Full-automatic multi-component glue distribution dynamic mixing proportion closed-loop control system

Abstract

The invention discloses a full-automatic multi-component dispensing dynamic mixing proportion closed-loop control system, which relates to the technical field of semiconductor manufacturing, and comprises a multi-component feeding module, a dynamic proportioning execution module, an online detection module, an intelligent control module, a mixing reaction module, an anti-pollution module and a data management module, the device has the advantages that anti-sticking conveying and rapid switching of multi-form raw materials are achieved through a plurality of independent flow channels and inner wall diamond-like coatings of the multi-component feeding module in cooperation with a flow channel inlet vibrating screen and an outlet electromagnetic valve, a double-metering-pump parallel structure and a double-axial-flow type pulse damper are adopted in the dynamic matching execution module, and the dynamic matching efficiency is improved. And by combining a nonlinear model prediction controller of the intelligent control module and BP neural network adaptive compensation, the dynamic adjustment precision of the multi-component mixing ratio is remarkably improved, the mixing uniformity and purity reach the standard, the product scrapping caused by the ratio error is greatly reduced, and the photoresist production yield and stability are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor manufacturing, in particular to a full-automatic multi-component glue dispensing dynamic mixing proportion closed-loop control system. BACKGROUND

[0002] In the field of semiconductor manufacturing, photoresist as the core material of the photolithography process, its preparation precision directly determines the resolution and yield of chip process, with the evolution of semiconductor technology to 7nm and below advanced process, photoresist formula from the traditional G / I line glue simple component to the multi-component complex system of EUV glue, the dispensing process is put forward the strict requirements of full-automatic, high precision, dynamic controllable; The prior art has certain defects, first, the multi-component dispensing precision of the prior art is low, the dynamic response is lagging, and the anti-pollution measures are insufficient, which easily leads to mixed liquid pollution and large dispensing error, second, the process optimization of the prior art relies on manual trial and error, the mixing uniformity is poor, and the cleaning is not thorough, it is difficult to maintain high-purity environment, which affects the yield of photoresist, therefore, we propose a full-automatic multi-component glue dispensing dynamic mixing proportion closed-loop control system. SUMMARY

[0003] The purpose of the present application is to provide a full-automatic multi-component glue dispensing dynamic mixing proportion closed-loop control system.

[0004] To achieve the above purpose, the present application provides the following technical scheme: a full-automatic multi-component glue dispensing dynamic mixing proportion closed-loop control system, the control system comprises: Multi-component feeding module: used for receiving liquid, powder and colloid-shaped resin, photoacid and solvent raw materials, based on independent flow channel to dynamic dispensing execution module; Dynamic dispensing execution module: receiving the flow adjustment instruction of the intelligent control module, dynamically adjusting the flow of each component feeding through the adjustable flow metering pump and proportional valve; Online detection module: set downstream of the mixing reaction module, real-time acquisition of the viscosity, density, granularity and component concentration data of the mixed liquid, and data transmission to the intelligent control module; The intelligent control module receives the real-time data of the online detection module, combines the preset formula parameters to generate dynamic flow adjustment instructions and mixing process instructions such as stirring speed and temperature, and transmits them to the dynamic dispensing execution module and the mixing reaction module respectively; The mixing reaction module receives the mixing process instruction of the intelligent control module, uniformly mixes the multi-component through stirring, dispersion and other ways, and outputs the mixed liquid to the storage unit; Anti-pollution module: integrated in the contact parts of the multi-component feeding module, dynamic dispensing execution module and mixing reaction module, to prevent raw material pollution and remove residues; Data management module: store the historical data of online detection module, process parameters of intelligent control module and running data of mixed reaction module, and provide historical data support for process optimization for intelligent control module.

[0005] As a further scheme of the present application: the multi-component feed module includes 3-10 independent flow channels, the inner wall of each flow channel is coated with a diamond-like coating, a vibrating screen with a frequency of 20-50 Hz is arranged at the inlet of the flow channel to handle the bridging and blocking of the powder raw material, and an electromagnetic valve with a response time of no more than 50 ms is arranged at the outlet of the flow channel to quickly switch different component feeds.

