3, InstrEndTime =t4, InstrMeta={gas_type = type, flow = F}},

[0104] The corresponding disturbance event is:

[0105] e_GAS={EventType=GAS, EventStartTime=InstrStartTime, EventEndTime =InstrEndTime, magnitude =α_gas(type)·F ·Δh_ref, Shape=step, EventMeta={InstrID=Num}},

[0106] in:

[0107] InstrID=Num is used to bind to process instructions;

[0108] α_gas(type) is an empirical coefficient related to the type of gas, obtained from the semiconductor device calibration.

[0109] Δh_ref is the reference enthalpy difference or thermal load conversion factor, which is obtained from the semiconductor device calibration.

[0110] Base temperature control setting adjustment command (CHUCK_SETPOINT):

[0111] CHUCK_SETPOINT(T_target, τ_ramp)={InstrID=Num, InstrType=CHUCK_SETPOINT, InstrStartTime=t4, InstrEndTime=t4+τ_ramp, InstrMeta ={T_target = T_target}},

[0112] The corresponding disturbance event is:

[0113] e_CHUCK={EventType=CHUCK, EventStartTime=InstrStartTime, EventEndTime=InstrEndTime, magnitude=α_chuck·(T_target-T_prev), Shape=ramp, EventMeta ={InstrID=Num, T_target = T_target}},

[0114] in:

[0115] InstrID=Num is used to bind to process instructions;

[0116] α_chuck is the base thermal coupling coefficient, obtained from semiconductor device calibration;

[0117] T_target is the preset target temperature of the temperature control system, which is configured by the process formula;

[0118] T_prev is the target temperature of the most recently activated temperature control system, maintained by the semiconductor device control system.

[0119] Heat Pulse Command (HEAT_PULSE):

[0120] HEAT_PULSE(T_target,τ_pulse) ={ InstrID=Num, InstrType=HEAT_PULSE,InstrStartTime=t5, InstrEndTime=t5+τ_pulse, InstrMeta={ T_target=T_target}},

[0121] The corresponding disturbance event is:

[0122] e_HEAT_PULSE ={EventType=PULSE, EventStartTime=InstrStartTime,EventEndTime=InstrEndTime, Shape=pulse, Magnitude=α_heat·(T_target - T_prev), EventMeta={InstrID=Num, T_target=T_target}},

[0123] in:

[0124] InstrID=Num uniquely identifies a process instruction and is used for binding and tracing that process instruction;

[0125] τ_pulse is the duration of the thermal pulse, configured in the process formulation;

[0126] α_heat is the equivalent thermal coupling coefficient of the heat source, obtained from the thermal calibration of the semiconductor device;

[0127] T_target is the preset target temperature of the temperature control system, configured in the process recipe;

[0128] T_prev is the most recently effective temperature setpoint before the start of the thermal pulse. It is maintained by the semiconductor device control system and is of the same origin as T_prev in the base temperature control setpoint adjustment instruction (CHUCK_SETPOINT).

[0129] It should be noted that other types of process instructions can also be extended according to this rule.

[0130] Step S102: Disturbance prediction. Based on the look-ahead window mechanism, disturbance instructions are screened from the process recipe. Based on the mapping rules, the disturbance instructions are converted into disturbance events. Each disturbance event is mapped to the prediction time axis to generate a corresponding disturbance prediction curve.

[0131] In this step, the disturbance prediction curve characterizes the temporal variation of the disturbance over a future period, serving as the input basis for subsequent calculations of integrated feedforward compensation control, feedback control, and coordinated output.

[0132] This step enables the prediction of disturbances over a future period before the execution of process instructions, obtains the trend of disturbance changes over time, and realizes the early identification and quantification of potential disturbances, providing a reliable basis for subsequent feedforward compensation, thereby improving the response speed and control accuracy of the temperature control system.

