Method and system for improving photoresist film thickness uniformity after gluing of gluing machine

Abstract

The invention belongs to the technical field of semiconductor manufacturing process control, and particularly discloses a method and a system for improving photoresist film thickness uniformity after gluing of a gluing machine, and the method comprises the following steps: obtaining photoresist transient thickness data online in real time through an optical system; generating a thickness deviation evolution rate map representing the radial non-uniform development trend through time differential and radial interpolation; reversely solving the radially symmetric solvent volatilization rate compensation field, and decoupling to generate a local environment regulation instruction for the annular partition environment regulation array and a global rotating speed correction instruction for the spindle motor; and finally, multi-physical field coupling regulation and control of the radial local viscosity and the global centrifugal force are realized until the whole-field thickness uniformity index is converged. According to the invention, a real-time closed-loop feedback control loop is constructed, the prospective identification of the radial nonuniformity of the film thickness and the compensation of the radial spatial resolution are realized, and the uniformity of the final film thickness and the robustness of process control are improved.

Description

A method and system for improving the uniformity of photoresist film thickness after coating. Technical Field

[0001] This invention belongs to the field of semiconductor manufacturing process control technology, and relates to a method and system for improving the uniformity of photoresist film thickness after coating by a coating machine. Background Technology

[0002] In the photolithography process of integrated circuit manufacturing, the uniformity of photoresist film thickness is one of the fundamental physical parameters that determines the consistency of critical dimensions in subsequent exposure, development, and other processes. As the feature size of semiconductor devices continues to shrink, more stringent control requirements are being placed on the uniformity of photoresist film thickness across the entire wafer surface.

[0003] Existing technologies, such as the invention application patent with publication number CN102122116A, disclose a closed-loop control method that adjusts the wafer rotation speed by acquiring the photoresist thickness in real time, comparing it with a target thickness, and then adjusting the rotation speed. It is evident that existing technologies have attempted to automatically adjust the final photoresist thickness by establishing a feedback loop. However, such methods primarily rely on the static deviation value of the thickness for global rotation speed correction. Their adjustment capability is limited for thickness deviations that are unevenly distributed in radial space during the spin-coating dynamic process, making it difficult to effectively suppress the non-uniform evolution of local radial regions.

[0004] Furthermore, existing technologies, such as the invention application patent with publication number CN113851370A, disclose a method for controlling coating thickness by pre-establishing the relationship between coating thickness and equipment exhaust pressure, and then setting the exhaust pressure accordingly. It is evident that existing technologies also recognize that adjusting the process environment, such as the chamber pressure affecting the solvent evaporation rate, is another effective way to control film thickness. However, this method typically relies on open-loop control based on fixed parameters calibrated offline, lacking real-time perception and response capabilities to random disturbances and dynamic evolution within a single process flow. Simultaneously, while this unified control of the entire process environment improves the overall average thickness, its adjustment accuracy and specificity for thickness deviations exhibiting non-uniform distribution in radial space still require improvement.

[0005] In summary, existing technologies have limitations in addressing real-time, dynamic, and spatially non-uniform film thickness deviations. Neither global rotational speed feedback based on static thickness deviations nor global environmental control based on preset parameters can effectively identify the radial non-uniformity formation process of film thickness or accurately compensate for radial spatial differences, thus affecting the uniformity of the final film thickness and the robustness of process control. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above objectives, the present invention proposes the following technical solution: a method for improving the uniformity of photoresist film thickness after coating by a coating machine, comprising: S1, controlling the spindle motor of the coating machine to drive the wafer to rotate according to a preset speed curve to establish a basic flow field, and simultaneously triggering a non-contact multi-point optical measurement system to acquire the original spectral reflection signal, and analyzing the original spectral reflection signal to generate photoresist transient thickness data.

[0007] S2. The instantaneous thickness deviation value is obtained by comparing the transient thickness data of the photoresist with the preset ideal thickness decay curve that varies with radial position. The instantaneous thickness deviation value is then differentiated over time and interpolated radially to generate a radial thickness deviation evolution rate map.

[0008] S3. Based on the radial thickness deviation evolution rate map, identify the thickening trend region and thinning trend region at different radial positions, calculate the corresponding solvent evaporation rate adjustment target value, and construct a radially symmetric solvent evaporation rate compensation field.

[0009] S4. Map the radially symmetric solvent evaporation rate compensation field to physical control parameters to generate local environmental control commands for the annular partitioned environmental control array, evaluate the residual error caused by the array execution saturation, and calculate the global speed correction command for the spindle motor based on the residual error.

[0010] S5 drives the annular partitioned environmental control array to respond to local environmental control commands to change the solvent evaporation rate of the corresponding radius annular region, and synchronously drives the spindle motor to respond to global speed correction commands to adjust the global centrifugal shear force. Through multi-physics field collaborative intervention, the non-uniform evolution trend on the corresponding radius is offset.

[0011] S6. While performing the intervention, the data acquisition and evolution rate analysis steps are executed in a loop, and the radial thickness deviation evolution rate map is updated in real time until the overall thickness uniformity index converges to the preset process tolerance window.

[0012] The second aspect of the present invention provides a system for improving the uniformity of photoresist film thickness after coating by a coating machine, comprising: a data acquisition module, which controls the spindle motor of the coating machine to drive the wafer to rotate according to a preset speed curve to establish a basic flow field, and simultaneously triggers a non-contact multi-point optical measurement system to acquire the original spectral reflection signal, and analyzes the original spectral reflection signal to generate photoresist transient thickness data.

[0013] The map construction module compares the transient thickness data of the photoresist with the preset ideal thickness decay curve that varies with radial position to obtain the instantaneous thickness deviation value. It then performs time differentiation and radial interpolation on the instantaneous thickness deviation value to generate a radial thickness deviation evolution rate map.

[0014] The compensation field calculation module identifies thickening and thinning trend regions at different radial positions based on the radial thickness deviation evolution rate map, calculates the corresponding solvent evaporation rate adjustment target value, and constructs a radially symmetric solvent evaporation rate compensation field.

[0015] The instruction generation module maps the radially symmetric solvent evaporation rate compensation field to physical control parameters to generate local environmental control instructions for the annular partitioned environmental control array, evaluates the residual error caused by the array execution saturation, and calculates the global speed correction instruction for the spindle motor based on the residual error.

[0016] The collaborative intervention module drives the annular partitioned environmental control array to respond to local environmental control commands to change the solvent evaporation rate of the corresponding radius annular region, and synchronously drives the spindle motor to respond to global speed correction commands to adjust the global centrifugal shear force. Through multi-physics field collaborative intervention, the non-uniform evolution trend on the corresponding radius is offset.

[0017] The closed-loop monitoring module performs data acquisition and evolution rate analysis steps in a loop while executing intervention, and updates the radial thickness deviation evolution rate map in real time until the overall thickness uniformity index converges to the preset process tolerance window.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) By constructing a dynamic model that characterizes the spatiotemporal changes of the photoresist flow state, the present invention can identify the thickening or thinning trend of different regions on the wafer surface online. This identification mechanism based on dynamic evolution rate rather than static deviation enables the control system to predict the development direction and intensity of non-uniformity, thereby providing a forward-looking decision basis for subsequent compensation intervention, realizing the direct suppression of the non-uniformity formation process, and improving the timeliness and initiative of the control response.