[0006] As a further scheme of the present application: the dynamic proportioning execution module includes a double-metering pump parallel structure composed of a large-displacement pump and a small-displacement pump and a double-shaft pulse damper, the double-metering pump is driven by an independent variable frequency motor with a speed regulation accuracy of ±0.1%, is suitable for flow control of high-viscosity raw materials with a viscosity greater than 3000 cP and low-viscosity raw materials with a viscosity less than 1 cP, and the double-shaft pulse damper reduces the flow fluctuation from ±2% to ±0.5%.

[0007] As a further scheme of the present application: the online detection module includes a Coriolis mass flowmeter with an accuracy of ±0.1% FS (for detecting the real-time mass flow of each component), a near-infrared spectrometer with a resolution of 5-20 nm (for detecting the characteristic absorption peaks of photoacid and resin in the mixed solution and calculating the component concentration), and a laser particle counter with a detection lower limit of 0.1-0.5 μm (for monitoring the number of particles with a size greater than or equal to 0.1 μm in the mixed solution), and the detection period of each sensor is no more than 200 ms.

[0008] As a further scheme of the present application: the intelligent control module includes a nonlinear model predictive controller and an adaptive compensation unit, the nonlinear model predictive controller establishes a state space model based on the viscosity-temperature coupling of photoresist components, the diffusion rate of photoacid and the molecular weight correlation of resin, generates a flow adjustment instruction based on a rolling optimization algorithm, and the adaptive compensation unit identifies the viscosity change of the raw material in real time through a BP neural network and corrects the flow set value of the metering pump.

[0009] As a further scheme of the present application: the mixed reaction module includes a double-shaft planetary stirrer and an ultrasonic disperser, the stirring paddle of the double-shaft planetary stirrer has a variable pitch structure with a pitch ratio of 1:2-3, the rotation speed range is 0-3000 rpm and the adjustment accuracy is ±1 rpm, and the frequency of the ultrasonic disperser is 20-40 kHz, which assists in dispersing the agglomerates of high-viscosity resin and photoacid.

[0010] As a further solution of the present invention, the anti-pollution module includes a perfluorinated material pipeline with a metal ion precipitation amount not exceeding 0.1 ppb (for preventing metal contamination), a nitrogen positive pressure environment control unit with an oxygen content not exceeding 1 ppm and a humidity not exceeding 5% RH (for preventing photoacid oxidation), and an ultrasonic vibrator with a frequency of 40 kHz to 50 kHz and arranged on the inner wall of the flow channel and the mixing chamber to remove the photoresist adhesion residue to an adhesion force not exceeding 0.5 N / cm 2 -1N / cm 2 .

[0011] As a further solution of the present invention: the data management module includes a process database and a neural network optimization unit. The process database stores no less than 5,000 sets of historical glue mixing data, including raw material batch data, process parameter data and yield result data. The neural network optimization unit trains a prediction model based on historical data and outputs optimal mixing process parameters, including stirring speed and temperature curve.

[0012] As a further solution of the present invention: the dynamic concentration prediction equation of the nonlinear model predictive controller is as follows: ; in, for Moment The concentration of the components, =1 represents resin, =2 represents photoacid, =3 represents the solvent, the unit is mass percentage, For the The feed composition The coupling weight matrix of the mixed components, 0< ≤1, reflecting the priority of the influence of feed flow rate on the mixed concentration, for Moment The flow rate of the feed component in liters per minute, For the Feed to the first The response delay time of the mixed components is determined by the sensor-actuator link, 0.1s≤ ≤0.5s, is the effective volume of the mixing chamber, For the The first mixed component and the The reaction rate constants of the feed components are related to the resin molecular weight (unit: Dalton) positive correlation, For the The mixed components A diffusion coefficient in a feed component, and a mixing temperature Satisfy the Arrhenius relationship = , is a pre-exponential factor, is an activation energy, is a gas constant, is a viscosity disturbance correction coefficient, 0.8≤ ≤1.2, is the viscosity disturbance value of the th mixed component at the moment, defined as the absolute deviation of the real-time viscosity from the reference viscosity; The dynamic concentration prediction equation determines the coupling weight matrix , the reaction rate constant and the diffusion coefficient by training historical rubber mixing data, predicts the change trend of the mixed liquid component concentration with time, and generates optimal flow regulation instructions through rolling optimization (optimization time domain N=20) to reduce the deviation of the predicted concentration from the target concentration.