[0133] Step S103: Feedforward compensation. Based on the disturbance prediction curve, multiple feedforward compensation components are generated according to the disturbance type. The feedforward compensation components are superimposed to obtain a comprehensive feedforward compensation control quantity. Each comprehensive feedforward compensation control quantity is mapped to the prediction time axis to obtain a feedforward compensation curve.

[0134] In this step, the integrated feedforward compensation control quantity can reflect the overall impact of disturbances within the look-ahead window, providing the temperature control system with a control quantity for advance correction, and serving as the input basis for subsequent feedback control and coordinated output.

[0135] In this embodiment, multiple feedforward compensation components can generate compensation control quantities in advance for disturbance events such as radio frequency (RF), gas (GAS), base temperature control (CHUCK), and heat pulse (PULSE). When a disturbance occurs, most of its impact can be quickly offset, so that the temperature control system does not need to rely on the feedback loop for large adjustments, thereby significantly reducing the transient deviation of the temperature control system.

[0136] Step S104: Feedback control. Based on the current temperature value of the temperature control system and the target temperature that is effective at the current moment, calculate the temperature error and obtain the feedback control quantity by combining it with the PID algorithm.

[0137] In this step, the feedforward compensation is used to generate a compensation control quantity in advance based on the disturbance prediction curve to offset the impact of the predicted disturbance on the temperature control system.

[0138] The feedback control is used to generate a correction control quantity based on the deviation between the actual temperature and the target temperature of the temperature control system, and to correct residual disturbances that are not covered by feedforward compensation.

[0139] The feedforward compensation and feedback control constitute a composite control architecture with division of labor and decoupling.

[0140] For calculating the feedback control quantity, firstly, the current temperature T_actual(t) of the temperature control system is collected and compared with the most recently effective target temperature (T_target) at the current moment to obtain the temperature error:

[0141] ,

[0142] The target temperature (T_target) is the target temperature set by the most recent base temperature control setpoint adjustment command (CHUCK_SETPOINT), or the target temperature set by the heat pulse command (HEAT_PULSE) within the effective period.

[0143] Then, based on the temperature error e(t), a PID algorithm is used to calculate the feedback control quantity U. fb (t):

[0144]

[0145] Among them, K p K i and K d The feedback control parameter is a preset global parameter, and its specific value can be set based on experimental measurements or experience.

[0146] In this embodiment, with the introduction of a feedforward compensation mechanism, the feedback control quantity only needs to handle the residual error and unknown disturbance that the feedforward compensation cannot fully cover, i.e., the so-called "residual". Since the main disturbance has been weakened by the feedforward compensation, the feedback control can always work in a relatively ideal linear region, avoiding integral saturation and frequent parameter tuning problems caused by large disturbances, thereby achieving faster and more stable closed-loop temperature control performance.

[0147] Step S105: Coordinated output. At the current moment, the integrated feedforward compensation control quantity and the feedback control quantity are superimposed to obtain the final control output quantity, and the final control output quantity is sent to the temperature control system to drive the heating or cooling device.

[0148] In this step, the combined feedforward compensation control quantity and the feedback control quantity are algebraically superimposed at the current moment to obtain the final control output quantity U. final :

[0149] .

[0150] Among them, U ff (t) represents the comprehensive feedforward compensation amount; U fb (t) represents the feedback control variable.

[0151] Finally, the final control output U final (t) is sent to the temperature control system to drive the heating or cooling device.

[0152] Figure 2 This is a schematic diagram illustrating the disturbance prediction process according to an example embodiment of this application. Figure 2 As shown, the disturbance prediction method provided in this embodiment includes:

[0153] Step S1021: Use a look-ahead window mechanism to filter disturbance commands within a future period of time, and convert the disturbance commands into disturbance events based on the mapping rules to form a disturbance event sequence.

[0154] In this step, the look-ahead window mechanism involves parsing process instructions for a fixed duration H (e.g., 10-60 seconds) based on the process recipe at the current time t0. This fixed duration H is the look-ahead window [t0, t0+H]. Within the look-ahead window, process instructions that affect the temperature control system are selected as disturbance instructions based on their instruction type. Each disturbance instruction is then converted into a disturbance event based on a preset mapping rule, thus obtaining the disturbance event sequence within the look-ahead window.