[0019] (2) This invention applies spatially differentiated physical interventions to independently and precisely adjust the local environment of specific regions on the wafer, such as the solvent evaporation rate that affects the viscosity of the photoresist. This mechanism can generate a physical compensation field that matches and has the opposite effect in spatial distribution based on the pre-identified non-uniformity development trend, thereby suppressing complex and asymmetric film thickness non-uniformity. This helps to avoid secondary non-uniformity problems that may be caused by global uniform control, and thus improves the uniformity of the final film thickness.

[0020] (3) This invention proposes a composite intervention strategy that combines local environmental control with macroscopic correction of the global centrifugal force field. When the compensation requirement of local control exceeds its physical execution limit, the system can convert this uncompensated deviation into a correction command for the global rotation speed and execute it synchronously. This dual control loop, which combines primary and secondary control, expands the dynamic adjustment range of the entire control system, ensuring that even when faced with large initial non-uniformity or strong disturbances, the system can still maintain stable control over the film thickness evolution trend through the coupling effect of multiple physical fields, thereby comprehensively improving the robustness and applicability of the process. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 is a schematic diagram of the implementation steps of the method of the present invention.

[0023] Figure 2 is a schematic diagram of the system module connection of the present invention.

[0024] Figure 3 is a schematic diagram of the logic control flow of the system modules of the present invention. Detailed Implementation

[0025] 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.

[0026] Example 1

[0027] Please refer to Figure 1. The present invention proposes a method to improve the uniformity of photoresist film thickness after coating by a coating machine, comprising: S1, controlling the spindle motor of the coating machine to drive the wafer to rotate according to a preset speed curve to establish a basic flow field, and simultaneously triggering a non-contact multi-point optical measurement system to acquire the original spectral reflection signal, and analyzing the original spectral reflection signal to generate photoresist transient thickness data.

[0028] Specifically, the main components executing this step are the process controller and integrated optical measurement system of the photoresist coater. The process controller sends a preset digital rotation speed command sequence to the servo driver of the spindle motor, driving the wafer stage carrying photoresist droplets to rotate according to the initial rotation speed curve. The initial rotation speed curve refers to a preset program of the angular velocity driving the wafer rotation over time, including at least one low-speed spreading stage and one high-speed planarization stage. In this method, the initial rotation speed curve includes a low-speed spin coating section accelerating from 10 rpm to 500 rpm, lasting from 0.5 s to 3 s, for the initial spreading of the photoresist, followed by a high-speed spin coating section jumping from 500 rpm to the range of 1500 rpm to 3000 rpm to establish a centrifugal-driven basic flow field. This basic flow field not only realizes the initial spreading and macroscopic shaping of the photoresist from a droplet state to a thin film covering the entire wafer, but more importantly, it constitutes the dynamic evolution starting point and reference state for all subsequent real-time monitoring and closed-loop intervention.

[0029] At the same moment the wafer begins to rotate, the process controller activates a non-contact multi-point optical measurement system located above the wafer via a digital trigger signal. This system, based on the principles of white light interferometry or spectral ellipsometrics, includes a multi-channel spectral light source and a radial array of 5 to 15 fiber optic probes, positioned, for example, at distances of 0 mm, 25 mm, 50 mm, 75 mm, and 95 mm from the wafer center. It synchronously captures the raw spectral signal reflected from each measurement point at a fixed sampling rate of 1 kHz. This signal represents the light intensity distribution with wavelengths from 400 nm to 800 nm and is converted into a digital matrix by a high-speed data acquisition card. The core function of the high-speed data acquisition card is to quickly and accurately convert analog signals from the physical world into digital signals that can be processed by a computer through an integrated high-speed analog-to-digital converter.

[0030] In a preferred embodiment, analyzing the original spectral reflection signal to generate photoresist transient thickness data includes: calling an analytical algorithm based on the principle of thin-film optical interference, and fitting the acquired original spectral reflection signal with a pre-stored photoresist and substrate material refractive index dispersion model for each measurement point at each sampling time.

[0031] The boundary conditions of Maxwell's equations on the interface of multilayer thin films are solved iteratively using the transfer matrix method, minimizing the residual between the measured spectrum and the simulated spectrum, and calculating the absolute film thickness at the current moment.

[0032] By associating the absolute film thickness value with the corresponding timestamp and location coordinates, structured transient thickness data of the photoresist is generated.

[0033] Specifically, the digital matrix is ​​fed into the real-time signal processing thread of the controller, invoking an analytical algorithm based on the principle of thin-film optical interference. For each measurement point at each sampling time, this algorithm first uses the transition matrix method to solve the boundary conditions of Maxwell's equations at the multilayer thin-film interface. Combined with a pre-stored refractive index dispersion model of the photoresist and substrate materials, and utilizing a pre-built high-dimensional lookup table or a trained neural network model, it establishes a relationship model between film thickness and reflection spectrum. Subsequently, a nonlinear least-squares optimization algorithm is used for rapid inversion, minimizing the residual between the original spectral reflection signal and the model-simulated spectrum to calculate the absolute film thickness of the photoresist physical layer at the current point. The inversion time should meet the real-time control cycle requirements, preferably less than 1 ms. For this purpose, the controller can be configured with a field-programmable gate array (FPGA) or a graphics processing unit (GPU) to provide the necessary parallel computing capabilities.

[0034] Each successful calculation generates a structured data record containing a precise millisecond timestamp from the start of spin coating, the radial position or radius value of the measurement point, and the calculated absolute film thickness value. This generates a photoresist transient thickness data sequence indexed by time and containing radial coordinates and thickness values ​​during continuous acquisition and analysis.

[0035] For example, suppose the coating machine's process controller retrieves a set of initial speed curves based on the formula, where the spindle speed is linearly accelerated from 0 rpm to 200 rpm between 0 ms and 200 ms. At that time, the rotation speed was 100 rpm. A fiber optic probe located 50 mm radially captured the reflected spectral signal at that point, which, after passing through a data acquisition card, yielded a light intensity array of 400 wavelength points. The controller's analytical algorithm called a pre-stored refractive index model specific to a particular photoresist type. , For the currently detected spectral data array, the simulated spectrum is calculated iteratively using the transfer matrix method. After 5 iterations, the root mean square error between the simulated and measured spectra reaches a minimum of 0.01, at which point the model calculates a film thickness of 1200 nm. The system then generates a transient thickness data record with a timestamp of 100 ms and location coordinates (…). , (Corresponding to a radial diameter of 50 mm and an angle of 45°), the film thickness is 1200.0 nm. Other measurement points located at the wafer center and edge are also synchronously acquired, analyzed, and have their own data records generated according to the above process, ultimately... A real-time data frame containing thickness information for all measurement points is generated at each moment.

[0036] S2. The instantaneous thickness deviation value is obtained by comparing the transient thickness data of the photoresist with the preset ideal thickness decay curve that varies with radial position. The instantaneous thickness deviation value is then differentiated over time and interpolated radially to generate a radial thickness deviation evolution rate map.

[0037] In a preferred embodiment, the instantaneous thickness deviation value is differentiated over time and interpolated radially to generate a radial thickness deviation evolution rate map, including: establishing a time window cache to store a sequence of instantaneous thickness deviation values ​​for multiple consecutive time steps.

[0038] The first derivative of the instantaneous thickness deviation at the current moment is calculated using the backward difference method to obtain the rate of change of the thickness deviation.

[0039] Using the wafer surface as a two-dimensional plane, the thickness deviation change rate calculated from discrete measurement points located at different radial measurement positions is used as a known sample point.

[0040] The radial spatial correlation of sample points is analyzed using a one-dimensional interpolation algorithm to estimate the rate of change at the unmeasured radius on the wafer surface.