[0013] As a further scheme of the application: the shear rate distribution of the double-shaft planetary stirrer satisfies the following relationship:

[0014] wherein, is the standard deviation of the shear rate distribution, is the number of sampling points in the mixing chamber (N≥100), is the shear rate of the th sampling point, is the average shear rate of the stirrer, calculated by , is the stirring speed, is the diameter of the stirring paddle, is the height of the mixing chamber; The standard deviation of the shear rate distribution is adjusted to be no more than 10%-15% of the average shear rate by adjusting the stirring speed , the diameter of the stirring paddle and the height of the mixing chamber , so as to avoid the rupture of the polymer chain.

[0015] Compared with the prior art, the beneficial effects of the application are as follows: 1、The present application realizes the anti-sticking conveying and rapid switching of multi-form raw materials through the multiple independent flow channels and the inner wall diamond-like coating of the multi-component feeding module, the inlet vibrating screen and the outlet electromagnetic valve, the dynamic proportioning execution module adopts the parallel structure of double metering pumps and double axial flow pulse dampers, combines the nonlinear model predictive controller and the BP neural network adaptive compensation of the intelligent control module, realizes the real-time decoupling of multivariable interference such as viscosity-temperature coupling, the on-line detection module realizes millisecond-level data acquisition feedback through the fusion of multiple sensors such as Coriolis mass flow meters and near-infrared spectrometers, the mixed reaction module utilizes the variable pitch structure of the double-shaft planetary stirrer and the ultrasonic disperser, and the anti-pollution module uses the full fluorine material pipeline, the nitrogen positive pressure environment and the ultrasonic vibrator to eliminate metal pollution and photoacid oxidation, so that the dynamic adjustment precision of the multi-component mixing ratio is significantly improved, the mixing uniformity and purity meet the standards, the product scrap caused by proportioning error is greatly reduced, and the yield and stability of the photoresist production are improved.

[0016] 2、The present application realizes the anti-sticking conveying and rapid switching of multi-form raw materials through the multiple independent flow channels and the inner wall diamond-like coating of the multi-component feeding module, the inlet vibrating screen and the outlet electromagnetic valve, the dynamic proportioning execution module adopts the parallel structure of double metering pumps and double axial flow pulse dampers, combines the nonlinear model predictive controller and the BP neural network adaptive compensation of the intelligent control module, realizes the real-time decoupling of multivariable interference such as viscosity-temperature coupling, the on-line detection module realizes millisecond-level data acquisition feedback through the fusion of multiple sensors such as Coriolis mass flow meters and near-infrared spectrometers, the mixed reaction module utilizes the variable pitch structure of the double-shaft planetary stirrer and the ultrasonic disperser, and the anti-pollution module uses the full fluorine material pipeline, the nitrogen positive pressure environment and the ultrasonic vibrator to eliminate metal pollution and photoacid oxidation, so that the dynamic adjustment precision of the multi-component mixing ratio is significantly improved, the mixing uniformity and purity meet the standards, the product scrap caused by proportioning error is greatly reduced, and the yield and stability of the photoresist production are improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The system flowchart in the embodiment of the present application is shown in the figure. Figure 2 The closed-loop control flowchart of the intelligent control module in the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0018] The specific embodiments of the present application will be further described below in conjunction with the drawings, and it should be noted that the description of these embodiments is used to help understand the present application and does not constitute a limitation on the present application.

[0019] In addition, the technical features involved in each of the embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0020] Please refer to the drawingsFigure 1 -attach Figure 2 The application is a full-automatic multi-component dispensing glue dynamic mixing ratio closed-loop control system, which comprises the following components: A multi-component feeding module is used to receive raw materials in the forms of liquid, powder and glue, photoacid and solvent, and deliver them to a dynamic dispensing execution module based on independent flow channels; The dynamic dispensing execution module receives flow adjustment instructions from the intelligent control module, and dynamically adjusts the flow of each component through the adjustable flow metering pump and proportional valve; An online detection module is arranged downstream of the mixing reaction module, which is used to collect the viscosity, density, granularity and component concentration data of the mixed liquid in real time, and transmit the data to the intelligent control module; The intelligent control module receives real-time data from the online detection module, combines preset formula parameters to generate dynamic flow adjustment instructions and mixing process instructions such as stirring speed and temperature, and transmits them to the dynamic dispensing execution module and the mixing reaction module, respectively; The mixing reaction module receives the mixing process instructions from the intelligent control module, uniformly mixes the multi-component through stirring and dispersion, and outputs the mixed liquid to a storage unit, which is a sealed tank made of perfluorinated material and has a built-in magnetic stirrer; A pollution prevention module is integrated with the contact components of the multi-component feeding module, the dynamic dispensing execution module and the mixing reaction module to prevent raw material pollution and remove residues; A data management module is used to store historical data of the online detection module, process parameters of the intelligent control module and operation data of the mixing reaction module, and provide historical data support for process optimization of the intelligent control module.