[0155] In this embodiment, the process commands that affect the temperature control system include the radio frequency power setting command (RF_SET), the radio frequency power on command (RF_ON), ​​the radio frequency power off command (RF_OFF), the gas switching command (GAS_SWITCH), the base temperature control setting adjustment command (CHUCK_SETPOINT), and the heat pulse command (HEAT_PULSE).

[0156] It should be noted that when the start and end times of a perturbation instruction cross the boundary of the current lookahead window [t0, t0+H], the effective execution time of the perturbation instruction (the difference between the instruction end time and the instruction start time) will be truncated within that lookahead window, retaining only the portion overlapping with the current lookahead window, i.e., [InstrStartTime, t0+H]. The remaining portion will continue to expand within the next lookahead window, meaning the effective execution time of the perturbation instruction in the next lookahead window is [t0+H, InstrEndTime]. This "intra-window truncation, cross-window continuation" approach ensures the continuity of the perturbation prediction curve and computational efficiency.

[0157] Step S1022: Construct a prediction time axis within the look-ahead window.

[0158] In this step, the prediction time axis is constructed by dividing the look-ahead window into several consecutive prediction time points with a fixed update period Δt (e.g., 100 milliseconds or 200 milliseconds), and these prediction time points are arranged sequentially to form the prediction time axis. For example, if the current time is t0=100s, H=30s, and Δt=0.1s, then the prediction time axis is t=[100.0,100.1,100.2,...,129.9], where t includes a total of 300 prediction time points.

[0159] This embodiment discretizes the look-ahead window (a continuous period of time in the future) to obtain a prediction time axis composed of several consecutive prediction time points. The prediction time axis provides a unified time reference for the mapping and superposition calculation of subsequent disturbance events in time, ensuring that the disturbance amplitudes of all disturbance events can be aligned and synthesized on the same time axis.

[0160] Step S1023: Based on the expansion rule, project each perturbation event in the look-ahead window onto the prediction time axis.

[0161] In this step, each perturbation event within the look-ahead window is projected onto the prediction time axis to form a perturbation prediction curve. Specifically:

[0162] The perturbation events within the look-ahead window are projected one by one onto the prediction timeline. Each perturbation event is then expanded into a corresponding time series curve segment based on a unified time reference according to expansion rules, thus providing a foundation for the generation of subsequent perturbation prediction curves. The expansion rules include:

[0163] When the perturbation shape is step-like, the corresponding perturbation magnitude is applied at the predicted time point corresponding to the event start time, and this perturbation magnitude remains unchanged at all subsequent predicted time points within the look-ahead window. A typical scenario is the RF power-on or power-off operation.

[0164] When the disturbance shape is ramp-like, it is expanded point-by-point according to a linear rule between the event start time and the event end time, so that the disturbance amplitude gradually transitions from the initial value to the target disturbance amplitude. This linear rule is determined by the preset rule of the semiconductor device control system. In this embodiment, only this linear rule is followed during disturbance prediction. A typical scenario is the adjustment of the target temperature of the base temperature control. If the ramp-like disturbance event process crosses the look-ahead window boundary, only the disturbance amplitude up to the end of the current look-ahead window is expanded in the current look-ahead window. The next look-ahead window continues to expand using the disturbance amplitude at the end of the previous look-ahead window as the initial value, ensuring the continuity of the disturbance prediction curve.

[0165] When the perturbation shape is pulse-type, a fixed perturbation magnitude is applied from the start time to the end time of the event, and the baseline value is restored immediately after the event ends. Typical scenarios include gas flushing pulses or thermal pulse operations. If the duration of a pulse-type perturbation event spans the look-ahead window boundary, only the overlapping section is expanded in the current look-ahead window, and the remaining portion continues to be expanded in the next look-ahead window with the fixed perturbation magnitude.