[0041] Arrange the rate of change values ​​of all radius points according to their spatial location to form a two-dimensional radial distribution matrix, and generate a radial thickness deviation evolution rate map that characterizes the development trend of radial film thickness non-uniformity across the entire wafer.

[0042] Specifically, the main entity executing this step is the real-time controller of the photoresist coater. The controller first reads the transient thickness data of the photoresist from memory, which contains a sequence of data frames arranged in chronological order. For each measurement point in each data frame, the real-time control algorithm calls a pre-set ideal thickness decay curve function that varies with radial position in memory. This function defines the theoretical value of the film thickness over time at different radii of the wafer under an ideal uniform flow field. The ideal thickness decay curve can be obtained by pre-calibrating a standard wafer (a reference workpiece whose surface characteristics and geometric dimensions conform to preset specifications) under strictly controlled environmental temperature, humidity, and chamber pressure conditions. Spin coating is performed, and the film thickness data at each radial position over time is recorded. A high-fidelity reference curve is obtained using statistical methods such as polynomial fitting or nonlinear regression. Alternatively, it can be calculated using a simulation model incorporating fluid dynamics parameters. This simulation model should at least couple the solution of the Navier-Stokes equations describing fluid motion, the concentration-diffusion equation describing solvent evaporation, and the photoresist viscosity model related to concentration and temperature. It should be verified by comparison with a small amount of experimental data before being used to generate the ideal curve to ensure its physical authenticity. For different types or batches of photoresist, only the corresponding pre-stored curve needs to be called or a rapid automated calibration process needs to be executed once to generate a new curve to ensure the accuracy of the reference. Due to different rotational linear velocities, the ideal decay characteristics at different radii have physical differences. The timestamp of the current data is extracted, and the target thickness value of the ideal curve at that moment is queried.

[0043] Subsequently, the actual film thickness at the current measurement point is subtracted from the target thickness at the same time point to obtain the instantaneous thickness deviation value at that measurement point at that moment. This value can be positive, negative, or zero. A positive value indicates that the film thickness is higher than the ideal value, and a negative value indicates that it is lower than the ideal value.

[0044] Next, the controller opens a time window buffer to store the instantaneous thickness deviation value sequence for multiple consecutive time steps at each measurement point, in order to ensure the stability of the differential calculation. For example, the system buffers data from the most recent 100 time steps.

[0045] For each measurement point, the controller processes its deviation sequence using a numerical differentiation method. Considering the causality of data in a real-time system, this implementation uses backward difference calculation to determine the current time. The rate of change of thickness deviation, i.e., the first derivative, is given by the formula:

[0046] in, This represents the instantaneous thickness deviation value. This represents the sampling time interval. The unit is . This represents the instantaneous thickness deviation at the measurement point at the previous sampling time, obtained by comparing the transient thickness data of the photoresist with the ideal curve. The rate of change of thickness deviation physically reflects whether the trend of the measurement point deviating from the ideal state is intensifying or mitigating.

[0047] Finally, the controller uses the radial coordinates of all measurement points and their corresponding rates of change to execute a radial spatial interpolation algorithm. This algorithm treats the wafer surface as a two-dimensional plane and knows the rates of change at a finite number of radial measurement points. The values ​​are calculated using methods such as spline interpolation or linear interpolation to estimate the rate of change at rotationally symmetric grid points covering the entire wafer surface. These grid point estimates are then arranged spatially to generate a two-dimensional matrix. All circumferential points on the same radius in this matrix have the same rate of change value. The row and column indices correspond to their spatial locations on the wafer surface, and the matrix element values ​​correspond to the calculated thickness deviation rate at each location. This two-dimensional matrix is ​​the radial thickness deviation evolution rate map characterizing the trend of film thickness inhomogeneity across the entire wafer.

[0048] Radial spatial interpolation algorithms are used to estimate the continuous spatial distribution along the radial direction based on discrete point data of radial distribution. In this scheme, it is assumed that the film thickness variation is continuous and smooth in radial space. Cubic spline interpolation or piecewise linear interpolation is used. This method can give the best smooth estimate based on the radial correlation of the measurement points. Subsequently, the radial distribution is extended to a two-dimensional plane based on the assumption of rotational symmetry. The density of the interpolation grid is set according to the control accuracy requirements, typically 50 to 200 points in radial resolution.

[0049] For example, following the example of step S1, suppose in At that time, the actual film thickness at the measurement point 50mm from the center was 1200.0nm. The controller retrieved the ideal thickness decay curve for the current formulation from the process database and obtained the thickness at that point. The target film thickness at that point is 1150 nm. The system calculates the instantaneous thickness deviation at that point. The system reads the cached data for that measurement point, assuming it was in the past. At that time, its instantaneous thickness deviation value Sampling interval The time is 1ms. The backward difference method is used for calculation. Rate of change of time: This means that at that point and at that moment, the actual film thickness is deviating from the ideal value at a rate of 300 nanometers per minute. Obtain all measurement points on the wafer (such as the center point, 25mm, 50mm, 75mm, and 95mm points) at... rate of change at time After the values, the system uses these discrete values. The value and its corresponding radial coordinates are used as input to perform radial spatial interpolation. The interpolation process estimates any radius position on the wafer (e.g., R=22.36mm, corresponding coordinates...). , The rate of change at (location) ultimately generates a The grid data. This grid data is... The radial thickness deviation evolution rate at time t is plotted, which shows a distribution of a series of concentric rings.

[0050] S3. Based on the radial thickness deviation evolution rate map, identify the thickening trend region and thinning trend region at different radial positions, calculate the corresponding solvent evaporation rate adjustment target value, and construct a radially symmetric solvent evaporation rate compensation field.

[0051] In a preferred embodiment, constructing a radially symmetric solvent evaporation rate compensation field includes: traversing each grid point in the radial thickness deviation evolution rate map; if the rate of change is positive, it is determined to be a region of thickening trend; and calculating a first solvent evaporation rate suppression target value lower than the base evaporation rate based on the absolute value of the rate of change and a preset suppression coefficient.

[0052] If the rate of change is negative, it is determined to be a thinning trend region. Based on the absolute value of the rate of change and the preset promotion coefficient, a second solvent evaporation rate promotion target value higher than the basic evaporation rate is calculated.

[0053] The first solvent evaporation rate suppression target value and the second solvent evaporation rate promotion target value are integrated according to spatial distribution to form a solvent evaporation rate compensation field.

[0054] Specifically, the main entity executing this step is the controller of the coating machine and its built-in coating dynamics inverse solution model. The controller receives the radial thickness deviation evolution rate map generated in step S2. This map is a two-dimensional data matrix, and its element values ​​represent the rate of change of thickness deviation at the wafer grid points. The coating kinetics inverse solution model refers to a mathematical model that inversely deduces the required solvent evaporation rate distribution based on the observed film thickness variation trend. In this step, it is specifically manifested by traversing all elements of the matrix and performing region identification logic by comparing each element value with zero: if... If so, the grid point corresponding to that element is marked as a "thickening trend region," indicating that the film thickness at that location is increasing; if If it is marked as a "thinning trend area", it indicates that the film thickness at that location is becoming increasingly thinner; if If the value is within a certain range, it is marked as a stable region. For example, at a certain moment, the controller calculates a point at the center of the wafer. (Thickening trend), while at a certain point on the edge (Thinning trend)

[0055] Subsequently, for each marked grid point, the model invokes a pre-defined internal mapping function. This mapping function takes the absolute value of the rate of change at the current point as input and calculates the target value for adjusting the solvent evaporation rate by querying a pre-stored linear interpolation lookup table or applying a pre-defined linear transformation coefficient.