[0021] In an embodiment of the application, the multi-component feeding module comprises 3-10 independent flow channels, the inner wall of each flow channel is coated with a diamond-like coating, a vibrating screen with a frequency of 20-50 Hz is arranged at the inlet of the flow channel to handle the bridging and clogging of powder raw materials, and an electromagnetic valve with a response time of no more than 50 ms is arranged at the outlet of the flow channel to quickly switch different component feedings.

[0022] In an embodiment of the application, the dynamic dispensing execution module comprises a double-metering pump parallel structure composed of a large-displacement pump and a small-displacement pump, and a double-shaft pulse damper, the double-metering pump is driven by an independent variable frequency motor with a speed regulation accuracy of ±0.1%, which is suitable for flow control of high-viscosity raw materials with a viscosity greater than 3000 cP and low-viscosity raw materials with a viscosity less than 1 cP, and the double-shaft pulse damper reduces the flow fluctuation from ±2% to ±0.5%.

[0023] In one embodiment of the present invention: the online detection module includes a Coriolis mass flowmeter with an accuracy of ±0.1% FS (used to detect the real-time mass flow of each component), a near-infrared spectrometer with a resolution of 5nm-20nm (used to detect the characteristic absorption peaks of photoacid and resin in the mixed liquid and calculate the component concentrations), and a laser particle counter with a detection limit of 0.1μm-0.5μm (used to monitor the number of particles with a size ≥0.1μm in the mixed liquid), and the detection cycle of each sensor does not exceed 200ms.

[0024] In one embodiment of the present invention: the intelligent control module includes a nonlinear model predictive controller and an adaptive compensation unit. The nonlinear model predictive controller establishes a state space model based on dynamic characteristics such as viscosity-temperature coupling of photoresist components and correlation between photoacid diffusion rate and resin molecular weight, and generates flow regulation instructions based on a rolling optimization algorithm. The adaptive compensation unit uses a BP neural network to identify changes in raw material viscosity in real time and correct the flow setting value of the metering pump.

[0025] In one embodiment of the present invention: the mixing reaction module includes a dual-shaft planetary agitator and an ultrasonic disperser, the stirring propeller of the dual-shaft planetary agitator is a variable pitch structure with a pitch ratio of 1:2-3, a speed range of 0-3000rpm and an adjustment accuracy of ±1rpm, and the frequency of the ultrasonic disperser is 20kHz-40kHz, which assists in dispersing the agglomerates of high viscosity resin and photoacid.

[0026] In one embodiment of the present invention, the anti-pollution module includes a perfluorinated material pipeline with a metal ion precipitation amount of no more than 0.1 ppb (to avoid metal contamination), a nitrogen positive pressure environment control unit with an oxygen content of no more than 1 ppm and a humidity of no more than 5% RH (to prevent photoacid oxidation), and an ultrasonic vibrator with a frequency of 40 kHz to 50 kHz and arranged on the inner wall of the flow channel and the mixing chamber to remove the photoresist adhesion residue to an adhesion force of no more than 0.5 N / cm 2 -1N / cm 2 .

[0027] In one embodiment of the present invention: the data management module includes a process database and a neural network optimization unit. The process database stores no less than 5,000 sets of historical glue mixing data, including raw material batch data, process parameter data and yield result data. The neural network optimization unit trains a prediction model based on the historical data and outputs the optimal mixing process parameters, including stirring speed and temperature curve.