[0166] By using the above-mentioned expansion rules and continuous expansion processing across look-ahead windows, it is possible to ensure that different types of disturbance events form continuous and superimposed time-series curve segments on the prediction time axis, thereby accurately reflecting the evolution characteristics of disturbances over time during the operation of semiconductor devices, and providing a reliable basis for the generation of subsequent disturbance prediction curves and feedforward compensation.

[0167] Step S1024: Based on the projection results of the disturbance event on the prediction time axis, calculate and generate the disturbance prediction curve point by point.

[0168] This step includes:

[0169] At each prediction time point on the prediction time axis, it is checked whether a disturbance event has occurred at that prediction time point;

[0170] If only one disturbance event occurs at the predicted time, the disturbance amplitude of the disturbance event is directly used as the predicted value at that predicted time.

[0171] If multiple disturbance events occur simultaneously at the predicted time point, the disturbance amplitudes of these disturbance events are added together to obtain the predicted value for that predicted time point.

[0172] In this way, all disturbance events can be synthesized point by point into a continuous prediction curve Q. predict (t), the disturbance prediction curve reflects the disturbance changes that the temperature control system may experience within the look-ahead window, and provides input data for the subsequent integrated feedforward compensation control.

[0173] Step S1025: During the operation of the semiconductor device, when the process recipe is modified, the corresponding disturbance event is updated synchronously in real time.

[0174] In this step, to address the possibility of changes to process steps during semiconductor equipment operation, this embodiment employs an event-triggered mechanism for dynamic synchronization adjustment. When the process recipe is updated, the semiconductor equipment control system sends a process recipe update message and simultaneously triggers a process recipe update operation, including:

[0175] Read and parse the latest process formula;

[0176] Based on the current look-ahead window, disturbance instructions are selected from the latest process formulation according to the instruction type;

[0177] Each perturbation command is converted into a perturbation event according to the preset mapping rules, and a perturbation event sequence arranged in chronological order is generated.

[0178] Based on the perturbation event sequence, a corresponding perturbation prediction curve is generated on the prediction time axis.

[0179] This method ensures that the disturbance prediction results remain effective and accurate even when process steps or parameters are adjusted.

[0180] Figure 3 This is a schematic diagram illustrating the feedforward compensation process according to an example embodiment of this application. Figure 3 As shown, the feedforward compensation method provided in this embodiment includes:

[0181] Step S1031: Based on the disturbance type, split the disturbance prediction curve into multiple disturbance component curves.

[0182] In this step, the perturbation prediction curve Q predict(t) represents the comprehensive prediction result after superimposing all disturbance events within the look-forward window. To facilitate feedforward compensation for different disturbance types, the disturbance prediction curve is split into multiple disturbance component curves according to the disturbance type. The disturbance prediction curve in this embodiment includes the radio frequency disturbance prediction curve Q. RF (t), gas-type disturbance prediction curve Q GAS (t), Prediction curve of base temperature control disturbance Q CHUCK (t) and thermal pulse disturbance prediction curve Q PULSE (t). Among them, the radio frequency disturbance prediction curve Q RF (t) The prediction curve formed by superimposing radio frequency disturbance events on the prediction time axis, and the gas disturbance prediction curve Q. GAS (t) The prediction curve formed by the superposition of gas-type disturbance events, and the prediction curve Q for base temperature control type disturbances. CHUCK (t) The prediction curve formed by superimposing disturbance events of the base temperature control type, the thermal pulse disturbance prediction curve Q. PULSE (t) is the prediction curve formed by superimposing disturbance events of the heat pulse type.

[0183] Step S1032: Calculate multiple feedforward compensation components based on the disturbance component curve.

[0184] In this step, the feedforward compensation components include the radio frequency feedforward compensation component, the gas feedforward compensation component, the base temperature control feedforward compensation component, and the thermal pulse feedforward compensation component.

