Interlock control method for semiconductor process machine side gas purification discharge

Abstract

The present application relates to the technical fields of semiconductor process gas purification control, and discloses a semiconductor process machine side gas purification discharge interlocking control method, which collects in real time the impurity concentration at the inlet of a purification unit, gas flow, body temperature, inlet pressure and outlet instantaneous concentration sequence; calculates the current residual adsorption capacity according to the inlet cumulative adsorption flux and temperature and pressure correction coefficients, and simultaneously performs sliding window residual accumulation processing on the outlet concentration sequence to obtain a slow release trend offset; reconstructs the residual adsorption capacity and the slow release trend offset in phase space and extracts the exponential divergence rate as a health deviation index; compares the health deviation index with a dynamic deviation threshold value, and generates a preventive trigger signal before the outlet instantaneous concentration reaches the alarm limit; and in response to the signal, performs standby path switching or suspends sensitive steps; the present application can intervene in advance during the early stage of saturation and slow release of the purification unit, and avoid batch wafer defects caused by delayed release.

Description

Technical Field

[0001] This invention relates to the field of semiconductor process gas purification control technology, specifically to a method for interlocking control of gas purification and emission on the semiconductor process equipment side. Background Technology

[0002] The gases used in semiconductor processes, such as silane and germane, are spontaneously combustible and highly toxic, while hydrogen chloride is highly corrosive. Therefore, the safe operation of their purification and supply systems is crucial. Existing interlocking control methods for gas purification emissions at the semiconductor process equipment side fail to identify the "delayed release" phenomenon of trace impurities after adsorption capacity saturation because they only monitor the instantaneous impurity concentration at the purification unit outlet. This leads to a technical problem where multiple process batches experience continuous wafer surface oxide defects or particulate contamination even when the outlet concentration remains below the alarm threshold, and the root cause of the failure is difficult to trace. Summary of the Invention

[0003] The purpose of this invention is to provide a method for interlocking control of gas purification and emission on the semiconductor process equipment side, so as to solve the problems mentioned above.

[0004] The objective of this invention can be achieved through the following technical solutions: A method for interlocking control of gas purification and emission on the semiconductor process equipment side includes the following steps: S1: Real-time acquisition of impurity concentration, gas flow rate, purification unit body temperature and inlet pressure at the purification unit inlet, and simultaneous acquisition of instantaneous impurity concentration sequence at the purification unit outlet; S2: Based on the time-cumulative value of inlet impurity concentration and gas flow rate, combined with the dynamic correction coefficient of the total adsorption capacity of the purification unit by the body temperature and inlet pressure, calculate the current remaining adsorption capacity. At the same time, perform sliding window residual accumulation processing on the outlet instantaneous impurity concentration sequence to obtain the outlet slow release trend offset. S3: Using the current remaining adsorption capacity and the outlet slow-release trend offset as two embedding dimensions, the two-dimensional time series formed by the two is reconstructed in phase space, and the exponential divergence rate of adjacent point pairs in the reconstruction trajectory is extracted. The exponential divergence rate is used as a single purification health deviation index. S4: Compare the single purification health deviation index with a deviation threshold. When the single purification health deviation index is higher than the deviation threshold, generate a preventive trigger signal before the outlet instantaneous concentration reaches the instantaneous alarm limit. S5: In response to a preventative trigger signal, execute at least one interlock action, the interlock action including automatically switching to a backup gas path or suspending a sensitive step in the current process batch.

[0005] As a further aspect of the present invention: obtaining the export slow-release trend offset specifically includes: Based on the body temperature and inlet pressure, temperature correction coefficients and pressure correction coefficients are fitted respectively. The baseline total adsorption capacity of the purification unit calibrated at the factory is multiplied by the product of the two correction coefficients to obtain the dynamic total adsorption capacity under the current operating conditions. The current remaining adsorption capacity is obtained by subtracting the cumulative value of inlet impurity concentration and gas flow rate over time from the dynamic total adsorption capacity. The instantaneous impurity concentration sequence at the outlet is divided into multiple overlapping windows according to time order. The residual between each concentration value and the mean value in each window is calculated. The absolute values ​​of the residuals in each window are then summed exponentially attenuated. The summed value is the outlet slow-release trend offset.

[0006] As a further aspect of the present invention: the step of fitting temperature correction coefficients and pressure correction coefficients based on the body temperature and inlet pressure respectively specifically includes: The temperature of the main body is compared with the optimal operating temperature of the purification unit as specified at the factory to obtain the temperature deviation. Based on the temperature deviation, a temperature correction coefficient is extracted on the preset thermal decay characteristic curve of the adsorption material through nonlinear interpolation. The pressure correction coefficient is obtained by comparing the inlet pressure with the critical pressure drop of the internal flow channel of the purification unit and by inversely mapping the gas compressibility factor offset induced by pressure fluctuation. The dynamic correction factor for correcting the total adsorption capacity is obtained by multiplying the temperature correction factor and the pressure correction factor by the square root.

