{"paper_id":"1d0bcec7-8384-4fa7-8caf-550e6b3a992c","body_text":"Intelligent configuration of integrated microwave photonic filter with programmable response and self-stabilization | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Intelligent configuration of integrated microwave photonic filter with programmable response and self-stabilization Yuan Yu, Yutong Shi, Yifan Liu, Kaixiang Cao, Mengmeng Deng, Fangzheng Zhang, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6669261/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Integrated microwave photonic filters (IMPFs) have emerged as promising candidates for advanced microwave systems owing to their distinctive combination of wide operational bandwidth, reconfigurable functionality, and compact size. Nevertheless, the complex and time-consuming manual manipulation of IMPFs remains a significant impediment to their widespread applications. Here, to the best of our knowledge, we experimentally demonstrate the first intelligent configuration of IMPF featuring wideband center frequency tunability, flexible bandwidth reconfigurability, self-stabilization, and excellent channel equalization simultaneously. The configuration strategy is enabled by our proposed universal hybrid collaboration algorithm, which can fully unleash the hardware potential of the optical device, thus enabling comprehensive synergy of multiple properties. Results show that the center frequency of IMPF is tuned from 2 to 48 GHz, and the bandwidth is reconfigured from 0.66 to 4.15 GHz, with a rejection ratio of up to 37.67 dB. The roll-off rate and shape factor reach as high as 17.50 dB/GHz and 0.78, respectively. Meanwhile, the maximum center frequency drift of IMPF in 3 hours is reduced from 11.950 to 0.051 GHz even without a thermo-electric cooler, indicating that the center frequency stability is enhanced by 234 times. The passband shape of the IMPF can also be dynamically adjusted to equalize frequency-dependent fading, achieving up to 2.42 dB compensation of intra-channel fading. Our work highlights the potential of IMPFs based on intelligent configuration, unlocking new avenues for practical applications of microwave photonic signal processing. Physical sciences/Optics and photonics/Applied optics/Microwave photonics Physical sciences/Optics and photonics/Applied optics/Integrated optics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Introduction Traditional radio frequency (RF) electronic technologies, constrained by limited sampling rates and component bandwidths, increasingly fail to meet the demands of next-generation communication systems, which require large bandwidth, flexible reconfigurability, excellent programmability, low cost, low consumption, and electromagnetic immunity 1 , 2 , 3 , 4 . In contrast, microwave photonics offers a breakthrough pathway by leveraging its inherent advantages, including ultra-wide bandwidth, high flexibility, low loss, and electromagnetic immunity 4 , 5 , 6 . As fundamental components, microwave photonic filters are used to select target signals from background noise or reject disturbing noise from signals. Advances in photonic integrated circuit (PIC) technology have driven the growing adoption of integrated microwave photonic filters (IMPFs) 7 , 8 , 9 , 10 , 11 , 12 , which offer significant advantages in terms of size, weight, power consumption, and cost 4 , 8 . Currently, IMPFs have been demonstrated on various platforms, including silicon-on-insulator (SOI) 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , silicon nitride 20 , 21 , 22 , 23 , indium phosphide 9 , 24 , 25 , lithium niobate 26 , and chalcogenide glass 27 , 28 , 29 . Compared with electronic filters, IMPFs exhibit superior tunability and reconfigurability, which drive their broad application prospects in wireless communications 8 , 29 , 30 , sensor networks 31 , and radar systems 4 , 10 . IMPFs require tailored configurations for various scenarios. For example, optoelectronic oscillators (OEOs) require IMPFs with a narrow bandwidth 32 , 33 , and radar systems demand IMPFs with a high rejection ratio (RR), wideband operation, and flexible tunability 19 , 34 , 35 . With the development of integrating multidimensional functions into a single system, reconfiguring IMPF responses according to diverse requirements significantly enhances economic efficiency. Although the flexibility of IMPFs has been demonstrated 12 , 14 , 20 , 26 , most reported reconfigurable IMPFs primarily depend on cumbersome manual manipulation for frequency tuning and bandwidth reconfiguration, which presents significant challenges for achieving real-time, dynamic, and self-adaptive configuration for complex applications. Many reconfigurable IMPFs utilizing various techniques have been demonstrated, such as Mach-Zehnder interferometer (MZI)-assisted microring resonators (MRRs) 13 , 36 , 37 , cascaded MRRs 21 , 26 , 29 , 