[0028] In one embodiment of the present invention, the dynamic concentration prediction equation of the nonlinear model predictive controller is as follows: ; in, for Time The concentration of the component, =1 represents resin, =2 represents photoacid, =3 represents solvent, unit: mass percentage, The coupling weight matrix of the The coupling weight matrix of the The coupling weight matrix of the The coupling weight matrix of the The coupling weight matrix of the The flow of the The flow of the The flow of the The flow of the The flow of the The flow of the The effective volume of the mixing chamber, The reaction rate constant of the The reaction rate constant of the The reaction rate constant of the The reaction rate constant of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The diffusion coefficient of the The viscosity disturbance value of the The viscosity disturbance value of the The viscosity disturbance value of the The dynamic concentration prediction equation determines the coupling weight matrix The dynamic concentration prediction equation determines the coupling weight matrix The dynamic concentration prediction equation determines the coupling weight matrix The dynamic concentration prediction equation determines the coupling weight matrix

[0029] In one embodiment of the present application: the shear rate distribution of the double-shaft planetary stirrer satisfies the following relationship:

[0030] wherein, is the standard deviation of the shear rate distribution, is the number of sampling points in the mixing chamber (N≥100), is the shear rate of the th sampling point, is the average shear rate of the agitator, calculated by is the stirring speed, is the diameter of the stirring paddle, is the height of the mixing chamber; The standard deviation of the shear rate distribution is not more than 10-15% of the average shear rate by adjusting the stirring speed , the diameter of the stirring paddle and the height of the mixing chamber to avoid polymer chain rupture.

[0031] Example 1, please refer to the attached Figure 1 - Appendix Figure 2 , 28 nm process ArF photoresist dispensing process parameter acquisition method: 1. Multi-component feeding module Number of flow channels (5): Acquisition method: based on the number of ArF photoresist components (resin, photoacid, solvent, dispersant, cleaning liquid), through historical dispensing data statistics (6000 groups of data stored in the process database), 5-component scene accounts for 70%, so 5 flow channels are set to balance compatibility and cost; Vibrating screen frequency (30 Hz): Acquisition method: through single factor experiment to test the influence of different frequencies (20-50 Hz) on powder resin (particle size 50 μm) bridging blockage, when the frequency is 30 Hz, the blockage frequency decreases from 0.5 times / hour to 0.1 times / hour, which is determined as the optimal value.

[0032] Solenoid valve response time (20 ms): Acquisition method: select commercially available high-precision solenoid valve (SMCVQZ series), its technical parameter nominal response time ≤50 ms, through field test to optimize the control circuit, actual reach 20 ms.

[0033] 2. Dynamic proportioning execution module Double metering pump speed regulation accuracy (±0.1%): Acquisition method: select servo motor + planetary reducer (reduction ratio 10:1), realize closed-loop control through encoder feedback (resolution 20 bit), calibrated and tested.

[0034] Flow fluctuation control (±0.25%): ​How to obtain: Calculate the volume of the dual-axial flow pulse damper (cavity volume = 500mL), and calculate the damper frequency by combining the Bernoulli equation. By optimizing the internal flow channel structure through fluid simulation, the flow fluctuation is reduced from ±2% to ±0.5%.

[0035] 3. Intelligent control module Rolling optimization time domain (N=15): Acquisition method: Based on the sensor detection cycle (100ms) and the actuator response delay (50ms), the total lag time is calculated to be 150ms. N is set to 15 (corresponding to a predicted time of 1.5 seconds) to ensure that the transmission delay of the mixed liquid from feeding to detection is covered.

[0036] BP neural network viscosity compensation accuracy (98%): Acquisition method: A neural network (3-layer structure, 1 node in the input layer (temperature), 5 nodes in the hidden layer, and 1 node in the output layer (viscosity)) was trained using 2000 sets of resin viscosity data from the process database (temperature 20°C-30°C, viscosity 2000cP-4000cP). The root mean square error (RMSE) of the test set was 40cP, and the compensation accuracy was 1-RMSE / average viscosity, which was 98%.

[0037] 4. Mixed reaction module Pitch ratio of stirring propeller (1:2): Acquisition method: Orthogonal experiments were conducted to test the effect of pitch ratios of 1:1, 1:2, and 1:3 on mixing uniformity (indicator: particle size distribution D90-D10). When the pitch ratio was 1:2, D90-D10 = 150nm (minimum), and the corresponding shear rate standard deviation accounted for 9%.