[0007] As a further aspect of the present invention: the extraction of the exponential divergence rate of adjacent point pairs in the reconstructed trajectory specifically includes: The current remaining adsorption capacity and the outlet slow release trend offset obtained from multiple consecutive sampling times are arranged in chronological order to form a two-dimensional state point sequence. For each state point, the dimension is expanded by a fixed time lag value. The state point is combined with its preceding and following adjacent points to construct a high-dimensional phase point. The nearest neighbor phase point of each phase point is searched in the high-dimensional space. Calculate the ratio of the distance between each phase point and its nearest neighbor at the next time step to the current distance, and take the arithmetic mean of the natural logarithms of all ratios. The resulting average is the exponential divergence rate.

[0008] As a further aspect of the present invention: for each state point, dimensional expansion is performed with a fixed time lag value, and the state point is combined with its preceding and following adjacent points to construct a high-dimensional phase point. The nearest neighbor phase point of each phase point is then searched in the high-dimensional space. Specifically, this includes: Obtain a number of consecutive state points before and after the current state point, arrange them in ascending order of time, copy them and delay them by a fixed time step, and cross-combine the original sequence and the delayed sequence to form a higher-dimensional embedding vector. Calculate the Euclidean distance between each embedding vector and all other embedding vectors one by one, and select the vector with the smallest distance other than itself as the nearest neighbor. Record the ratio of the vector distance between each pair of nearest neighbor points at the next time step to the vector distance at the current time step, for use in subsequent calculations of the exponential divergence rate.

[0009] As a further aspect of the present invention: the generation of the preventive triggering signal specifically includes: Obtain the single purification health deviation index for multiple consecutive sampling times prior to the current sampling time, and calculate the moving median and moving absolute deviation of the corresponding sequences; The sum of the sliding median and five times the sliding absolute deviation is used as the deviation threshold at the current moment; If the single purification health deviation index at the current moment exceeds the deviation threshold, and the single purification health deviation index does not fall back below the threshold in the next two consecutive sampling moments, a preventive trigger signal will be generated before the instantaneous concentration at the outlet reaches the instantaneous alarm limit.

[0010] As a further aspect of the present invention: the generation of the preventive triggering signal specifically includes: Based on the extent to which the single purification health deviation index exceeds the deviation threshold, three progressively increasing risk levels are defined. When the first risk level is reached, a first sub-signal is generated. The first sub-signal is used to prohibit the current process equipment from starting a new process batch, but does not interfere with the batch that is already running. After the first sub-signal is generated, if the single purification health deviation index continues to rise to the second risk level within two consecutive sampling cycles, a second sub-signal is generated based on the retention of the first sub-signal. The second sub-signal is used to trigger the preheating and purging preparation of the backup gas path. After the second sub-signal is generated, if the single purification health deviation index further reaches the third risk level, a third sub-signal is generated based on the first and second sub-signals. The third sub-signal is used to close the gas supply valve upstream of the purification unit and discharge the residual gas in the purification unit to the waste gas treatment device.

[0011] As a further aspect of the present invention: the automatic switching to the backup gas path specifically includes: After confirming that the inlet valve of the backup gas passage is closed and the difference between the outlet pressure and the current process gas pressure is less than the set tolerance, the preheating temperature of the backup gas passage is increased to be consistent with the outlet gas temperature of the current purification unit. The inlet valve of the backup gas passage is opened with an initial flow rate of 1 / 10 of the current process gas flow rate, while the output flow rate of the current purification unit is decreased by the same amount until the current purification unit is completely shut down and the backup gas passage reaches full load flow rate. After the current output flow of the purification unit is zero, close its outlet valve and open its vent valve to direct the residual gas in the purification unit to the waste gas treatment pipeline.

[0012] The beneficial effects of this invention are: (1) This invention can identify the risk state of adsorption capacity approaching saturation and potential delayed release before the instantaneous concentration at the outlet of the purification unit reaches the traditional alarm limit. By combining the cumulative adsorption flux at the inlet with the residual of the sliding window at the outlet concentration, and further using the phase space reconstruction extraction index divergence rate as the health deviation index, this method effectively captures the dynamic characteristics of the purification unit entering the early stage of saturated slow release. Thus, it triggers preventive interlocking actions before impurities are released into the process chamber in batches, avoiding the continuous generation of oxide defects or particulate contamination in multiple process batches due to delayed release, and solving the problem of difficulty in tracing the root cause after defects occur in the prior art.

