A laser interferometric lithography fringe phase-locked control method and system based on PID and ILC composite control
By using a PID+ILC composite control method, the phase error in laser interference lithography is decoupled, thereby improving the phase-locked loop accuracy and fringe contrast without changing the stability of the feedback loop. This solves the problem of phase drift in interference fringes and meets the quality requirements for grating fabrication.
Patent Information
- Application Number
- CN202611576103.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-12
- Publication Date
- 2026-11-17
AI Technical Summary
In laser interference lithography, changes in ambient temperature, air disturbances, and mechanical vibrations cause phase shifts in the interference fringes. Traditional PID feedback control struggles to simultaneously suppress periodic ripples and maintain stability margins, making predictive compensation impossible and affecting the quality of the grating mask.
The PID+ILC composite control method is adopted. By decoupling the phase error into repetitive periodic disturbances and non-repetitive random disturbances, the repetitive disturbances are compensated by ILC feedforward between exposure batches, and random disturbances are suppressed in real time by PID feedback to ensure the stability of the feedback loop remains unchanged.
It significantly improves phase-locked loop accuracy, reduces phase error RMS by about 20%, and improves stripe contrast by about 1.5%, meeting the requirements for grating fabrication, reducing control bandwidth requirements, and improving lithography quality.
Smart Images

Figure CN123113261A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ultra-precision micro-nano manufacturing technology, and in particular to a laser interference lithography stripe phase-locked control method and system based on PID and ILC composite control. Background Technology
[0002] Laser interferometry (LIL) utilizes the superposition of coherent light fields to form periodic interference fringes. It can fabricate periodic micro / nano structures such as holographic gratings on a large area in a single exposure, offering advantages such as simple structure, high efficiency, and controllable cost, making it a key technology for high-performance optical devices and precision manufacturing equipment. However, the area of a single exposure is limited by the spot size and light field area, and the fabrication of ultra-large area gratings typically employs a step-repeating interferometry exposure technique. During exposure, changes in ambient temperature, air disturbances, and mechanical vibrations can cause phase shifts in the interference fringes, severely reducing exposure contrast and grating mask quality. Therefore, phase-locked loop control of the interference fringes is a crucial step in ensuring processing quality.
[0003] In practical systems, phase errors arise from the coupling of multiple physical disturbances: repetitive periodic disturbances (mains frequency coupling, mechanical resonance, measurement nonlinearity) coexist with non-repetitive random disturbances (air turbulence, electronic noise, ground vibration). Constrained by the system's inherent delay and Bode's integral theorem, simply increasing the PID gain is insufficient to achieve zero steady-state error suppression of periodic ripples; instead, it easily reduces phase margin and induces high-frequency oscillations. Furthermore, traditional single-loop feedback control can only passively respond after an error occurs, unable to utilize the recurrence patterns between exposure batches for predictive compensation, and also difficult to further improve phase-locked loop accuracy without sacrificing feedback stability. Therefore, a composite phase-locked loop control method is needed that can sense the time-frequency structure of disturbances, utilize batch repeatability, and does not alter the stability of the feedback loop. Summary of the Invention
[0004] This invention discloses a phase-locked loop (PLL) control method for laser interferometric lithography fringes based on PID+ILC composite control. Addressing the issues of multi-source heterogeneous disturbances causing fringe phase drift in step-repetitive laser interferometric lithography, and the difficulty of traditional PID feedback in simultaneously suppressing periodic ripple and maintaining stability margin, this method decouples the phase error into repetitive periodic disturbances and non-repetitive random disturbances based on the time-frequency characteristics of the disturbance, establishing an "integral + delay" controlled object model. The bottom layer uses high-bandwidth PI feedback to suppress random disturbances in real time, while the upper layer uses an additional ILC feedforward to learn and compensate for repetitive disturbances between exposure batches. The ILC feedforward does not enter the feedback loop and does not change the closed-loop characteristic equation, thus continuously improving the PLL accuracy without sacrificing feedback stability.