38 , coupled resonator optical waveguides (CROWs) 15 , 39 , 40 , 41 , and programmable two-dimensional (2D) mesh network topologies 12 , 20 , 42 . These IMPFs exhibit superior performance in key properties, including RR, tunable range of center frequency, reconfigurable range of full width at half-maximum (FWHM) bandwidth, shape factor (SF), and multifunctional integration. Nonetheless, these designs often face performance compromises due to insufficient configuration, making it difficult to achieve synergy of all properties. Furthermore, their sustained and stable operation and resistance to random environmental disturbances have not been sufficiently addressed. Overall, traditional manual manipulation and offline design methods fail to meet the requirements for synergy of all properties, dynamic configuration, and sustained and stable operation of IMPFs, which significantly hinder their application prospects. Therefore, developing a supporting intelligent configuration framework is essential for advancing IMPF toward practical wireless applications. In this work, we propose a universal multi-objective and multi-stage hybrid collaboration (MOMS-HC) algorithm for the intelligent configuration of IMPFs. The algorithm innovatively leverages the complementary strengths of particle swarm optimization (PSO) 43 , genetic algorithm (GA) 44 , and simulated annealing (SA) 45 through a staged collaboration mechanism. Specifically, the task-driven specialization of each sub-algorithm at different stages governs the collaboration, enabling multi-algorithm hybrid execution and complementary strengths integration. The algorithm integrates global-local exploration, robust multi-objective equilibrium, efficient self-optimization, and environmental robustness empowerment, and fully unleashes the hardware potential of IMPF. Utilizing the MOMS-HC algorithm, we implement, to the best of our knowledge, the first intelligently configured bandpass IMPF, whose hardware is based on a 4th-order coupled resonator optical waveguide (CROW) on a commercial SOI wafer. Experimental results demonstrate that the IMPF achieves unprecedented synergy of almost all properties, including a tuning range of 2–48 GHz, an FWHM bandwidth reconfiguration range of 0.66–4.15 GHz, a roll-off rate of up to 17.50 dB/GHz, an in-band ripple of below 0.91 dB, an RR of up to 37.67 dB, and an SF of up to 0.78. Notably, with the MOMS-HO algorithm, the maximum center frequency drift of IMPF is only 0.051 GHz in 3 hours. Compared with the free-running case, the frequency stability is enhanced by 234 times. Results also show that the MOMS-HO algorithm enables the IMPF to quickly recover to its target operating state when significant disturbances occur. Additionally, IMPF-based equalization of the frequency-dependent fading in a microwave photonic link based on the MOMS-HO algorithm is demonstrated, and up to 2.42 dB improvement for intra-channel fading is obtained by reshaping the passband of IMPF. The proposed intelligent configuration approach provides critical support for practical applications of IMPFs and is expected to accelerate the deployment of high-performance integrated optical devices. Results Principle of the MOMS-HC algorithm and device The integrated microwave photonic signal processing is fundamentally implemented based on the optical-to-electrical mapping technique. Therefore, the integrated optical processing unit (IOPU) must be carefully designed and configured to achieve high-performance signal processing. As a result, we first consider the general problem of configuring numerous electrically controlled adjustable parameters to ensure that the IOPU converges to the desired response. The challenges of this problem include the vast parameter space resulting from complex structures and the inter-coupling between multiple adjustable parameters. Therefore, we propose to consider the utilization of advanced machine learning algorithms to transform the configuration problem of IOPU into an optimization problem. The optimization objective and variables are the required frequency response and voltage applied to the adjustable parameters. However, to realize an IOPU with superior overall performance, we must prioritize the multi-objective equilibrium of the configuration algorithm. Based on these ideas, we propose the MOMS-HC algorithm. The execution flowchart of the MOMS-HC algorithm, which consists of four stages, is shown in Fig. 1a. The “Preprocessing Stage” (Stage I) first programmatically defines various properties of the IOPU and a figure of merit (FOM) to guide subsequent optimization. The “Global Search Stage” (Stage II) executes PSO optimization, where swarm intelligence drives rapid global optimization using a single critical property of the IOPU as the FOM. The “Dynamic Configuration Stage” (Stage III) employs GA, which specializes in precision optimization under multi-objective