[0013] (2) This invention employs an interlocking execution strategy that combines multi-level risk levels with cross-flow switching, maximizing process continuity while ensuring safety. When the health deviation index exceeds the dynamic threshold, three sub-signals are generated sequentially based on the magnitude of the deviation: prohibiting the start-up of new batches, preheating and purging the backup gas path, and shutting off the gas supply and venting the system. This achieves a smooth transition from early warning to intervention. In particular, by using a synchronous control method that decreases the flow rate of the current purification unit at the same rate while increasing the flow rate of the backup path, the total flow rate of the process gas remains constant during the switching process, avoiding disturbances to the process chamber pressure or plasma stability caused by sudden changes in flow rate. This allows for redundant switching of the gas path without interrupting the currently running batch. Attached Figure Description

[0014] The invention will now be further described with reference to the accompanying drawings.

[0015] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart of the process for obtaining the export slow-release trend offset in this invention; Figure 3 This is a flowchart of the process for extracting the exponential divergence rate of adjacent point pairs in the reconstructed trajectory in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 As shown, this invention is a gas purification and emission interlock control method for semiconductor process equipment, comprising the following steps: S1: Real-time acquisition of impurity concentration, gas flow rate, purification unit body temperature and inlet pressure at the purification unit inlet, and simultaneous acquisition of instantaneous impurity concentration sequence at the purification unit outlet; S2: Based on the time-cumulative value of inlet impurity concentration and gas flow rate, combined with the dynamic correction coefficient of the total adsorption capacity of the purification unit by the body temperature and inlet pressure, calculate the current remaining adsorption capacity. At the same time, perform sliding window residual accumulation processing on the outlet instantaneous impurity concentration sequence to obtain the outlet slow release trend offset. S3: Using the current remaining adsorption capacity and the outlet slow-release trend offset as two embedding dimensions, the two-dimensional time series formed by the two is reconstructed in phase space, and the exponential divergence rate of adjacent point pairs in the reconstruction trajectory is extracted. The exponential divergence rate is used as a single purification health deviation index. S4: Compare the single purification health deviation index with a deviation threshold. When the single purification health deviation index is higher than the deviation threshold, generate a preventive trigger signal before the outlet instantaneous concentration reaches the instantaneous alarm limit. S5: In response to a preventative trigger signal, execute at least one interlock action, the interlock action including automatically switching to a backup gas path or suspending a sensitive step in the current process batch.

[0018] In S1, the impurity concentration, gas flow rate, purification unit temperature, and inlet pressure at the purification unit inlet are collected in real time, and the instantaneous impurity concentration sequence at the purification unit outlet is collected simultaneously, specifically including: A trace moisture analyzer is installed on the inlet pipe of the purification unit. This analyzer uses tunable diode laser absorption spectroscopy technology, with a detection limit not exceeding 0.1 ppb, to measure the concentration of impurities in the gas entering the purification unit in real time. These impurities include at least water vapor, oxygen, or carbon dioxide. Simultaneously, a thermal mass flow meter is connected in series on the inlet pipe, with a measurement range of 0 to 50 standard liters per minute and an accuracy of ±1% of full scale, for continuous gas flow rate acquisition. A platinum resistance temperature sensor is attached to the surface of the purification unit's outer casing, with a range of 0 to 100 degrees Celsius and an accuracy of ±0.1 degrees Celsius, to acquire the operating temperature of the purification unit. A piezoresistive pressure transmitter is installed on the inlet pipe near the purification unit flange, with a range of 0 to 0.5 MPa and an accuracy of ±0.5% of full scale, to acquire the inlet pressure.

[0019] All the aforementioned sensors collect data in parallel at a sampling frequency of twice per second and connect to the programmable logic controller (PLC) via an analog input channel. The controller has an internal data buffer that records a set of data at each sampling moment, including impurity concentration, gas flow rate, body temperature, and inlet pressure, along with a timestamp of the current moment. Simultaneously, the controller reads the instantaneous impurity concentration at the outlet from another trace moisture analyzer on the purification unit's outlet pipeline. This analyzer also uses laser absorption spectroscopy, has a detection limit of 0.05 ppb, and maintains the same sampling frequency as the inlet side, forming a sequence of instantaneous impurity concentrations at the outlet.

[0020] Each time the purification unit is activated or a regeneration operation is performed, the timer inside the controller is automatically reset to zero, and the runtime is accumulated starting from the first sampling moment after the reset. The runtime is recorded in seconds and stored in a circular storage queue along with the collected data. The queue depth is at least one hour of historical data. All collected raw data is used directly for calculations in subsequent steps without any filtering or smoothing.

[0021] To ensure the reliability of data acquisition, the controller automatically performs a zero-point calibration of the sensors every 24 hours. During calibration, the process gas supply is suspended, and the pipeline is purged with high-purity nitrogen. If any sensor output exceeds its normal operating range or a signal is lost, the controller immediately issues a fault alarm and suspends the current process batch. The process can only resume after the fault is manually resolved.