[0038] Ultrasonic frequency (30kHz): Acquisition method: Calculate the resonance frequency based on the characteristic absorption frequency of the photoacid agent (PAG) (248nm corresponds to the acoustic wavelength ≈ 5.9μm) =1500 / 5.9≈254kHz, but considering the equipment cost and cavitation effect, 30kHz (close to the frequency doubling relationship) was selected, and the dispersion effect was verified by particle size test.

[0039] Example 2, please refer to the attached Figure 1 -Attached Figure 2 , 3nm process EUV photoresist dispensing process: 1. Application Scenario Preparation of metal oxide photoresist suitable for 3nm EUV lithography process, the main components are SnO2 nanoresin (viscosity 5000 cP), EUV photoacid (photosensitivity <50 mJ / cm2), fluorinated solvent (viscosity 1 cP), the target ratio is resin:photoacid:solvent=4:3:3 (mass ratio), the particle size requirement is ≥0.05 μm particle ≤1 / mL, the metal ion content is <0.5 ppb.

[0040] 2. Parameter setting of each module Multi-component feed module Flow channel configuration: 8 independent flow channels (resin, photoacid, solvent, surfactant, chelating agent, cleaning agent 1, cleaning agent 2, nitrogen purge), inner wall diamond-like coating (Ra=0.1 μm); Powder treatment: SnO2 nanoresin (particle size 20 nm) is transported by airflow, the inlet vibration screen frequency is 40 Hz, and the outlet electromagnetic valve response time is 15 ms.

[0041] Dynamic ratio execution module Double metering pump configuration: Large displacement pump (solvent / chelating agent): magnetic drive centrifugal pump, flow range 0-15 L / min, speed regulation accuracy ±0.08%; Small displacement pump (resin / photoacid): screw metering pump, flow range 0-3 L / min, speed regulation accuracy ±0.05%; Pulse damper: three-stage series structure, flow fluctuation reduced to ±0.2%.

[0042] Online detection module Coriolis mass flowmeter: accuracy ±0.08% FS, monitoring resin flow 1.6 L / min, photoacid flow 1.2 L / min, and solvent flow 1.2 L / min; Near-infrared spectrometer: resolution 10 nm, detecting SnO2 resin characteristic peak (λ=193 nm), concentration deviation ≤±0.15%; Laser particle counter: detection lower limit 0.1 μm, real-time particle number display 0.6 / mL (up to standard).

[0043] Intelligent control module Nonlinear model predictive controller: State vector: resin concentration , photoacid concentration , solvent concentration , viscosity ; Coupling weight matrix: =0.7 (resin feed has the greatest impact on resin concentration), =0.2 (solvent feed has the least impact on photoacid concentration); Arrhenius model: Diffusion system = ,in =40kJ / mol, real-time temperature compensation accuracy ±0.1℃.

[0044] Mixed reaction module Twin-shaft planetary mixer: The pitch ratio is 1:3, the diameter of the stirring paddle is 0.2m, and the height of the mixing chamber is 0.4m; Speed ​​2000rpm, average shear rate =2000s −1 , standard deviation =200s −1 (accounting for 10%); Ultrasonic disperser: frequency 40kHz, action time 45min, the dispersion of SnO2 nanoparticles decreased from D90=300nm to D90=150nm.

[0045] Anti-pollution module Perfluorinated material piping: Metal ion precipitation amount 0.03ppb, nitrogen positive pressure environment (oxygen content 0.3ppm, humidity 2%RH); Ultrasonic vibrator: frequency 50kHz, activated for 8s every 20min, the adhesion force on the inner wall of the flow channel is reduced to 1.0N / cm².

[0046] Data Management Module Process database: stores 8,000 sets of data, including resin batches (SnO2 particle size distribution 10-30nm), mixing temperature (15±0.1℃); Neural network optimization: Outputs the optimal temperature curve (heating to 20°C in the first 10 minutes and cooling to 15°C in the last 35 minutes), with a predicted yield of 92% (actual yield 91.5%, error ±0.5%).