[0056] It should be noted that, for the thickening trend region, the target value for suppressing the first solvent evaporation rate is calculated. Its value is less than the basic volatility. The aim is to slow down the increase in photoresist viscosity and enhance its fluidity by reducing the local evaporation rate. The calculation formula is as follows: ,in This represents the inhibition coefficient. Calculation results. This indicates the expected evaporation rate, which is lower than the baseline value. The rate of change of thickness deviation at this grid point, read from the radial thickness deviation evolution rate map, in units of... . This represents the preset base solvent evaporation rate under stable and uniform flow conditions, in units of... Or equivalent evaporation flux units. Base solvent evaporation rate. This is a process preset constant, representing the average solvent evaporation rate required to achieve the target film thickness under ideal and uniform conditions. Its value depends on the photoresist formulation, ambient temperature, and atmospheric pressure. For example, for a certain type of ArF photoresist, under process conditions of 23°C, standard atmospheric pressure, and a target film thickness of 100nm, extensive experiments have shown that when the average solvent evaporation rate is maintained at a certain level... When the root mean square error is minimized, the overall uniformity can be obtained, and this value will be written into the formulation of the process.

[0057] For the thinning trend region, calculate the target value for the second solvent evaporation rate. Its value is greater than The aim is to accelerate the increase in photoresist viscosity by increasing the local evaporation rate, thereby hindering its excessive flow. The calculation formula is as follows: ,in The calculation results are for the promotion coefficient. This indicates the expected evaporation rate, which is higher than the baseline value.

[0058] Inhibition coefficient and promotion coefficient It is a proportional constant in the mapping relationship, which characterizes the strength of the influence of the thickness change trend on the required volatile matter adjustment amount; and The dimensions of the product make the product and and The units are consistent, and the specific values ​​are obtained by performing a series of spin-coating experiments on a specific photoresist under different evaporation rate conditions, and then fitting the relationship between film thickness uniformity and evaporation rate gradient. For example... The value ranges from 0.05 to 0.2. , The value ranges from 0.02 to 0.1. between.

[0059] For the stable region, the target value is directly set to .

[0060] After calculating the target values ​​for all grid points, the controller reassembles these target values ​​according to the spatial positions of the original grid points, forming a two-dimensional matrix with the same dimensions as the input map. This matrix defines the desired solvent evaporation rate at each point on the wafer surface, thus constructing a solvent evaporation rate compensation field that is spatially continuous on the wafer surface.

[0061] For example, following the example of step S2, in At that time, the rate of change of grid points located at 50 mm radially This area was identified as a region showing a thickening trend. (Controller preset) And read the suppression coefficient calibrated for the current photoresist from the parameter table. The system calculates the target value for suppressing the first solvent evaporation rate at this point: Meanwhile, assuming a grid point at 95mm from the wafer edge is identified as a thinning trend region, its System read promotion coefficient Calculate the second solvent evaporation rate at this point to promote the target value: For the wafer center For regions close to 0, the target value is directly set to 10.0. The controller follows this method, traversing each grid point of the radial thickness deviation evolution rate map (e.g., ...). (10,000 points), for each point, a target evaporation rate value is calculated. These 10,000 values ​​are then arranged in their original spatial order, ultimately generating a 100x100 matrix. This matrix represents the solvent evaporation rate compensation field at the current moment. This compensation field clearly indicates where the evaporation rate needs to be reduced to 9.5%. Where is it necessary to increase it to 10.1? And where to maintain 10.0 .

[0062] S4. Map the radially symmetric solvent evaporation rate compensation field to physical control parameters to generate local environmental control commands for the annular partitioned environmental control array, evaluate the residual error caused by the array execution saturation, and calculate the global speed correction command for the spindle motor based on the residual error.

[0063] In a preferred embodiment, mapping the radially symmetric solvent evaporation rate compensation field to physical control parameters to generate local environmental control commands for the annular partitioned environmental control array includes: mapping the target value in the radially symmetric solvent evaporation rate compensation field to the preliminary command value of each control unit based on the spatial position of each independent control unit in the annular partitioned environmental control array.

[0064] Spatial gradient smoothing is performed on the initial command values ​​of adjacent control units, and convolution operation is performed to eliminate parameter abrupt changes between adjacent regions, so as to obtain the final local environment control command.

[0065] In a further preferred embodiment, the residual error caused by array execution saturation is evaluated, and a global speed correction command for the spindle motor is calculated based on the residual error, including: comparing the local environmental control command of each control unit with the maximum and minimum values ​​of the physical output capability of that unit.

[0066] If the instruction value exceeds the physical output capability range, the unit will execute at the limit value and calculate the difference between the instruction value and the limit value as the single-point residual error.

[0067] The net residual error is obtained by summing the single-point residual errors of all control units. If the net residual error exceeds the preset threshold, the net residual error is converted into speed increments to generate a global speed correction command.

[0068] Specifically, the main entity executing this step is the controller of the coating machine. The controller first reads the solvent evaporation rate compensation field generated in step S3. The controller has a pre-stored physical parameter mapping database, which defines the target solvent evaporation rate value and the executable physical parameters (such as the power density of thermal radiation flux) of each independent control unit in the partitioned environmental control array. or the volume concentration of a specific solvent in a solvent atmosphere curtain. The correspondence between the physical parameter mapping database and the control unit should be defined. The physical parameter mapping database should contain parameters describing the dynamic response characteristics of each control unit, such as time constants, so that the controller can take into account the response lag of the physical actuators when generating instructions.

[0069] The partitioned environment control array refers to a physical device consisting of multiple independently addressable and controllable units arranged in a ring or array above the wafer. Each unit can respond to actuators controlled by digital or analog signals, such as miniature infrared heaters or miniature airflow nozzles, to change the microenvironment of the local area below it.

[0070] The controller traverses each grid point in the compensation field, determines which control unit it belongs to based on its radial distance (radius) from the wafer center, and the control unit is an annular region covering a specific radius. The controller then calls a mapping function to convert the target volatility value of the grid point into a preliminary local environmental control command value suitable for the control unit, such as a voltage setpoint representing the desired power level or a digital quantity representing the gas flow rate.

[0071] Next, the controller performs spatial gradient smoothing on the initial command values ​​of all adjacent control units. Spatial gradient smoothing is an image or signal processing method used to eliminate high-frequency spatial components in data and prevent drastic changes in adjacent control commands. In this step, it is specifically implemented using discrete convolution operations. The size and type of the filter kernel (mean, Gaussian) are selected based on the requirements for spatial resolution and smoothness of the control response. Specifically, the controller treats the initial command value as a one-dimensional discrete signal and applies a low-pass filter convolution kernel to perform a convolution operation on this signal. The convolution operation traverses all command points except for edges, replacing the original value with the weighted average of its own and the command values ​​of its surrounding neighboring points, thereby eliminating parameter abrupt changes between adjacent regions. The smoothed command value matrix is ​​the final local environment control command set.

[0072] Let the first The initial command value of each ring control unit is The final command value is obtained after spatial gradient smoothing. The smoothing process using a mean filter kernel can be represented as a convolution operation:

[0073] in, The set of indices representing the units radially adjacent to the i-th ring unit (including the i-th unit itself), i.e. To handle boundary cases located at the wafer center (index i=0) and wafer edge (index i=N), this implementation adopts a mirror-fill boundary condition. That is, for a central cell, the instruction value of its non-existent inner neighbor is considered to be the same as the instruction value of its outer neighbor; for an edge cell, the instruction value of its non-existent outer neighbor is considered to be the same as the instruction value of its inner neighbor, thereby ensuring that the smooth calculation of all cells can be completed within the effective neighborhood. This represents the initial command value for the j-th ring control unit. Indicates the range of the summation: j will take all elements of the set. Each element in it.