[0005] The specific technical solution for achieving the objective of this invention is as follows: 1. A phase-locked control method for laser interferometric lithography fringes based on PID+ILC composite control, characterized by comprising the following steps: Step 1: System Modeling and Disturbance Decomposition. Establish the "integral + delay" controlled object model of the frequency shift execution chain; based on the time-frequency domain characteristics, decouple the phase error of the interference fringes into a linear superposition of repetitive periodic disturbances and non-repetitive random disturbances; Step 2: PID Feedback Controller Design. Based on the controlled object model, a Type II PI feedback controller is designed to dynamically suppress non-repetitive random disturbances in real time, and the recursive control law is obtained by discretization with a preset servo period. Step 3: ILC Feedforward Controller Design. A P-type iterative learning control learning law is designed and a zero-phase-shift low-pass Q-filter is introduced. The ILC feedforward signal is injected into the output of the PID feedback controller in an additional structure to learn and compensate for repetitive periodic disturbances between exposure batches without changing the characteristic equation of the feedback loop. Step 4: Online Phase Locking and Offline Learning Execution. In the online phase, the PID feedback controller performs real-time phase locking and superimposes the ILC feedforward signal within each servo cycle. In the offline phase, during exposure intervals, a batch of phase error trajectories are collected, ILC learning updates are performed, and the data is written to the working area for use in the next exposure.
[0006] The beneficial effects of this invention are: (1) ILC feedforward is injected into the PID output terminal with an additional structure. It does not enter the feedback loop and does not change the characteristic equation of the closed loop system. The stability margin of the feedback loop is not affected by the iterative update of ILC. There is no need to re-verify the stability boundary of the feedback loop due to the introduction of ILC. (2) Based on the dual-channel division of labor mechanism of "ILC learns repetitive disturbances and PID resists random disturbances", ILC learns repetitive disturbances within the frequency band limited by Q-filter, and PID suppresses random disturbances in real time across the entire frequency band, thus coordinating to cover disturbances across the entire frequency band. (3) By utilizing the time integral smoothing characteristics of step-repeated exposure, the control bandwidth requirement is greatly reduced, providing feasibility for PC host computer digital control; (4) Simulation results show that the phase error RMS decreased from 0.2850 rad to 0.2261 rad within 40 iterations, a decrease of about 20%. The first 10 iterations contributed about 62% of the decrease. The fringe contrast retention rate increased from 96.0% to 97.5%, both of which meet the requirements of interference grating fabrication for linewidth uniformity and diffraction efficiency. Attached Figure Description
[0007] Figure 1 This is a schematic diagram of the architecture of the step-repetitive interference fringe phase-locked control system in an embodiment of the present invention; Figure 2This is a block diagram of the PID+ILC composite control architecture in an embodiment of the present invention; Figure 3 This is a flowchart of the PID+ILC composite control algorithm in an embodiment of the present invention; Figure 4 This is a time-domain simulation waveform of the phase error of a pure PID phase-locked loop in an embodiment of the present invention; Figure 5 This is an example of the ILC iterative convergence curve, the error comparison between the 1st and 40th batches, and the evolution diagram of the ILC feedforward signal in this embodiment of the invention. Detailed Implementation
[0008] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so as to more clearly illustrate the advantages and features of the present invention and enable those skilled in the art to more easily understand the essence of the present invention. The description of specific embodiments is intended to further clarify the scope of protection of the present invention and provide a basis for defining the claims.
[0009] refer to Figures 1-5 A laser interferometric lithography fringe phase-locked control method based on PID+ILC composite control includes the following steps: Step 1: System modeling and disturbance decomposition.
[0010] (1) System architecture. For example... Figure 1 As shown, the system consists of four parts: an optical interferometer platform, a phase detection module, a control module, and an actuator module. The ultraviolet laser beam is split into two paths by a beam splitter after passing through a fixed-frequency acousto-optic modulator AOM3 (frequency shift +110MHz): one path passes through a fixed-frequency acousto-optic modulator AOM1 as a reference path; the other path passes through a frequency-shifting acousto-optic modulator AOM2 (110MHz±Δf) as a modulation path. The two beams converge through a mirror to form interference fringes on the substrate surface. The fringe period is determined by the beam angle. ,in The wavelength of the laser. The beam angle is denoted as . Phase detection employs a PDB balanced detector, positioned at a π / 2 bias point in the interference field. Within the locked residual range, the phase error and the detector output voltage satisfy a linear approximation relationship: Detection gain The detection signal is acquired by the data acquisition card and sent to the host computer, where a PID+ILC composite control algorithm is run to output the frequency correction value. The signal is sent to the DDSPA frequency synthesizer via the USB interface, and then driven by the power amplifier to shift the frequency AOM2 to complete the phase closed-loop compensation.