[0022] Please see Figure 2 As shown, in S2, based on the time-cumulative values ​​of inlet impurity concentration and gas flow rate, combined with the dynamic correction coefficient of the total adsorption capacity of the purification unit based on the bulk temperature and inlet pressure, the current remaining adsorption capacity is calculated. Simultaneously, the outlet instantaneous impurity concentration sequence is processed by sliding window residual accumulation to obtain the outlet slow-release trend offset, specifically including: The process for obtaining the outlet slow-release trend offset is as follows: The instantaneous impurity concentration sequence at the purification unit outlet is arranged chronologically. A fixed-length time window is set, containing 50 continuously collected concentration values. Multiple overlapping windows are generated by sliding one sampling point forward each time, with each window overlapping the previous one by 49 sampling points. Within each window, the arithmetic mean of the 50 concentration values ​​within that window is calculated. Then, the average value is subtracted from each concentration value within the window to obtain the residual corresponding to each concentration value. The absolute value of each residual is taken, and these absolute values ​​are then exponentially decayed and weighted: a weight is assigned to each absolute value of the residual in chronological order from oldest to newest. The oldest data has a weight of 1. Each time a sampling point is moved to the new data, the weight is multiplied by a decay factor of 0.95. The sum of the absolute values ​​of all weighted residuals is divided by the sum of the weights for normalization. The final value is the outlet slow-release trend offset at the current sampling time, which is used to characterize the cumulative trend of the outlet concentration fluctuation relative to its own average value.

[0023] The fitting process for the temperature correction coefficient is as follows: Obtain the temperature point corresponding to the baseline adsorption capacity of the adsorbent material at its optimal operating temperature from the purification unit's factory technical documents, for example, 25 degrees Celsius. Compare the real-time collected bulk temperature with this optimal temperature point and calculate the temperature deviation, which is the absolute value of the bulk temperature minus 25 degrees Celsius.

[0024] Obtain the thermal decay characteristic curve of the adsorbent material before it leaves the factory. This curve records the decay ratio of the adsorption capacity relative to the reference value under different temperature deviations. The horizontal axis represents the temperature deviation (in degrees Celsius), and the vertical axis represents the temperature correction factor (range 0 to 1). Substitute the calculated temperature deviation into the curve and use linear interpolation to determine the corresponding vertical axis value: First, find two calibration points on the curve adjacent to the deviation, and mark them as the first calibration point and the second calibration point, respectively. Add the vertical axis of the first calibration point to the difference between the vertical axes of the first and second calibration points, multiplied by the ratio of the deviation between the first and second calibration points. The result is the temperature correction factor.

[0025] The fitting process for the pressure correction coefficient is as follows: The critical pressure drop of the internal flow channel is obtained from the purification unit design data. This critical pressure drop refers to the value at which the gas flow enters a blocked flow state and the compressibility factor shifts significantly when the difference between the inlet and outlet pressures exceeds this value; for example, the critical pressure drop is 0.05 MPa. The actual pressure drop is obtained by subtracting the real-time collected inlet pressure from the purification unit outlet pressure (obtained through the outlet pipeline pressure transmitter). The ratio of the actual pressure drop to the critical pressure drop is calculated. A ratio less than 1 indicates no blocked flow, while a ratio greater than or equal to 1 indicates blocked flow. An empirical mapping table of gas compressibility factor variation with pressure drop is used. This table takes the ratio of the actual pressure drop to the critical pressure drop as input and outputs a pressure correction coefficient (range 0.8 to 1.0). The nearest neighbor lookup method is employed, finding the two ratio nodes closest to the current ratio in the mapping table and taking the average of their corresponding pressure correction coefficients to obtain the final pressure correction coefficient.

[0026] Regarding the acquisition of the dynamic correction coefficient and the calculation of the current remaining adsorption capacity: Multiply the temperature correction coefficient and pressure correction coefficient obtained above, and take the square root of the product. The result of the square root operation is the dynamic correction coefficient. Read the baseline total adsorption capacity from the purification unit's factory calibration data. This capacity is the total mass of adsorbable impurities measured under optimal operating conditions of 25 degrees Celsius, inlet pressure of 0.1 MPa, and gas flow rate of 10 standard liters per minute, in micrograms. Multiply the baseline total adsorption capacity by the dynamic correction coefficient to obtain the dynamic total adsorption capacity under the current operating conditions. Then, subtract the cumulative value of the inlet impurity concentration and gas flow rate over time (i.e., the total amount accumulated from the time the purification unit is activated or regenerated, multiplied by the gas flow rate and the sampling interval, up to the present) from the dynamic total adsorption capacity. The difference obtained is the current remaining adsorption capacity, in the same unit as the baseline total adsorption capacity.

[0027] Please see Figure 3 As shown, in S3, the current remaining adsorption capacity and the outlet slow-release trend offset are used as two embedding dimensions. The two-dimensional time series formed by the two dimensions is reconstructed in phase space, and the exponential divergence rate of adjacent point pairs in the reconstructed trajectory is extracted. The exponential divergence rate is used as a single purification health deviation index, specifically including: The calculated remaining adsorption capacity and the outlet slow-release trend offset are arranged chronologically to form a two-dimensional state point sequence. Each sampling time corresponds to a two-dimensional state point, where the first dimension represents the current remaining adsorption capacity (in micrograms) and the second dimension represents the outlet slow-release trend offset. Data is collected twice per second for 600 consecutive sampling times, i.e., 300 state points within the most recent 5 minutes, forming the original sequence for phase space reconstruction. The data in the sequence undergoes no smoothing or filtering to preserve the original dynamic characteristics.