[0074] Subsequently, the controller evaluates the saturation of the partitioned environmental control array's execution of the solvent evaporation compensation field: it compares the final command value of each control unit with the maximum and minimum values ​​that the unit can physically output. If the command value of a unit exceeds its maximum capacity, the unit executes at its maximum value and records a positive residual error value, which is the command value minus the maximum value; if the command value is lower than its minimum capacity, it executes at its minimum value and records a negative residual error value.

[0075] Execution saturation refers to the degree to which the actual physical output capability of the partitioned environment control array satisfies the control command requirements; when the command value exceeds the physical limit of the unit, the unit is in a saturated state.

[0076] Residual error refers to the portion of compensation demand that cannot be executed due to the saturation of the control unit, and it is the direct input for calculating the global speed correction.

[0077] The controller sums the positive and negative residual errors of all units and calculates the net residual error. If the absolute value of the net residual error exceeds a preset threshold, for example, if the equivalent volatility error exceeds 0.05... Based on this net residual error, the controller synchronously calculates a global speed correction command for the spindle motor using a proportional-integral (PI) control algorithm implemented in a digital signal processor. This correction command is a speed increment. The unit is usually rpm, and its sign is used to offset the overall unevenness represented by the net residual error. Its calculation follows the PI control law: ,in, This is the global speed correction command calculated at time t, i.e., the speed increment; This is the net residual error calculated at time t; For time pointers, Indicates the net residual error The time integral from the start of the process to the current time t represents the historical cumulative error; The proportional gain determines the system's response speed to the current error; This is the integral term gain, used to eliminate steady-state errors caused by model mismatch or continuous perturbations, ensuring that residual inhomogeneities are ultimately and completely compensated.

[0078] For example, if the overall thickness is too large, the rotational speed is slightly increased to enhance the centrifugal shearing effect. The magnitude of this increase is dynamically determined by the proportional-integral control algorithm based on the instantaneous and historical cumulative values ​​of the net residual error. To prevent excessively large rotational speed adjustments from disrupting the stability of the basic flow field, the rotational speed increment... The calculation should be subject to a preset amplitude limit, such as the single adjustment amount not exceeding five percent of the current speed, and the adjustment direction should conform to physical laws, such as increasing the speed when it is necessary to reduce the overall film thickness, and decreasing the speed when it is necessary to reduce the overall film thickness.

[0079] In the saturation assessment, let the first... The physical output capability range of each unit is Then the residual error of the unit Defined as:

[0080] Net residual error The sum of residual errors of all elements: .

[0081] like , If the preset threshold is used, then the global speed correction command will be applied. The threshold is calculated using the PI control algorithm described above. This is used to determine whether the residual error is significant enough to require initiating global speed correction. Its setting is based on the process's tolerance for minor inhomogeneities, for example, equivalent to a film thickness variation rate of less than 0.01%. The error is negligible.

[0082] For example, following the example of step S3, the controller reads the solvent evaporation rate compensation field, where the target evaporation rate at a grid point located in a radial 50mm region is 9.5%. Through coordinate mapping, this point belongs to unit A of the control array. The controller queries the physical parameter mapping database and assigns 9.5... The target value is converted into a preliminary heater PWM duty cycle command, assumed to be 45%. The preliminary command for adjacent unit B of unit A is 48%, and for unit C it is 45%. To prevent these three unit command values ​​(45%, 48%, 45%) from stepping at the boundaries, the controller applies a preliminary command matrix to the entire region. Mean filtering. After filtering, the final command value of unit A becomes the average of its own and the commands of its radially adjacent units, which is assumed to be 46% after calculation. Next, the execution saturation is evaluated: assuming that the maximum output of unit A's heater corresponds to a duty cycle of 50% and a minimum of 10%, its final command of 46% falls within this range, therefore its residual error... Suppose a unit D in the edge region has a final instruction requirement of 55%, but its maximum capacity is only 50%. In this case, the unit will execute at 50% saturation, generating a positive residual error. The controller aggregates the positive and negative residual errors of all units and calculates the net residual error. (Equivalent volatility error). Preset threshold. It is 0.5%, because Therefore, a global speed correction needs to be calculated. A complete PI controller will calculate both the proportional and integral terms simultaneously. The controller parameters are set as follows: proportional coefficient. Integral coefficient And assume that the integral value of the error accumulated from the beginning of the process up to the current time. for This value represents a small positive error that has persisted in the system for a period of time and has not been completely eliminated. Therefore, the calculation of the speed correction command Δω is divided into two parts: the proportional part, which responds to the current error. The integral part compensates for accumulated historical errors. Final speed correction command This means that while maintaining the output of each unit according to the final instruction (such as 46% for unit A and 50% for unit D), the controller will send an instruction to the spindle motor, requiring it to temporarily reduce its speed by 11 rpm from the current speed, so as to more forcefully drive the system to eliminate steady-state deviation.

[0083] S5 drives the annular partitioned environmental control array to respond to local environmental control commands to change the solvent evaporation rate of the corresponding radius annular region, and synchronously drives the spindle motor to respond to global speed correction commands to adjust the global centrifugal shear force. Through multi-physics field collaborative intervention, the non-uniform evolution trend on the corresponding radius is offset.

[0084] In a preferred embodiment, driving the annular partitioned environmental control array to respond to local environmental control commands to change the solvent evaporation rate of the corresponding radius annular region includes: sending digital control signals to control units located above the wafer in concentric rings, driving micro infrared heaters to apply thermal radiation flux of intensity corresponding to the local environmental control commands to the corresponding region of the wafer, thereby increasing the solvent evaporation rate by raising the local temperature.

[0085] Alternatively, a gas flow controller can be driven to spray a solvent atmosphere curtain of a concentration corresponding to the local environmental control command into the corresponding area of ​​the wafer, thereby adjusting the solvent evaporation rate by changing the local gas phase solvent partial pressure.

[0086] In a further preferred embodiment, the non-uniform evolution trend on the corresponding radius is offset by multi-physics field synergistic intervention, including: using a partitioned environmental control array to differentially control the solvent evaporation rate of the photoresist in a local area, thereby achieving micro-adjustment of the local viscosity of the photoresist to impede or promote local flow.

[0087] By adjusting the wafer rotation angular velocity using a spindle motor, the macroscopic centrifugal shear force acting on the photoresist fluid across the entire field can be globally adjusted.

[0088] Through the coupling effect of local viscosity micro-control and macro-centrifugal shear force global adjustment, a net flow effect is generated that is opposite to the non-uniform trend in the radial thickness deviation evolution rate spectrum.

[0089] Specifically, the main components performing this step are the controller of the glue applicator, the drive circuit of the zoned environmental control array, and the servo driver of the spindle motor. The drive circuit of the zoned environmental control array refers to the electronic hardware necessary to convert digital control commands into physical actuator actions, including digital-to-analog converters, power amplifiers, solid-state relays, and protection circuits.

[0090] The controller simultaneously sends commands to two independent execution channels according to a preset control cycle. In the first channel, the controller sends the local environmental control command set to each independent control unit of the partitioned environmental control array via a digital communication bus or parallel I / O port. Considering the potential physical response lag in thermal or gas control, the controller can employ a Smith predictor compensation strategy or adjust the integral time constant in the PID parameters to match the dynamic characteristics of the actuator. Each control unit generates a corresponding analog control signal based on the received digital command value through its internal or external digital-to-analog converter circuit.