[0011] (2) Modeling of the controlled object. The frequency shift execution chain is the core controlled object of the control system. It consists of a frequency synthesizer, a power amplifier and a frequency shift acousto-optic modulator cascaded together. Its input is the frequency correction command. (Hz), the output is the phase change of the interference fringes. The AOM shifts the optical frequency of the beam by changing its driving frequency. When the driving frequency of AOM2 is at the center frequency... =110MHz superimposed At that time, the frequency difference between the two beams of light in the interference field is... Interference fringe phase With frequency difference The integral is proportional to the time integral. After Laplace transform and gain normalization, the controlled object is represented as an "integral + delay" model: In the formula, the total system delay is... The latency is approximately 1.2ms, almost entirely contributed by the USB communication delay. The latency of other components is on the order of nanoseconds, which is much smaller than the servo period Ts=5ms and can be ignored. Gain normalization does not change the dynamic characteristics of the system, but only maps the control signal to the phase domain.
[0012] (3) Analysis of disturbance sources. Based on the time-frequency domain characteristics, the total phase tracking error of the system can be decoupled into a linear superposition of repetitive periodic disturbances and non-repetitive random disturbances. Repetitive disturbances originate from internal electrical coupling, inherent mechanical modes, and measurement nonlinearity: the AOM RF drive and photoelectric detection amplifier circuits are affected by mains radiation and ground loops, exhibiting significant 50Hz and higher harmonic peaks on the amplitude spectral density curve; peripheral equipment such as water chillers and vacuum pumps excite the optical platform to generate 190–200Hz mechanical resonance; the non-ideal extinction ratio of the polarization beam splitter introduces a periodic nonlinear error that is an integer multiple of the spatial frequency of the interference signal. Such disturbances are highly reproducible across batches, providing a theoretical basis for the non-causal feedforward compensation of ILC. Non-repetitive random disturbances are unpredictable background noise in the time domain and have a wide frequency domain distribution, including low-frequency random walks caused by air turbulence and temperature drift, external ground vibrations, and shot and thermal noise from photoelectric devices, which can only be suppressed in real time by high-bandwidth closed-loop feedback.
[0013] Step 2: PID feedback controller design.
[0014] (1) Design Requirements Analysis. In step-repeat exposure, the photoresist records the time integral exposure amount of a single 1-second exposure. After hundreds of complete oscillation cycles, the net contribution of high-frequency vibrations above 100 Hz approaches zero. Therefore, the system only needs a control bandwidth of tens of Hz to meet the accuracy requirements, which provides feasibility for PC-based digital control. The design of the PID controller needs to meet two indicators: first, the control bandwidth must be higher than the frequency band where the main energy of the disturbance is located and lower than the Nyquist frequency; second, the phase margin must be ≥60° to ensure robust stability under system delay.
[0015] (2) PI parameter design. A parallel PI controller is designed using the frequency domain method: The open-loop system is a Type II structure with no steady-state error for step reference; the crossover frequency is 378 rad / s (60.2 Hz), the phase margin is 62.4°, and the amplitude margin is 10.8 dB; the peak time of the step response is 6.86 ms, the overshoot is 4.17%, and the settling time is about 38 ms, which are much smaller than the 1 s exposure cycle.
[0016] (3) Discretization implementation. Forward Euler discretization is performed with a servo period Ts = 5ms to obtain the recursive control law: In the formula, This is the servo cycle number. The phase error in the k-th servo cycle is... This is the frequency correction value output by the PID feedback controller. The controller output is set with a saturation limit of ±10, corresponding to the frequency adjustment range of the frequency shift AOM based on the center frequency of 110MHz, to prevent the integrator from generating excessive control quantity under large errors, which could lead to actuator saturation.