[0028] For each state point in the above two-dimensional state point sequence, perform the dimension expansion operation in phase space reconstruction. Set a fixed time lag of 3 sampling intervals (i.e., 1.5 seconds) and an embedding dimension of 4. For the i-th state point, take the state points of that point and the third sampling point thereafter, a total of two points, each of which is two-dimensional, and combine them to form a four-dimensional initial embedding vector.

[0029] The initial embedding vector is then copied backward and delayed by one sampling step to obtain another four-dimensional delayed vector. These two four-dimensional vectors are then juxtaposed to form an eight-dimensional high-dimensional phase point. Specifically, the high-dimensional phase point is formed by sequentially concatenating four two-dimensional points: the original two-dimensional state point, a two-dimensional state point delayed by three steps, a two-dimensional state point delayed by one step, and a two-dimensional state point delayed by four steps, resulting in a total of eight coordinate components. For points near the boundary in the original sequence, if subsequent points of the required step size cannot be obtained, the last valid point in the sequence is used to fill the gap, ensuring that each state point can generate an eight-dimensional phase point.

[0030] For each of the generated eight-dimensional phase points, search for its nearest neighbor. During the search, calculate the Euclidean distance between the current phase point and all other phase points. The Euclidean distance between two eight-dimensional phase points is determined using the following mathematical formula: ;in, Indicates the first The first phase point and the second phase point The distance between each phase point Indicates the first The first phase point Each coordinate component Indicates the first The first phase point Each coordinate component, the subscript under the summation symbol Iterate from 1 to 8. After calculating all distances, exclude points with a distance of zero from the current point. Select the point with the smallest distance as the nearest neighbor of the current point and record the minimum distance value and the corresponding position of the point at the next moment. If there are multiple minimum distances with the same value, select the one with the smallest index.

[0031] For each pair of nearest neighbor phase points, obtain the distance between the two phase points at the current time, and the distance between the two new phase points formed by these two phase points at the next sampling time (i.e., shifting backward by one sampling interval). The phase points at the next time step are obtained by incrementing the time index of each state point in the original sequence by 1, and constructed according to the same dimensional expansion rule. Calculate the ratio of these two distances. Then, obtain the exponential divergence rate according to the following mathematical formula: ;in, Indicates the exponential divergence rate. This represents the total number of nearest neighbor pairs. It is the index of the nearest neighbor pair. Indicates the first The Euclidean distance between the nearest neighbor points at the current moment, This represents the Euclidean distance between the same pair of phase points at the next moment. The expression represents the natural logarithm operation. The summation symbol indicates that the ratios of all nearest-neighbor pairs are taken as natural logarithms, summed, and then divided by the total logarithm N to obtain the arithmetic mean. During calculation, if... If the value is zero or extremely small (less than 10 to the power of -6), then the pair of phase points is ignored and not included in N.

[0032] The calculated exponential divergence rate is output as a single purification health deviation index. This index is a dimensionless real number; a positive value indicates that adjacent trajectories diverge at an exponential rate (the system tends to be unstable), a negative value indicates convergence (the system tends to be stable), and zero indicates neutrality. In actual operation, this index typically changes slowly in the negative range, and rises rapidly to a positive value when the purification unit enters the early stage of saturated sustained release. This index does not depend on any preset health center point or historical distribution threshold; it is entirely reconstructed from the current dynamic data itself, thus possessing adaptive anomaly sensitivity. The index is recalculated every second, using data from the most recent 600 sampling times for each calculation, and is updated in a sliding manner, thereby achieving continuous dynamic monitoring of the purification unit's health status.

[0033] In S4, a single purification health deviation index is compared with a deviation threshold. When the single purification health deviation index is higher than the deviation threshold, a preventative trigger signal is generated before the instantaneous concentration at the outlet reaches the instantaneous alarm limit. Specifically, this includes: A sliding window with a length of 15 sampling times is defined. At each sampling time, a single purification health deviation index is calculated; this index is a dimensionless real number. At each sampling time, the controller acquires the 15 most recent index values ​​within the current window, including the current time, and arranges these values ​​in ascending order. The 8th value (i.e., the median) is taken as the sliding median. Then, for each index value within the window, the absolute difference between it and the sliding median is calculated, resulting in a set of absolute values. The median of these absolute values ​​is then taken as the sliding absolute deviation. The sliding median is added to five times the sliding absolute deviation; the sum is the deviation threshold for the current time. This threshold is dynamically updated with each sampling time and is entirely determined by recent data, requiring no preset value.