[0047] 3. Run the process Feeding stage: 8 flow channels are opened in stages. Resin / photoacid is first fed at a low speed (0.5L / min) by a small displacement pump, and solvent / chelating agent is then fed at a high speed (10L / min) by a large displacement pump. The feeding is completed within 15 minutes. Mixing stage: The stirrer was premixed at a medium speed of 800 rpm for 8 min, then increased to a high speed of 2000 rpm for 40 min, and the ultrasonic disperser was operated intermittently for 5 min every 10 min; Detection and adjustment: The online detection module found that the photoacid concentration was 0.2% lower. The intelligent control module activated the small-displacement pump to add 0.03L of photoacid, and the correction was completed within 1 minute. Discharge and cleaning: the mixed solution is discharged through a 0.02 μm filter, cleaning agent 1 (fluorinated solvent) and cleaning agent 2 (ultra-pure water) pass through the flow channel in turn, combined with ultrasonic vibration, and the single cleaning time is 60 min, and the residual rate is <0.1% 4. Effect verification Proportioning accuracy: the proportioning error of resin / photoacid / solvent is +0.15%, -0.1% and +0.05% respectively, which meets the requirement of ±0.3%; Particle size: ≥0.05 μm particles 0.6 per mL, which meets the standard; Yield: used for 3 nm EUV photoetching process, chip line width roughness (LWR) 2.5 nm (requirement <3 nm), yield 91.5%.

[0048] Through the multiple independent flow channels of the multi-component feeding module, the inner wall diamond-like coating, the inlet vibrating screen and the outlet electromagnetic valve of the flow channel, the parallel structure of the double metering pumps of the dynamic proportioning execution module and the double-shaft flow type pulse damper, the nonlinear model predictive controller and the BP neural network adaptive compensation of the intelligent control module, the multi-sensor fusion of the Coriolis mass flowmeter and the near-infrared spectrometer of the online detection module, the variable pitch structure of the double-shaft planetary stirrer and the ultrasonic disperser of the mixed reaction module, the full fluorine material pipeline, the nitrogen positive pressure environment and the ultrasonic vibrator of the pollution prevention module, the process database and the neural network optimization unit of the data management module and other multiple modules, the dynamic adjustment precision of the multi-component mixing ratio is effectively improved, the mixing uniformity and purity meet the standard, the high molecular chain rupture is avoided, the heavy metal pollution and photoacid oxidation are eliminated, the photoresist formula optimization period is shortened, the mixed liquid particle size distribution uniformity is improved, the influence of raw material batch fluctuation on yield is reduced, and by means of the cooperation of multiple technologies in each module, the product scrap caused by proportioning error can be greatly reduced, and the yield and stability of the photoresist production are improved.

[0049] Although the present application is disclosed in the preferred embodiments as above, it is not intended to limit the present application, and any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, any modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, which does not deviate from the technical solutions of the present application, falls within the protection scope defined by the claims of the present application.

Claims

1. A fully automatic multi-component glue dynamic mixing ratio closed-loop control system, characterized in that: The control system includes: Multi-component feeding module: used to receive raw materials of resin, photoacid and solvent in liquid, powder and colloid form; Dynamic proportioning execution module: receives the flow adjustment instruction from the intelligent control module and dynamically adjusts the feed flow of each component through the adjustable flow metering pump and proportional valve; Online detection module: real-time collection of viscosity, density, particle size and component concentration data of the mixed liquid; The intelligent control module receives real-time data from the online detection module, combines it with preset recipe parameters, generates dynamic flow adjustment instructions and mixing process instructions, and transmits them to the dynamic proportioning execution module and the mixing reaction module respectively; The mixing reaction module receives the mixing process instruction from the intelligent control module, uniformly mixes the multiple components by stirring and dispersing, and outputs the mixed liquid to the storage unit; Anti-pollution module: integrated into the contact parts of the multi-component feeding module, dynamic proportioning execution module and mixing reaction module to prevent raw material contamination and remove residues; Data management module: stores historical data of the online detection module, process parameters of the intelligent control module and operating data of the mixing reaction module, and provides historical data support for process optimization of the intelligent control module.

2. The fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 1, characterized in that: The multi-component feeding module includes 3-10 independent flow channels, the inner wall of each flow channel is coated with a diamond-like coating, a vibrating screen with a frequency of 20Hz-50Hz is set at the flow channel inlet to deal with bridging blockage of powder raw materials, and a solenoid valve with a response time of no more than 50ms is set at the flow channel outlet to quickly switch between different component feeds.