[0091] For thermal radiation control units, the analog signal (typically 0-10V DC voltage or 4-20mA current) drives a solid-state relay or linear power amplifier, thereby precisely controlling the current of the miniature infrared heater. This allows it to apply a thermal radiation flux to the corresponding annular region of the wafer directly below, corresponding to the intensity of the local environmental control command. The unit is typically 1000 kJ / m². Thermal radiation flux refers to the heat energy radiated by the heater per unit area per unit time. Its intensity is controlled by adjusting the electrical power of the heating element. The applied heat flux must be sufficient to change the surface temperature of the photoresist to affect the solvent evaporation rate, but at the same time, excessively high temperatures must be avoided to prevent thermal decomposition or cross-linking reactions of the photoresist.

[0092] For the solvent atmosphere curtain-type control unit, the analog signal controls a high-precision mass flow controller and a solvent vapor generator to regulate the mixed airflow of solvent vapor and inert carrier gas corresponding to the concentration of the local environment control command, forming a vertical or inclined curtain covering the target annular region at a specific flow rate and direction. By changing the partial pressure of solvent vapor in the curtain, the solvent saturation level in the gas phase above the annular region on the wafer surface can be adjusted, thereby suppressing or promoting solvent evaporation. The application of thermal radiation or solvent atmosphere, with its spatial distribution and intensity strictly corresponding to the local environment control command, aims to change the solvent evaporation rate of the photoresist in the local area. The entire application process is controlled within the intensity range verified by hydrodynamic simulation to ensure that no severe airflow disturbances or temperature gradients that would disrupt the stability of the basic flow field are introduced.

[0093] The intensity range verified by fluid dynamics simulation refers to the upper limit of the control intensity determined by simulating a specific chamber structure and spin coating parameters using computational fluid dynamics software, without inducing flow instability (such as the generation of eddies or Taylor vortices); for example, the heat flux generally does not exceed 5 The airflow velocity generally does not exceed the value corresponding to the laminar critical Reynolds number.

[0094] In the second channel, the controller synchronously sends a global speed correction command to the servo driver of the spindle motor via a high-speed analog output module or a dedicated motion control bus. This command is an instantaneous speed increment superimposed on the original speed curve. The servo driver responds to this command by rapidly adjusting the current output to the spindle motor armature through its internal current loop, thereby fine-tuning the wafer's rotational angular velocity. This fine-tuning alters the macroscopic centrifugal shear force acting on the photoresist fluid. Macroscopic centrifugal shear force refers to the shear stress formed by the combined effect of the centrifugal acceleration generated by the wafer's rotation and the viscosity of the photoresist fluid itself; it is the main force field driving the radial spreading and thinning of the photoresist. Fine-tuning the rotational speed directly changes the intensity of this force field.

[0095] Ultimately, through the precise temporal synchronization and coupling of the differentiated spatial distribution of solvent evaporation rate in the local annular region (achieved by the first channel) and the overall adjustment of global centrifugal shear force (achieved by the second channel), the system dynamically counteracts and offsets the non-uniform evolution trend characterized by the radial thickness deviation evolution rate spectrum. The coupling effect refers to the physical process in which local viscosity changes (caused by evaporation rate regulation) and global shear force changes jointly affect the photoresist flow state; its synergistic goal is to generate a net flow effect opposite to the natural non-uniform trend.

[0096] For example, following the example of step S4, in During the control cycle, the controller simultaneously executes two operations. First, it sends a data frame via the CAN bus to the partitioned environment control array, containing the final command value of unit A (46% duty cycle) and the saturation execution value of unit D (50% duty cycle). Upon receiving the digital value 46, the drive circuit of unit A outputs a 4.6V analog voltage from its built-in DAC to its power amplifier, driving its miniature infrared heater to operate at 46% of its maximum rated power, assuming a power output of 2.3. The thermal radiation flux irradiates a radial 50mm area on the wafer surface. The heater in unit D operates at 50% power (2.5). Simultaneously, other curtain-type units located at the wafer edge also output nitrogen curtains of solvent vapor at specific concentrations (e.g., 200 ppm) according to their command values. All these local interventions aim to bring the photoresist solvent evaporation rate of the corresponding annular region closer to the target value calculated in step S3. Next, the controller sends a voltage step signal of -11 rpm to the spindle servo drive through the analog output module. The servo drive detects this command and, under its internal speed loop control, reduces the actual speed of the spindle motor from the current value (e.g., 1500 rpm) to 1489 rpm within milliseconds. This slight increase in speed slightly reduces the centrifugal force acting on the entire photoresist film. Initially, localized heat / gas regulation alters the rheological properties (viscosity) of a specific region, while global rotational speed fine-tuning changes the overall shear field. These two interventions work together to suppress the edge thickening trend detected in the example of step S2. The thinning trend in the central region and the surrounding area leads to a more uniform film thickness evolution across the entire wafer surface. The system continuously monitors the process in step S6 to verify the actual effect of this coupled intervention.

[0097] S6. While performing the intervention, the data acquisition and evolution rate analysis steps are executed in a loop, and the radial thickness deviation evolution rate map is updated in real time until the overall thickness uniformity index converges to the preset process tolerance window.

[0098] In a preferred embodiment, the radial thickness deviation evolution rate map is updated in real time until the overall thickness uniformity index converges to a preset process tolerance window, including: calculating the sample standard deviation of film thickness values ​​at all measurement points based on the latest acquired photoresist transient thickness data.

[0099] Compare the sample standard deviation with the upper limit of the preset process tolerance window.

[0100] If the sample standard deviation is less than or equal to the upper limit, and the current average film thickness is within the preset target film thickness range, or the current process time reaches the preset spin coating time endpoint, then it is determined to be converged, intervention is stopped and the process proceeds to the next step; otherwise, it is determined to be non-converged, and the adjustment instructions continue based on the updated spectrum.

[0101] Specifically, the main entity executing this step is the controller of the glue applicator. While continuously executing the multi-physics collaborative intervention in step S5, this step also cyclically runs a monitoring and decision-making thread with a fixed control cycle. The monitoring and decision-making thread refers to a real-time task within the controller with a fixed execution frequency, the cycle of which must be less than the main time constant of the glue applicator process, typically 5ms to 50ms.

[0102] Within each control cycle, the controller first commands the non-contact multi-point optical measurement system to synchronously acquire the raw spectral reflection signal of the new frame, and then generates the photoresist transient thickness data at the current moment in real time according to the analytical algorithm in step S1.

[0103] Subsequently, the controller immediately invokes the algorithm flow of step S2, using the newly generated transient thickness data as input, to calculate and update the radial thickness deviation evolution rate map.

[0104] Next, the controller calculates a global thickness uniformity index based on the latest transient thickness data of the photoresist, such as calculating the sample standard deviation of the thickness values ​​at all valid measurement points. Alternatively, calculate the difference between the maximum and minimum thickness values. The overall thickness uniformity index is a statistic used to quantify the dispersion of film thickness on the wafer surface. This step uses the sample standard deviation as an example, and its calculation covers all real-time measurement points. The controller compares the overall thickness uniformity index with a process tolerance window in a pre-existing process recipe, which defines the upper limit of the uniformity index. or The process tolerance window is the allowable deviation range set according to the specific requirements of subsequent photolithography processes for film thickness uniformity. For example, for a process that requires film thickness uniformity better than ±1.5%, if the target film thickness is 100nm, then... The setting can be from 1.0nm to 1.5nm. Simultaneously, the controller reads the system clock in real time to obtain the accumulated process time since the spin coating began. .