[0034] The single purification health deviation index calculated at the current sampling time is compared with the aforementioned deviation threshold. If the current index is less than or equal to the threshold, no action is triggered. If the current index is greater than the threshold, the verification phase begins: the controller continues to receive index values ​​for the next two sampling times (each sampling interval is 0.5 seconds). If the index value does not fall below the corresponding deviation threshold in either of these two consecutive times (the threshold for each time is calculated separately), the index is deemed to be validly exceeded. After being deemed valid, even if the instantaneous impurity concentration at the purification unit outlet is still far below the instantaneous alarm limit set by the process (for example, the instantaneous alarm limit for outlet water vapor concentration is 1 ppb, while the current measured value is only 0.3 ppb), the controller will still immediately generate a preventative trigger signal. If the index falls below the threshold at any point during the verification period, the trigger is canceled, the counter is reset, and monitoring resumes.

[0035] The preventative trigger signal comprises three sub-signals, increasing in risk level. The risk level is determined by the percentage by which a single purification health deviation index exceeds a deviation threshold: a deviation of less than 30% is the first risk level; a deviation of 30% or more but less than 70% is the second risk level; and a deviation of 70% or more is the third risk level. When a level of risk is determined to be first, the first sub-signal is generated. This sub-signal is transmitted to the main controller of the process equipment to prevent the equipment from starting a new batch of processes. Specifically, the "New Batch Allowed" flag in the equipment's status word is forcibly cleared, but the currently running batch remains unaffected and continues to complete normally.

[0036] If, after the first sub-signal has been generated, the single purification health deviation index continues to rise and eventually enters the second risk level in two consecutive sampling cycles (each cycle being 2 seconds), a second sub-signal is generated in addition to the first sub-signal. The second sub-signal triggers the preheating and purging preparation of the backup gas path: the controller sends a start command to the heating element in the backup path, raising the path temperature to within 2 degrees Celsius of the outlet gas temperature of the current purification unit; simultaneously, the purging nitrogen valve at the inlet of the backup path is opened, purging for 30 seconds at a flow rate of 5 standard liters per minute to ensure no residual impurities remain in the path. If the index jumps to the third risk level during its rise, it is directly processed according to the third risk level.

[0037] If, after the second sub-signal has been generated, the single purification health deviation index further increases and reaches the third risk level, a third sub-signal will be generated while retaining the first and second sub-signals. The third sub-signal performs two actions: First, it closes the gas supply valve upstream of the purification unit. This valve is a normally closed pneumatic diaphragm valve and will completely shut off within 0.3 seconds of receiving the signal. Second, it opens the vent valve downstream of the purification unit, discharging residual gas inside the purification unit to the plant's waste gas treatment device through a separate waste gas pipeline. The venting duration is 10 seconds. The three sub-signals are arranged in ascending order of priority. Once the third sub-signal is generated, it ignores any subsequent index decline and maintains all interlocking actions for at least 60 seconds, only releasing them after manual reset by the operator.

[0038] In S5, in response to a preventative trigger signal, at least one interlock action is executed, including automatically switching to a backup gas path or pausing sensitive steps in the current process batch, specifically including: Upon receiving the second or third sub-signal from the preventative trigger signal, the controller begins preparation for switching the backup gas path. First, the controller reads the on / off status feedback signal of the backup gas path inlet valve to confirm that the valve is fully closed (valve position feedback is 0%). Simultaneously, it reads the outlet pressure value via a pressure transmitter installed at the backup gas path outlet and compares it with the process gas pressure value at the outlet of the current primary purification unit. The absolute difference between the two is calculated. If this difference is less than the set tolerance of 0.01 MPa, the switching condition is considered met. If the difference exceeds the tolerance, the controller automatically adjusts the back pressure valve at the backup path outlet to reduce the pressure difference to within the tolerance range. After the pressure condition is met, the controller sends a control signal to the heating element of the backup gas path, gradually increasing the path temperature from the standby temperature (25 degrees Celsius) until it matches the outlet gas temperature of the current purification unit. The heating rate does not exceed 10 degrees Celsius per minute to avoid thermal shock.

[0039] The switching process employs a synchronous control strategy of alternating flow rate decreases and increases. The controller reads the current process gas flow rate, for example, 10 standard liters per minute (SPM). One-tenth of this flow rate, or 1 SPM, is used as the initial target flow rate for the backup gas path. The controller gradually opens the inlet mass flow controller of the backup gas path at a rate increasing by 0.2 SPM per second, while simultaneously gradually closing the mass flow controller at the output of the current purification unit at the same rate decreasing by 0.2 SPM per second. During this process, the controller continuously monitors the instantaneous sum of the two flow rates, ensuring that the sum always equals the target flow rate required by the process (i.e., 10 SPM), with fluctuations not exceeding ±0.1 SPM. The entire cross-switching process lasts approximately 45 seconds, until the output flow rate of the current purification unit drops to zero and the output flow rate of the backup gas path rises to full capacity of 10 SPM.