[0017] Step 3: Design of ILC feedforward controller.
[0018] (1) Learning law design. For example... Figure 2 As shown, the composite controller consists of two parallel branches: a PID feedback controller outputs a correction value based on the real-time phase error at the time domain level (within 5ms of the servo cycle), which is the basic stabilization mechanism for all disturbances; and an ILC feedforward controller collects the entire batch of error trajectories and performs learning updates at the iterative domain level (offline updates during exposure intervals), possessing learning capabilities only for repetitive disturbances. The step-repetitive interferometric fringe phase-locked loop system meets the three applicable prerequisites of ILC: each exposure is locked to the same reference phase, the controlled object and reference trajectory remain unchanged across cycles, and repetitive disturbances have high predictability across cycles. This invention adopts a P-type learning law, the expression of which is: In the formula, For the batch number to be exposed, For a single exposure cycle, the time variable is... The ILC feedforward signal applied in the kth batch, This is the phase error trajectory for the kth batch. For proportional learning gain, It is a zero-phase-shift low-pass filter.
[0019] (2) Q-filter design. The Q-filter is a key component in ensuring robust convergence of the ILC. It applies selective attenuation to the learning update in the frequency domain: maintaining unity gain in the low-frequency band where repetitive disturbances occur to allow the ILC to learn fully; and applying attenuation in the high-frequency band where measurement noise and non-repetitive disturbances occur to prevent noise accumulation from causing iterative divergence. This invention uses a second-order Butterworth zero-phase-shift (filtfilt) low-pass filter with a cutoff frequency of 40Hz. The 40Hz cutoff frequency is mainly constrained by three factors: not significantly attenuating the frequency band where repetitive disturbances occur, effectively suppressing measurement noise near and above the Nyquist frequency (100Hz), and being lower than the starting frequency where the uncertainty of the controlled object model increases significantly. The zero-phase-shift filter eliminates the net phase delay through two filtering operations, forward and reverse, so that the Q-filter only performs amplitude attenuation without introducing phase distortion.
[0020] (3) ILC update timing. For example... Figure 3 As shown, the online phase operates within the servo cycle: at the start of the k-th exposure, the ILC feedforward signal for the current batch is loaded from the working area. The error is superimposed with the real-time correction output from the PID feedback controller to drive the frequency-shifted AOM, completing the phase-locked loop of the interference fringes; the complete phase error trajectory of the batch is recorded synchronously during exposure. The offline phase is executed during the step displacement interval after exposure, without occupying a servo cycle: the completed error trajectory is read, the temporary control signal is updated according to the P-type learning law, and after being low-pass filtered by a Q-filter with zero phase shift, it is written to the working area, awaiting the next exposure call. Convergence is determined based on the root mean square value of the phase error; if it is lower than a preset threshold, the iteration terminates.
[0021] (4) Convergence analysis. Define the closed-loop sensitivity function S and the process sensitivity function T: The phase error of the kth batch can be expressed as: Substitute into the P-type learning law and eliminate The error propagation equation in the iterative domain is obtained as follows: ILC monotonically converges if the amplitude of each frequency component of the error strictly decreases with each iteration, which is equivalent to the H∞ norm of the error propagation matrix being less than 1. Frequency band analysis: Low frequency band ≈1、 ≈1, the convergence condition degenerates to 0< <2; High-frequency Q-filter stopband attenuation will Suppress the error to a value much less than 1 to ensure high-frequency convergence and block noise accumulation. As the iterations approach infinity, the asymptotic error becomes: when =1 →0, but the measurement noise does not decay and accumulate in the iteration domain; when When <1, the frequency components within the stopband of the Q-filter exist permanently as residual errors, which are effectively attenuated within the bandwidth by the PID feedback loop. The two complement the ILC across the entire frequency band.
[0022] Step 4: Simulation verification.