[0185] When the disturbance type is radio frequency (RF), the RF feedforward compensation component U corresponding to the disturbance prediction curve RF (t) is:

[0186] ,

[0187] Among them, K rf This is the preset RF gain coefficient.

[0188] When the disturbance type is gas, the gas feedforward compensation component U corresponding to the disturbance prediction curve GAS (t) is:

[0189] ,

[0190] Among them, K gas This is a preset gain coefficient curve for gas types.

[0191] When the disturbance type is base temperature control type, the base temperature control feedforward compensation component U corresponding to the disturbance prediction curve CHUCK (t) is:

[0192] ,

[0193] Among them, K chuck This is the preset gain coefficient for base temperature control.

[0194] When the disturbance type is thermal pulse, the thermal pulse feedforward compensation component U corresponding to the disturbance prediction curve PULSE (t) is:

[0195] ,

[0196] Among them, K pulse This is the preset thermal pulse gain coefficient.

[0197] It should be noted that K rf K gas K chuck and K pulse Used to reflect the response sensitivity of a temperature control system to disturbance events of different types. K rf K gas K chuck and K pulse Determined during the equipment calibration phase.

[0198] Step S1033: At the same prediction time point on the prediction time axis, each feedforward compensation component is superimposed to generate a comprehensive feedforward compensation control quantity.

[0199] In this step, at the same prediction time point on the prediction time axis, each feedforward compensation component is algebraically superimposed to obtain the comprehensive feedforward compensation amount:

[0200] ,

[0201] Among them, U ff (t) represents the comprehensive feedforward compensation amount.

[0202] It should be noted that at any prediction time point on the prediction time axis, if there is no disturbance event of any type, such as radio frequency (RF), gas (GAS), base temperature control (CHUCK), or thermal pulse (PULSE), the feedforward compensation component corresponding to that disturbance type will be zero and will not participate in the calculation at that time point.

[0203] Step S1034: Segment the comprehensive feedforward compensation amount at each predicted time point to obtain the feedforward compensation curve.

[0204] In this step, a comprehensive feedforward compensation value is generated point-by-point on the prediction time axis, and these comprehensive feedforward compensation values ​​are spliced ​​together to generate a feedforward compensation curve. This comprehensive feedforward compensation curve is used to represent the variation of the feedforward compensation power required by the temperature control system over time within a future look-ahead window. This comprehensive feedforward compensation curve is cached in the storage module of the semiconductor device control system for real-time retrieval during operation.

[0205] By using the above methods, a complete integrated feedforward compensation curve can be obtained in advance before the disturbance occurs, providing an input basis for the subsequent coordinated output with the feedback control quantity, and ensuring that the temperature control system can compensate in a timely and accurate manner when the disturbance occurs.

[0206] Figure 4 This is a schematic diagram illustrating the structure of a semiconductor device disturbance prediction and control system according to an example embodiment of this application. Figure 4 As shown, the disturbance prediction and control system 400 for semiconductor devices provided in this embodiment includes: a process recipe analysis module 410, a disturbance prediction module 420, a feedforward compensation module 430, a feedback control module 440, and a cooperative output module 450.

[0207] The process recipe analysis module 410 presets the mapping rules between process instructions and disturbance events before the semiconductor equipment is run, and analyzes the process instructions in the process recipe during the operation of the semiconductor equipment.

[0208] The disturbance prediction module 420 filters disturbance instructions from the process recipe based on the look-ahead window mechanism, converts the disturbance instructions into disturbance events based on the mapping rules, and maps each disturbance event to the prediction time axis to generate a corresponding disturbance prediction curve.

[0209] The feedforward compensation module 430 generates multiple feedforward compensation components according to the type of disturbance based on the disturbance prediction curve, and superimposes the feedforward compensation components to obtain a comprehensive feedforward compensation control quantity. Each comprehensive feedforward compensation control quantity is mapped to the prediction time axis to obtain the feedforward compensation curve.

[0210] The feedback control module 440 calculates the temperature error based on the current temperature value of the temperature control system and the target temperature that is effective at the current moment, and obtains the feedback control quantity by combining the PID algorithm.