[0040] After the output flow rate of the current purification unit reaches zero, the controller maintains this state for 2 seconds to confirm that the flow meter reading is zero and without fluctuation. The controller sends a closing command to the outlet valve of the current purification unit; this valve is a pneumatic shut-off valve, and the execution time is 0.5 seconds. After the valve closes, the controller immediately opens the vent valve connected between the outlet of the current purification unit and the exhaust gas treatment pipeline, while keeping the inlet valve of the purification unit closed (this inlet valve was already closed upon receiving the third sub-signal). After the vent valve opens, the residual gas inside the purification unit flows naturally to the exhaust gas treatment pipeline under the pressure difference, and the venting time is 15 seconds. After the venting is completed, the controller closes the vent valve and marks the current purification unit as "isolated and awaiting maintenance".

[0041] Throughout the switching process, the controller records the flow, pressure, and temperature values ​​of both channels every second, creating a switching log. If any abnormality occurs during the switching process (e.g., a flow rate failing to follow the set value, a sudden drop in pressure, or valve feedback error), the controller immediately stops the switching process, maintains the current purification unit's output flow rate, and issues a fault alarm signal to prompt manual intervention. If the switching is successfully completed, the controller updates the process gas source flag to "backup gas path" and continues to use the backup gas path until the primary purification unit completes regeneration or is replaced and manually restored.

[0042] The working principle of this invention is as follows: Real-time acquisition of impurity concentration, gas flow rate, body temperature, inlet pressure, and instantaneous impurity concentration sequence at the purification unit inlet and outlet; calculation of the current remaining adsorption capacity based on the time-cumulative values ​​of inlet impurity concentration and gas flow rate, combined with the dynamic correction coefficients of temperature and pressure on the total adsorption capacity; simultaneous processing of the outlet impurity concentration sequence using a sliding window residual accumulation method to obtain the outlet slow-release trend offset; phase space reconstruction of the two-dimensional time series using the remaining adsorption capacity and slow-release trend offset as two embedding dimensions; extraction of the exponential divergence rate of adjacent point pairs in the reconstruction trajectory as a single purification health deviation index; comparison of the health deviation index with a deviation threshold dynamically generated from the sliding median and sliding absolute deviation; when the health deviation index exceeds the threshold, a preventative trigger signal containing three risk levels is generated before the outlet instantaneous concentration reaches the instantaneous alarm limit; in response to this trigger signal, interlocking actions such as automatic switching to the backup gas path or pausing sensitive steps in the current process batch are executed, thereby intervening in advance before the purification unit enters the early stage of saturated slow release to avoid batch wafer defects.

[0043] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for interlocking control of gas purification and emission on the side of a semiconductor process equipment, characterized in that, Includes the following steps: S1: Real-time acquisition of impurity concentration, gas flow rate, purification unit body temperature and inlet pressure at the purification unit inlet, and simultaneous acquisition of instantaneous impurity concentration sequence at the purification unit outlet; S2: Based on the time-cumulative value of inlet impurity concentration and gas flow rate, combined with the dynamic correction coefficient of the total adsorption capacity of the purification unit by the body temperature and inlet pressure, calculate the current remaining adsorption capacity. At the same time, perform sliding window residual accumulation processing on the outlet instantaneous impurity concentration sequence to obtain the outlet slow release trend offset. S3: Using the current remaining adsorption capacity and the outlet slow-release trend offset as two embedding dimensions, the two-dimensional time series formed by the two is reconstructed in phase space, and the exponential divergence rate of adjacent point pairs in the reconstruction trajectory is extracted. The exponential divergence rate is used as a single purification health deviation index. S4: Compare the single purification health deviation index with a deviation threshold. When the single purification health deviation index is higher than the deviation threshold, generate a preventive trigger signal before the outlet instantaneous concentration reaches the instantaneous alarm limit. S5: In response to a preventative trigger signal, execute at least one interlock action, the interlock action including automatically switching to a backup gas path or suspending a sensitive step in the current process batch.

2. The interlock control method for gas purification and emission on the semiconductor process equipment side according to claim 1, characterized in that, The acquisition of the export mitigation trend offset specifically includes: Based on the body temperature and inlet pressure, temperature correction coefficients and pressure correction coefficients are fitted respectively. The baseline total adsorption capacity of the purification unit calibrated at the factory is multiplied by the product of the two correction coefficients to obtain the dynamic total adsorption capacity under the current operating conditions. The current remaining adsorption capacity is obtained by subtracting the cumulative value of inlet impurity concentration and gas flow rate over time from the dynamic total adsorption capacity. The instantaneous impurity concentration sequence at the outlet is divided into multiple overlapping windows according to time order. The residual between each concentration value and the mean value in each window is calculated. The absolute values ​​of the residuals in each window are then summed exponentially attenuated. The summed value is the outlet slow-release trend offset.