3. The fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 2, characterized in that: The dynamic proportioning execution module includes a dual-metering pump parallel structure consisting of a large-displacement pump and a small-displacement pump, and a dual-axial flow pulse damper. The dual-metering pump is driven by an independent variable-frequency motor with a speed adjustment accuracy of ±0.1%, and is suitable for flow control of high-viscosity raw materials with a viscosity greater than 3000cP and low-viscosity raw materials with a viscosity less than 1cP.

4. The fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 3, characterized in that: The online detection module includes a Coriolis mass flowmeter with an accuracy of ±0.1% FS, a near-infrared spectrometer with a resolution of 5nm-20nm, and a laser particle counter with a detection limit of 0.1μm-0.5μm, and the detection cycle of each sensor does not exceed 200ms.

5. The fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 4, characterized in that: The intelligent control module includes a nonlinear model predictive controller and an adaptive compensation unit. The nonlinear model predictive controller establishes a state space model based on the dynamic characteristics of the photoresist components and generates flow adjustment instructions based on a rolling optimization algorithm. The adaptive compensation unit uses a BP neural network to identify changes in raw material viscosity in real time and correct the flow setting value of the metering pump.

6. The fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 5, characterized in that: The mixing reaction module includes a dual-shaft planetary agitator and an ultrasonic disperser. The agitator of the dual-shaft planetary agitator has a variable pitch structure with a pitch ratio of 1:2-3, a speed range of 0-3000rpm and an adjustment accuracy of ±1rpm. The frequency of the ultrasonic disperser is 20kHz-40kHz, which assists in dispersing the agglomerates of high-viscosity resin and photoacid.

7. The fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 6, characterized in that: The anti-pollution module includes a perfluorinated material pipeline with a metal ion precipitation amount of no more than 0.1ppb, a nitrogen positive pressure environment control unit that controls the oxygen content to no more than 1ppm and the humidity to no more than 5%RH, and an ultrasonic vibrator with a frequency of 40kHz-50kHz and set on the inner wall of the flow channel and the mixing chamber to remove the photoresist adhesion residue until the adhesion force does not exceed 0.5N / cm 2 -1N / cm 2 .

8. The fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 7, characterized in that: The data management module includes a process database and a neural network optimization unit. The process database stores no less than 5,000 sets of historical glue mixing data, including raw material batch data, process parameter data and yield result data. The neural network optimization unit trains a prediction model based on historical data and outputs the optimal mixing process parameters, including stirring speed and temperature curve.

9. A fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 8, characterized in that: The dynamic concentration prediction equation of the nonlinear model predictive controller is as follows: ； in, for Moment The concentration of the components, =1 represents resin, =2 represents photoacid, =3 represents the solvent, the unit is mass percentage, For the The feed composition The coupling weight matrix of the mixed components, 0< ≤1, for Moment The flow rate of the feed component in liters per minute, For the Feed to the first The response delay time of the mixed components is determined by the sensor-actuator link, 0.1s≤ ≤0.5s, is the effective volume of the mixing chamber, For the The first mixed component and the The reaction rate constants of the feed components are related to the resin molecular weight Positive correlation, For the The mixed components The diffusion coefficients of the feed components are related to the mixing temperature Satisfy the Arrhenius relation = , is the pre-exponential factor, is the activation energy, is the gas constant, is the viscosity disturbance correction coefficient, 0.8≤ ≤1.2, for Moment The viscosity disturbance value of the mixed component is defined as the absolute deviation between the real-time viscosity and the reference viscosity; The dynamic concentration prediction equation is trained by historical rubber data to determine the coupling weight matrix , reaction rate constant and diffusion coefficient , predict the changing trend of the concentration of the mixed liquid components over time, and generate flow adjustment instructions through rolling optimization to reduce the deviation between the predicted concentration and the target concentration.

10. A fully automatic multi-component glue dispensing dynamic mixing ratio closed-loop control system according to claim 9, characterized in that: The shear rate distribution of the twin-shaft planetary mixer satisfies the following relationship: in, is the standard deviation of the shear rate distribution, is the number of sampling points in the mixing chamber, For the The shear rate at each sampling point, is the average shear rate of the agitator, calculate, is the stirring speed, is the diameter of the stirring paddle, is the height of the mixing chamber; By adjusting the stirring speed , stirring paddle diameter and mixing chamber height , so that the standard deviation of the shear rate distribution does not exceed 10%-15% of the average shear rate, avoiding polymer chain breakage.

Images