[0023] A simulation model of a step-repetitive interferometric fringe phase-locked control system was built in MATLAB / Simulink. The controlled object adopted the aforementioned "integral + delay" transfer function, with a delay of... =1.2ms; PID uses the above PI parameters and is discretized with Ts=5ms; ILC learning gain =1.50, Q-filter cutoff frequency 40Hz, ILC signal injected into PID output via FromWorkspace in an additional manner. Perturbation configuration is based on dual-channel classification settings: repetitive perturbations include a sine component with amplitude 0.5rad and frequency 5Hz, a sine component with amplitude 0.1rad and frequency 50Hz, and a step sequence; non-repetitive perturbations are simulated by band-limited white noise with variance 0.002rad²; the reference phase is fixed at π / 2, single exposure duration is 1s, and a total of 40 batches are iterated.
[0024] Under pure PID phase-locked loop, the time-domain waveform of phase error is as follows: Figure 4 As shown: the error fluctuates with a period of 0.2s (corresponding to 5Hz mechanical vibration), and a plateau jump with an amplitude of 0.3rad occurs at approximately 0.7s (thermal drift), superimposed with dense high-frequency white noise spikes. The phase error RMS for the entire exposure cycle is 0.2850rad. After introducing ILC feedforward, as... Figure 5 As shown, the phase error RMS decreases monotonically with iteration and without oscillation: in the first 10 iterations, it decreases from 0.2850 rad to 0.2472 rad, contributing approximately 62% of the total reduction; from the 10th to the 40th iteration, it slowly converges to 0.2261 rad, a reduction of approximately 20% compared to pure PID. The lower limit of convergence is determined by the non-repetitive random perturbation, consistent with the asymptotic error prediction. In the step-repetitive process, the number of exposure fields on a single substrate is limited, and most of the accuracy improvement can be obtained in the first 10 iterations, with the convergence speed meeting the constraints of the actual process.
[0025] To evaluate the impact of control precision on the final grating quality, phase error is mapped to fringe contrast. During exposure, the photoresist records the exposure amount over time; random fluctuations in phase error affect the fringe contrast according to... Degradation. Substituting the steady-state RMS of pure PID and composite control, the contrast retention rates are 96.0% (loss of 4.0%) and 97.5% (loss of 2.5%), respectively, both meeting the linewidth uniformity and diffraction efficiency requirements of over 90% contrast retention in interference grating fabrication. Phase error mainly affects fringe contrast and absolute position, but does not change the fringe period, therefore it has no direct impact on grating period uniformity.
[0026] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A phase-locked control method for laser interferometric lithography fringes based on PID and ILC composite control, characterized in that, Includes the following steps: Step 1: Establish the "integral + delay" controlled object model of the frequency shift execution chain, which consists of a frequency synthesizer, a power amplifier and a frequency shift acousto-optic modulator cascaded together; based on the time-frequency domain characteristics, the phase error of the interference fringes is decoupled into a linear superposition of repetitive periodic perturbation and non-repetitive random perturbation; Step 2: Design a Type II PI feedback controller based on the controlled object model to dynamically suppress non-repetitive random disturbances in real time; Discretize the PI feedback controller with a preset servo period to obtain a discrete recursive control law; Step 3: Design a P-type iterative learning control learning law and introduce a zero-phase-shift low-pass Q-filter. Inject the ILC feedforward signal into the output of the PID feedback controller in an additional structure to learn and compensate for repetitive periodic disturbances between exposure batches without changing the characteristic equation of the feedback loop. Step 4: In the online phase, the PID feedback controller locks the phase in real time and superimposes the ILC feedforward signal in each servo cycle to drive the frequency shift execution chain to complete the phase closed-loop compensation; in the offline phase, the entire batch of phase error trajectories is collected during the exposure interval, ILC learning and updating is performed and written to the working area for use in the next exposure.
2. The laser interferometric lithography fringe phase-locked control method based on PID and ILC composite control according to claim 1, characterized in that, Step 1, establishing the controlled object model, specifically involves: The ultraviolet laser beam is split into two paths by a beam splitter after passing through a fixed-frequency acousto-optic modulator AOM3: one path passes through a fixed-frequency acousto-optic modulator AOM1 as a reference path, and the other path passes through a frequency-shifting acousto-optic modulator AOM2 as a control path. The two beams converge on the substrate surface to form interference fringes. When the driving frequency of the frequency-shifting AOM2 is superimposed with a frequency correction Δf on the center frequency, the frequency difference between the two beams in the interference field is Δf. The fringe phase φ is proportional to the time integral of the frequency difference Δf. After Laplace transform and gain normalization, the controlled object is represented as an "integral + delay" model. ,in To perform chain gain, The total system delay; the total system delay The latency is approximately 1.2ms, mainly contributed by USB communication delay. The latency of other components is on the order of nanoseconds, which is much smaller than the servo cycle Ts=5ms.