3. The interlock control method for gas purification and emission on the semiconductor process equipment side according to claim 2, characterized in that, The process of fitting temperature correction coefficients and pressure correction coefficients based on the body temperature and inlet pressure respectively includes: The temperature of the main body is compared with the optimal operating temperature of the purification unit as specified at the factory to obtain the temperature deviation. Based on the temperature deviation, a temperature correction coefficient is extracted on the preset thermal decay characteristic curve of the adsorption material through nonlinear interpolation. The pressure correction coefficient is obtained by comparing the inlet pressure with the critical pressure drop of the internal flow channel of the purification unit and by inversely mapping the gas compressibility factor offset induced by pressure fluctuation. The square root of the product of the temperature correction factor and the pressure correction factor is used to obtain the dynamic correction factor for correcting the total adsorption capacity.

4. The interlock control method for gas purification and emission on the semiconductor process equipment side according to claim 1, characterized in that, The extraction of the exponential divergence rate of adjacent point pairs in the reconstructed trajectory specifically includes: The current remaining adsorption capacity and the outlet slow release trend offset obtained from multiple consecutive sampling times are arranged in chronological order to form a two-dimensional state point sequence. For each state point, the dimension is expanded by a fixed time lag value. The state point is combined with its preceding and following adjacent points to construct a high-dimensional phase point. The nearest neighbor phase point of each phase point is searched in the high-dimensional space. Calculate the ratio of the distance between each phase point and its nearest neighbor at the next time step to the current distance, and take the arithmetic mean of the natural logarithms of all ratios. The resulting average is the exponential divergence rate.

5. The interlock control method for gas purification and emission on the semiconductor process equipment side according to claim 4, characterized in that, For each state point, dimensional expansion is performed with a fixed time lag value. The state point is then combined with its preceding and following neighboring points to construct a high-dimensional phase point. Finally, the nearest neighbor phase point for each phase point is searched in the high-dimensional space. Specifically, this includes: Obtain a number of consecutive state points before and after the current state point, arrange them in ascending order of time, copy them and delay them by a fixed time step, and cross-combine the original sequence and the delayed sequence to form a higher-dimensional embedding vector. Calculate the Euclidean distance between each embedding vector and all other embedding vectors one by one, and select the vector with the smallest distance other than itself as the nearest neighbor. Record the ratio of the vector distance between each pair of nearest neighbor points at the next time step to the vector distance at the current time step, for use in subsequent calculations of the exponential divergence rate.

6. The interlock control method for gas purification and emission on the semiconductor process equipment side according to claim 1, characterized in that, The generation of the preventative trigger signal specifically includes: Obtain the single purification health deviation index for multiple consecutive sampling times prior to the current sampling time, and calculate the moving median and moving absolute deviation of the corresponding sequences; The sum of the sliding median and five times the sliding absolute deviation is used as the deviation threshold at the current moment; If the single purification health deviation index exceeds the deviation threshold at the current moment, and the single purification health deviation index does not fall back below the threshold in the next two consecutive sampling moments, a preventive trigger signal will be generated before the instantaneous concentration at the outlet reaches the instantaneous alarm limit.

7. The interlock control method for gas purification and emission on the semiconductor process equipment side according to claim 6, characterized in that, The generation of the preventative trigger signal specifically includes: Based on the extent to which the single purification health deviation index exceeds the deviation threshold, three progressively increasing risk levels are defined. When the first risk level is reached, a first sub-signal is generated. The first sub-signal is used to prohibit the current process equipment from starting a new process batch, but does not interfere with the batch that is already running. After the first sub-signal is generated, if the single purification health deviation index continues to rise to the second risk level within two consecutive sampling cycles, a second sub-signal is generated based on the retention of the first sub-signal. The second sub-signal is used to trigger the preheating and purging preparation of the backup gas path. After the second sub-signal is generated, if the single purification health deviation index further reaches the third risk level, a third sub-signal is generated based on the first and second sub-signals. The third sub-signal is used to close the gas supply valve upstream of the purification unit and discharge the residual gas in the purification unit to the waste gas treatment device.

8. The interlock control method for gas purification and emission on the semiconductor process equipment side according to claim 1, characterized in that, The automatic switching to the backup gas path specifically includes: After confirming that the inlet valve of the backup gas passage is closed and the difference between the outlet pressure and the current process gas pressure is less than the set tolerance, the preheating temperature of the backup gas passage is increased to be consistent with the outlet gas temperature of the current purification unit. The inlet valve of the backup gas passage is opened with an initial flow rate of 1 / 10 of the current process gas flow rate, while the output flow rate of the current purification unit is decreased by the same amount until the current purification unit is completely shut down and the backup gas passage reaches full load flow rate. After the current output flow of the purification unit is zero, close its outlet valve and open its vent valve to direct the residual gas in the purification unit to the waste gas treatment pipeline.

Images