[0105] Assuming the system collects data within a control cycle... The transient thickness of the photoresist at each measurement point constitutes a dataset. The sample standard deviation was used as the index for overall thickness uniformity. The calculation is performed using the following formula:

[0106] in, Indicates the first The current film thickness value at each measurement point, in nm. Indicates all currently The arithmetic mean of the film thickness at each measurement point, i.e. . The unit is nm, and a smaller value indicates better film thickness uniformity. Simultaneously, all... Arithmetic mean of film thickness at each measurement point And determine whether it satisfies: ,in For the target film thickness, The allowable average film thickness deviation. The convergence criterion is... And the average film thickness meets the above conditions. The upper limit of the allowable standard deviation defined for the process tolerance window.

[0107] The controller performs a logical judgment: if the current uniformity index is less than or equal to the tolerance limit and the average film thickness meets the standard, and the current process time is... Less than the preset spin coating time endpoint Or if the current process time Reached or exceeded If the process meets the convergence or completion criteria, then the process is considered to have achieved the desired result. The preset spin coating time endpoint... The total process time is determined based on the evaporation and leveling characteristics of the photoresist solvent, typically ranging from 20 to 60 seconds. Upon reaching this time, the process will be forcibly terminated regardless of whether the uniformity index has fully converged, to prevent over-drying or defects. If the time reaches... If only some of the conditions in the convergence criterion are met, the system can record the abnormal state and issue an alarm, but it will still force the spin coating to end in order to proceed to the next process.

[0108] If the conditions are met, the controller immediately sends a stop intervention command to the drive circuit of the partitioned environment control array and the servo driver of the spindle motor, terminating the collaborative intervention in step S5, and triggering the glue applicator to enter the subsequent curing or cooling process.

[0109] If the conditions are not met, the controller will use the latest updated radial thickness deviation evolution rate map as input and re-execute steps S3 to S5 sequentially. Based on the new map, it will calculate a new solvent evaporation rate compensation field, generate new local environmental control commands and global speed correction commands, and drive the actuator to perform a new round of intervention, thus forming a dynamic closed-loop control loop until the process converges or reaches the end of the time. The dynamic closed-loop control loop refers to a complete control loop consisting of data acquisition, state analysis, decision calculation, command execution, and effect feedback. Its purpose is to make the controlled variable (film thickness uniformity) automatically approach and stabilize within the target range.

[0110] For example, following the example of step S5, suppose in After the intervention is implemented, the system enters a cyclical monitoring phase. In the next control cycle... The controller first acquires new transient thickness data, assuming the film thickness at the 50mm measurement point now becomes 1195nm. The system immediately executes step S2 based on the new data to calculate the new rate of change. This indicates that the thickening trend has slowed down, and the radial thickness deviation evolution rate map has been updated. Next, the controller calculates the film thickness values ​​at all five current measurement points (0, 25, 50, 75, 95 mm), assumed to be 1150 nm, 1160 nm, 1195 nm, 1180 nm, and 1200 nm, respectively. The average value is then calculated. Sample standard deviation Preset process tolerance It is 3.0nm, the current process time seconds, far from reaching Seconds. Due to ,and The convergence was deemed non-convergent. The controller then... Using the newly generated radial thickness deviation evolution rate map as input, the process in step S3 is restarted to calculate the new solvent evaporation rate compensation field, generate a new round of local environmental control commands and global rotation speed correction commands, and then... The drive array and motors execute in real time. This cycle continues. Assuming a process time... At that time, the system collected and calculated the standard deviation of the film thickness. It is already smaller than ,at the same time At this point, convergence is determined. The controller immediately sends a stop command, shuts off the heat source and air curtain of the partition control array, and smoothly reduces the spindle speed to 0 rpm, then triggers the glue applicator to enter the subsequent soft drying process.

[0111] Example 2

[0112] Please refer to Figures 2 and 3. Based on Embodiment 1, the second aspect of the present invention provides a system for improving the uniformity of photoresist film thickness after coating by a coating machine, comprising: a data acquisition module, a pattern construction module, a compensation field calculation module, an instruction generation module, a collaborative intervention module, and a closed-loop monitoring module.

[0113] The data acquisition module controls the spindle motor of the coating machine to drive the wafer to rotate according to a preset speed curve to establish the basic flow field. At the same time, it triggers a non-contact multi-point optical measurement system to acquire the original spectral reflection signal and analyzes the original spectral reflection signal to generate photoresist transient thickness data.

[0114] The map construction module compares the transient thickness data of the photoresist with the preset ideal thickness decay curve that varies with radial position to obtain the instantaneous thickness deviation value. It then performs time differentiation and radial interpolation on the instantaneous thickness deviation value to generate a radial thickness deviation evolution rate map.

[0115] The compensation field calculation module identifies thickening and thinning trend regions at different radial positions based on the radial thickness deviation evolution rate map, calculates the corresponding solvent evaporation rate adjustment target value, and constructs a radially symmetric solvent evaporation rate compensation field.

[0116] The instruction generation module maps the radially symmetric solvent evaporation rate compensation field to physical control parameters to generate local environmental control instructions for the annular partitioned environmental control array, evaluates the residual error caused by the array execution saturation, and calculates the global speed correction instruction for the spindle motor based on the residual error.

[0117] The collaborative intervention module drives the annular partitioned environmental control array to respond to local environmental control commands to change the solvent evaporation rate of the corresponding radius annular region, and synchronously drives the spindle motor to respond to global speed correction commands to adjust the global centrifugal shear force. Through multi-physics field collaborative intervention, the non-uniform evolution trend on the corresponding radius is offset.

[0118] The closed-loop monitoring module performs data acquisition and evolution rate analysis steps in a loop while executing intervention, and updates the radial thickness deviation evolution rate map in real time until the overall thickness uniformity index converges to the preset process tolerance window.

[0119] It should be noted that the formulas described above, through the principle of dimensional consistency and mathematical standardization methods (such as normalization, dimensionless parameter conversion, or unit system unification), can translate physical quantities with different properties into unitless standard values ​​or superimposed parameters of the same dimension. This eliminates the interference of different dimensions on the computational logic, allowing the formulas to retain the original data distribution characteristics while possessing mathematical rationality and adaptability to objective laws. The descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the invention.

Claims

1. A method for improving the uniformity of photoresist film thickness after coating by a coating machine, characterized in that, include: S1. Control the spindle motor of the coating machine to drive the wafer to rotate according to the preset speed curve to establish the basic flow field. At the same time, trigger the non-contact multi-point optical measurement system to acquire the original spectral reflection signal and analyze the original spectral reflection signal to generate photoresist transient thickness data. S2. Compare the transient thickness data of the photoresist with the preset ideal thickness decay curve that varies with radial position to obtain the instantaneous thickness deviation value. Perform time differentiation and radial interpolation on the instantaneous thickness deviation value to generate a radial thickness deviation evolution rate map. S3. Based on the radial thickness deviation evolution rate map, identify the thickening trend region and thinning trend region at different radial positions, calculate the corresponding solvent evaporation rate adjustment target value, and construct a radially symmetric solvent evaporation rate compensation field. S4. Map the radially symmetric solvent evaporation rate compensation field to physical control parameters to generate local environmental control commands for the annular partitioned environmental control array, and evaluate... S5. Estimate the residual error caused by the array's saturation, and calculate the global speed correction command for the spindle motor based on the residual error; S6. Drive the annular partitioned environmental control array to respond to the local environmental control command to change the solvent evaporation rate of the corresponding radius annular region, and simultaneously drive the spindle motor to respond to the global speed correction command to adjust the global centrifugal shear force, and offset the non-uniform evolution trend on the corresponding radius through multi-physics field collaborative intervention; S7. While performing the intervention, cyclically execute the data acquisition and evolution rate analysis steps, and update the radial thickness deviation evolution rate map in real time until the overall thickness uniformity index converges to within the preset process tolerance window.