[0211] The collaborative output module 450 superimposes the integrated feedforward compensation control quantity and the feedback control quantity at the current moment to obtain the final control output quantity, and sends the final control output quantity to the temperature control system to drive the heating or cooling device.

[0212] Figure 5 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. For example... Figure 5 As shown, the electronic device 500 provided in this embodiment includes: a processor 501 and a memory 502; wherein:

[0213] Memory 502 is used to store computer programs, and the memory may also be flash memory.

[0214] Processor 501 is used to execute the execution instructions stored in the memory to implement the various steps in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0215] Alternatively, the memory 502 can be either standalone or integrated with the processor 501.

[0216] When the memory 502 is a device independent of the processor 501, the electronic device 500 may further include:

[0217] Bus 503 is used to connect the memory 502 and the processor 501.

[0218] This embodiment also provides a readable storage medium storing a computer program, which, when executed by at least one processor of an electronic device, enables the electronic device to perform the methods provided in the various embodiments described above.

[0219] This embodiment also provides a program product including a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the methods provided in the various embodiments described above.

[0220] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the foregoing claims.

[0221] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for predicting and controlling disturbances in semiconductor devices, characterized in that, The method comprises the following steps: process recipe analysis, mapping rules between process instructions and disturbance events are preset before the semiconductor equipment is running, and the process instructions in the process recipe are analyzed during the running of the semiconductor equipment; disturbance prediction, disturbance instructions are screened from the process recipe based on a look-ahead window mechanism, the disturbance instructions are converted into disturbance events based on the mapping rules, each disturbance event is mapped to a predicted time axis, and a corresponding disturbance prediction curve is generated; feedforward compensation, a plurality of feedforward compensation components are generated based on the disturbance prediction curve according to disturbance types, the feedforward compensation components are superimposed, a comprehensive feedforward compensation control quantity is obtained, and each comprehensive feedforward compensation control quantity is mapped to the predicted time axis to obtain a feedforward compensation curve; feedback control, a temperature error is calculated based on a current temperature value of a temperature control system and a target temperature effective at a current time, and a feedback control quantity is obtained by combining a PID algorithm; coordinated output, the comprehensive feedforward compensation control quantity and the feedback control quantity are superimposed at the current time to obtain a final control output quantity, and the final control output quantity is sent to the temperature control system to drive a heating or cooling device; the feedforward compensation is used to generate compensation control quantities in advance based on the disturbance prediction curve to offset the influence of predicted disturbances on the temperature control system; the feedback control is used to generate correction control quantities based on the deviation between the actual temperature of the temperature control system and the target temperature to correct residual disturbances not covered by the feedforward compensation; the feedforward compensation and the feedback control constitute a decoupled composite control architecture; the disturbance prediction comprises the following steps: disturbance instructions in a future period of time are screened by using a look-ahead window mechanism, and the disturbance instructions are converted into disturbance events based on the mapping rules to form a disturbance event sequence; a predicted time axis is constructed within the look-ahead window; each disturbance event within the look-ahead window is projected onto the predicted time axis based on an expansion rule; a disturbance prediction curve is calculated and generated point by point based on the projection result of the disturbance event on the predicted time axis; when the process recipe is modified during the running of the semiconductor equipment, the corresponding disturbance events are updated in real time.

2. The semiconductor device perturbation prediction and control method of claim 1, wherein, the feedforward compensation comprises the following steps: the disturbance prediction curve is split into a plurality of disturbance component curves based on the disturbance types; a plurality of feedforward compensation components are calculated based on the disturbance component curves; each feedforward compensation component is superimposed at the same predicted time point on the predicted time axis to generate a comprehensive feedforward compensation control quantity; the comprehensive feedforward compensation quantities at each predicted time point are spliced to obtain a feedforward compensation curve.