3. The laser interferometric lithography fringe phase-locked control method based on PID and ILC composite control according to claim 1, characterized in that, Step 1 decouples the phase error, specifically as follows: Repetitive periodic disturbances originate from internal electrical coupling, inherent mechanical modes, and measurement nonlinearities within the system. These include power frequency and high-order harmonic coupling generated by the mains radiation and ground loop effects on the acousto-optic modulator RF drive and photoelectric detection amplifier circuits, mechanical resonance generated by the steady-state operation of peripheral equipment excitation of the optical platform, and periodic nonlinear errors introduced by the non-ideal extinction ratio of the polarization beam splitter. Non-repetitive random disturbances are unpredictable background noise in the time domain and have a broadband distribution in the frequency domain. These include low-frequency random walks caused by air turbulence and temperature drift, broadband mechanical shocks introduced by external ground vibrations, and shot noise and thermal noise from optoelectronic devices.
4. The laser interferometric lithography fringe phase-locked control method based on PID and ILC composite control according to claim 1, characterized in that, Step 2, designing the Type II PI feedback controller, specifically involves: A parallel PI controller is designed using the frequency domain method, with a transfer function of C_PI(s) = K_p + Ki / s, making the open-loop system a type II structure with no steady-state error for a step reference input and ensuring preset phase and gain margins. Forward Euler discretization is performed over the servo period Ts to obtain the recursive control law. ,in The phase error in the k-th servo cycle is... This is the frequency correction value output by the PID feedback controller in the k-th servo cycle; the controller output is set... Saturation limiting is used to prevent actuator saturation.
5. The laser interferometric lithography fringe phase-locked control method based on PID and ILC composite control according to claim 1, characterized in that, Step 3, designing the P-type ILC learning law, specifically involves: The P-type learning law is Where k is the exposure batch number and t is the time variable within a single exposure cycle. For proportional learning gain, The filter is a zero-phase-shift low-pass Q-filter. The Q-filter is a second-order Butterworth zero-phase-shift low-pass filter with a cutoff frequency lower than the starting frequency of significant growth in uncertainty of the controlled object model. It maintains unity gain in the passband to ensure the learning ability for low-frequency repetitive disturbances, and applies attenuation in the stopband to prevent high-frequency measurement noise from entering the iteration accumulation. The ILC feedforward signal is injected into the PID output terminal in an additional structure, does not enter the feedback loop, does not change the closed-loop characteristic equation, and the stability margin of the feedback loop is not affected by the ILC iteration update.
6. The laser interferometric lithography fringe phase-locked control method based on PID and ILC composite control according to claim 1, characterized in that, In step 4, the ILC offline learning update is performed during the step displacement interval after exposure is completed, without occupying the servo cycle. The learning update process is as follows: read the phase error trajectory of the kth batch that has just been completed, update the temporary control signal according to the P-type learning law, and obtain the feedforward signal of the next batch after zero phase shift low-pass filtering by Q-filter, and write it into the working area; use the root mean square value of the phase error as the convergence criterion, if it is lower than the preset threshold, the iteration is terminated, otherwise the batch number is incremented, and the process returns to the online stage to continue execution.
7. A laser interferometric lithography fringe phase-locked control system based on PID and ILC composite control, characterized in that, include: An optical interference platform is used to generate two coherent beams of light and converge them on the surface of a substrate to form interference fringes; The phase detection module uses a balanced detector placed at the π / 2 offset point of the interference field to detect the phase error of the interference fringes; The control module is used to execute the phase-locked control method according to any one of claims 1 to 6 and output a frequency correction amount; The actuator module includes a frequency synthesizer, a power amplifier, and a frequency-shifting acousto-optic modulator, which is used to drive the frequency-shifting acousto-optic modulator according to the frequency correction amount to complete phase closed-loop compensation.