2. The method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 1, characterized in that, The process of analyzing the raw spectral reflection signal to generate transient photoresist thickness data includes: calling an analytical algorithm based on the principle of thin-film optical interference to fit the acquired raw spectral reflection signal with a pre-stored refractive index dispersion model of the photoresist and substrate materials for each measurement point at each sampling time; using the transfer matrix method to iteratively solve the boundary conditions of Maxwell's equations at the interface of the multilayer thin film, minimizing the residual between the measured spectrum and the simulated spectrum, and calculating the absolute film thickness value at the current time; and associating the absolute film thickness value with the corresponding timestamp and position coordinates to generate structured transient photoresist thickness data.

3. The method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 1, characterized in that, The instantaneous thickness deviation value is differentiated over time and interpolated radially to generate a radial thickness deviation evolution rate map. This process includes: establishing a time window buffer to store a sequence of instantaneous thickness deviation values ​​for multiple consecutive time steps; calculating the first derivative of the instantaneous thickness deviation value at the current moment using the backward difference method to obtain the thickness deviation change rate; using the wafer surface as a two-dimensional plane, the thickness deviation change rate calculated at each discrete measurement point located at different radial measurement positions is used as known sample points; the radial spatial correlation of the sample points is analyzed using a one-dimensional interpolation algorithm to estimate the change rate value at the unmeasured radius of the wafer surface; and the change rate values ​​of all radius points are arranged according to their spatial positions to form a two-dimensional radial distribution matrix, generating a radial thickness deviation evolution rate map that characterizes the development trend of radial film thickness inhomogeneity across the entire wafer.

4. The method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 3, characterized in that, Constructing a radially symmetric solvent evaporation rate compensation field includes: traversing each grid point in the radial thickness deviation evolution rate map; if the rate of change is positive, it is determined to be a thickening trend region; and calculating a first solvent evaporation rate suppression target value lower than the base evaporation rate based on the absolute value of the rate of change and a preset suppression coefficient; if the rate of change is negative, it is determined to be a thinning trend region; and calculating a second solvent evaporation rate promotion target value higher than the base evaporation rate based on the absolute value of the rate of change and a preset promotion coefficient; and integrating the first solvent evaporation rate suppression target value and the second solvent evaporation rate promotion target value according to spatial distribution to form a solvent evaporation rate compensation field.

5. The method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 1, characterized in that, Mapping a radially symmetric solvent evaporation compensation field to physical control parameters to generate local environmental control commands for a ring-shaped partitioned environmental control array includes: mapping the target value in the radially symmetric solvent evaporation compensation field to the preliminary command value of each control unit based on the spatial position of each independent control unit in the ring-shaped partitioned environmental control array; performing spatial gradient smoothing on the preliminary command values ​​of adjacent control units and performing convolution operations to eliminate parameter abrupt changes between adjacent regions to obtain the final local environmental control command.

6. The method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 1, characterized in that, The residual error generated by the array execution saturation is evaluated, and a global speed correction command for the spindle motor is calculated based on the residual error. This includes: comparing the local environment control command of each control unit with the maximum and minimum values ​​of the physical output capability of that unit; if the command value exceeds the range of the physical output capability, the unit executes at the limit value and calculates the difference between the command value and the limit value as the single-point residual error; summing up the single-point residual errors of all control units to obtain the net residual error; if the net residual error exceeds a preset threshold, the net residual error is converted into a speed increment to generate a global speed correction command.

7. The method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 1, characterized in that, The method of driving the annular partitioned environmental control array to respond to local environmental control commands to change the solvent evaporation rate of the corresponding annular region includes: sending digital control signals to the control units located above the wafer in concentric rings to drive the micro infrared heater to apply a thermal radiation flux of intensity corresponding to the local environmental control command to the corresponding region of the wafer, thereby increasing the solvent evaporation rate by raising the local temperature; or driving the gas flow controller to spray a solvent atmosphere curtain of concentration corresponding to the local environmental control command to the corresponding region of the wafer, thereby adjusting the solvent evaporation rate by changing the local gas phase solvent partial pressure.

8. The method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 1, characterized in that, The non-uniform evolution trend on the corresponding radius is offset by multi-physics field synergistic intervention, including: using a partitioned environmental control array to differentially control the solvent evaporation rate of the photoresist in local areas, thereby achieving micro-adjustment of the local viscosity of the photoresist to impede or promote local flow; using a spindle motor to adjust the wafer rotation angular velocity, thereby achieving global adjustment of the macro-centrifugal shear force acting on the photoresist fluid throughout the field; and through the coupling effect of micro-adjustment of local viscosity and global adjustment of macro-centrifugal shear force, a net flow effect opposite to the non-uniform trend in the radial thickness deviation evolution rate spectrum is generated.

9. A method for improving the uniformity of photoresist film thickness after coating by a coating machine according to claim 1, characterized in that, The radial thickness deviation evolution rate map is updated in real time until the overall thickness uniformity index converges to the preset process tolerance window. This includes: calculating the sample standard deviation of the film thickness value at all measurement points based on the latest acquired photoresist transient thickness data; comparing the sample standard deviation with the upper limit of the preset process tolerance window; if the sample standard deviation is less than or equal to the upper limit, and the current overall average film thickness value is within the preset target film thickness range, or the current process time reaches the preset spin coating time endpoint, then convergence is determined, intervention is stopped, and the process proceeds to the next step; otherwise, it is determined as non-convergence, and the adjustment instructions continue based on the updated map.

10. A system for improving the uniformity of photoresist film thickness after coating by a coating machine, characterized in that, include: The data acquisition module controls the spindle motor of the coating machine to drive the wafer to rotate according to a preset speed curve to establish a basic flow field. At the same time, it triggers a non-contact multi-point optical measurement system to acquire the original spectral reflection signal and analyzes the original spectral reflection signal to generate photoresist transient thickness data. The image construction module compares the transient thickness data of the photoresist with a preset ideal thickness decay curve that varies with radial position to obtain the instantaneous thickness deviation value. It then performs time differentiation and radial interpolation on the instantaneous thickness deviation value to generate a radial thickness deviation evolution rate image. The compensation field calculation module, based on the radial thickness deviation evolution rate image, identifies thickening and thinning trend regions at different radial positions, calculates the corresponding solvent evaporation rate adjustment target values, and constructs a radially symmetric solvent evaporation rate compensation field. The instruction generation module maps the radially symmetric solvent evaporation rate compensation field to physical control parameters to generate local environmental control instructions for the annular partitioned environmental control array. The module assesses the residual error caused by array saturation and calculates a global speed correction command for the spindle motor based on the residual error. The collaborative intervention module drives the annular partitioned environmental control array to respond to local environmental control commands to change the solvent evaporation rate of the corresponding radius annular region, and simultaneously drives the spindle motor to respond to the global speed correction command to adjust the global centrifugal shear force. The non-uniform evolution trend on the corresponding radius is offset by multi-physics field collaborative intervention. The closed-loop monitoring module performs data acquisition and evolution rate analysis steps in a loop while performing intervention, and updates the radial thickness deviation evolution rate map in real time until the overall thickness uniformity index converges to the preset process tolerance window.

Images