3. The method of claim 1, wherein the step of determining the semiconductor device perturbation comprises: the expansion rule comprises the following steps: when the disturbance shape is a step type, a corresponding disturbance amplitude is applied at the predicted time point corresponding to the event start time, and the disturbance amplitude remains unchanged at all subsequent predicted time points within the look-ahead window; when the disturbance shape is a ramp type, the disturbance amplitude is gradually transitioned from an initial value to a target disturbance amplitude according to a linear rule point by point between the event start time and the event end time; when the disturbance shape is a pulse type, a fixed disturbance amplitude is applied between the event start time and the event end time, and the baseline value is restored immediately after the end.

4. The method of claim 1, wherein the step of determining the semiconductor device perturbation comprises: determining a semiconductor device perturbation based on a semiconductor device parameter and a semiconductor device parameter perturbation. The prediction time axis is constructed by dividing the look-ahead window into a plurality of continuous prediction time points in a fixed update period, and the prediction time points are sequentially arranged to form a prediction time axis; Each disturbance event in the look-ahead window is projected onto the prediction time axis to form a disturbance prediction curve.

5. The method of claim 1, wherein the step of determining the semiconductor device perturbation comprises: determining a semiconductor device perturbation based on a semiconductor device parameter and a semiconductor device parameter perturbation. At each prediction time point on the prediction time axis, it is checked whether there is a disturbance event acting on the prediction time point; If only one disturbance event occurs at the prediction time point, the disturbance amplitude of the disturbance event is directly taken as the prediction value of the prediction time point; If multiple disturbance events occur at the prediction time point, the disturbance amplitudes of the multiple disturbance events are added to obtain the prediction value of the prediction time point.

6. A method system for semiconductor device perturbation prediction and control, comprising: It comprises: a process recipe analysis module, which predefines a mapping rule between process instructions and disturbance events before the semiconductor equipment is operated, and analyzes the process instructions in the process recipe during the operation of the semiconductor equipment; a disturbance prediction module, which screens disturbance instructions from the process recipe based on a look-ahead window mechanism, converts the disturbance instructions into disturbance events based on the mapping rule, and maps each disturbance event to a prediction time axis to generate a corresponding disturbance prediction curve; a feedforward compensation module, which generates a plurality of feedforward compensation components based on the disturbance prediction curve according to disturbance types, superimposes the feedforward compensation components to obtain comprehensive feedforward compensation control quantities, maps each comprehensive feedforward compensation control quantity to the prediction time axis to obtain a feedforward compensation curve; a feedback control module, which calculates a temperature error based on a current temperature value of the temperature control system and a target temperature effective at the current time, and obtains a feedback control quantity in combination with a PID algorithm; a cooperative output module, which superimposes the comprehensive feedforward compensation control quantity and the feedback control quantity at the current time to obtain a final control output quantity, and sends the final control output quantity to the temperature control system to drive the heating or cooling device; The feedforward compensation is used to generate compensation control quantities in advance based on the disturbance prediction curve to offset the influence of predicted disturbances on the temperature control system. The feedback control is used to generate correction control quantities based on the deviation of the actual temperature of the temperature control system from the target temperature to correct residual disturbances that are not covered by the feedforward compensation. The feedforward compensation and the feedback control constitute a decoupled composite control architecture. The disturbance prediction comprises: screening disturbance instructions in a future period of time using a look-ahead window mechanism, and converting the disturbance instructions into disturbance events based on the mapping rule to form a disturbance event sequence; constructing a prediction time axis within the look-ahead window; projecting each disturbance event in the look-ahead window onto the prediction time axis based on an unfolding rule; point-by-point calculating and generating a disturbance prediction curve based on the projection result of the disturbance event on the prediction time axis; During the operation of the semiconductor equipment, when the process recipe is modified, the corresponding disturbance events are updated in real time.

7. An electronic device, comprising: It comprises: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the method of any one of claims 1 to 5 by executing the executable instructions.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer-executable instructions, which, when executed by a processor, implement the method according to any one of claims 1 to 5.

Images