constraints. The GA performs collaborative multi-objective optimization with a redefinable FOM. It enables IMPF to achieve synergy of multiple properties due to leverage roulette selection, gene crossover, and mutation operations, while a flexible population management mechanism balances search efficiency and robustness. In the “Self-stabilization Stage” (Stage IV), the SA algorithm is employed to suppress performance degradation caused by environmental disturbances, such as temperature fluctuations 46 , mechanical vibrations, and power supply noise 47 , 48 . Through a customized temperature decay mechanism, SA enables rapid local optimization while progressively constraining the amplitude of voltage adjustments. This mechanism suppresses secondary performance fluctuations induced by voltage searches, ensuring sustained consistency and self-stabilization of the IOPU under complex operating conditions (see Methods for more details of the MOMS-HC algorithm). The principle of the integrated microwave photonic signal processing based on intelligent configuration is shown in Fig. 1b. The continuous wave (CW) light emitted by a laser source (LS) is modulated by microwave signals via a modulator. Then, the modulated microwave photonic signal is processed by IOPU. Following processing, the microwave signal is extracted by a photodetector (PD) through optical-electrical (O/E) conversion. The resulting microwave signal is fed back to the intelligent configuration module, which subsequently calculates the control signals based on the configuration targets. After calculation, the output control signals are applied to the IOPU, thereby enabling dynamic reconfiguration of the IOPU. Conventionally, dynamic configuration of high-order CROW is quite challenging because of inherent structural complexity and strong inter-coupling between adjustable parameters in compact size 15 , 41 . Meanwhile, filtering is one of the most classical and commonly used signal processing techniques. Therefore, the implementation of intelligent configuration of IMPF based on 4th-order CROW 39 is utilized to demonstrate the superior performance of our proposed MOMS-HC algorithm. The 4th-order CROW is composed of four identical MRRs integrated with 14 titanium nitride (TiN) microheaters, as shown in Fig. 1c. The detailed parameters of each MRR are shown in Fig. 1d. The MRR coupling regions are implemented as MZI couplers consisting of two identical directional couplers (DCs) and two phase shifters. These couplers are designed as two single-mode waveguides with three 60° arcs and a bending radius of 20 μm, as illustrated in Fig. 1e. The device employs ridge waveguides with a ridge height of 130 nm to reduce the waveguide loss, as shown in Fig. 1f. The values of r and k should satisfy in order to obtain the maximum tuning range of the equivalent coupling coefficients 38 , 40 . Fig. 1g shows the false-color image of the fabricated 4th-order CROW serving as the verification device. The device is fabricated based on an SOI wafer with a 220 nm-thick silicon layer and a 2 μm-thick buried oxide (BOX) layer. The four microheaters (H 1 -H 4 ) deposited over individual MRR waveguides are used for tuning the resonant wavelength of each MRR. The remaining microheaters (H 5 -H 14 ) deposited over the upper and lower arms of MZI couplers are used for adjusting equivalent coupling coefficients between coupled MRRs, as shown in Fig. 1c and 1g. In practical operation, only one microheater (H 5 -H 9 ) of the adjustable MZI coupler actively controls coupling coefficients, and the others (H 10 -H 14 ) are used for backup. These microheaters enable refractive index modulation in waveguide structures by changing local temperature 20 . The measured linear resistances of the nine microheaters H 1 -H 9 are 1.5491, 1.5532, 1.6022, 1.5848, 0.2109, 0.2001, 0.2184, 0.2361, and 0.2222 kΩ, respectively. To ensure the stability of the optical and electrical ports during experimental tests, the on-chip device is optically and electrically packaged, as shown in Fig. 1h. The inset shows the zoomed-in view of the fabricated device. All pads fabricated on the chip are connected to DuPont connectors on a printed circuit board (PCB) via gold wires. The four grating couplers (GCs) are coupled and packaged with a standard fiber array (FA). When the microwave photonic signal is coupled into the device through GC 1 , the signal will be processed and then coupled out of the device through GC 3 . Experimental setups Fig. 2 illustrates the experimental setup for dynamically configuring the frequency response of the IMPF based on our proposed MOMS-HC algorithm. A CW light at 1550 nm generated by a laser source (LS, Koheras BasiK E15) is divided into two branches by an optical coupler (OC 1 ) with a 7:3 splitting ratio. In the upper branch, 70% of the CW light is utilized for microwave signal modulation. The state of polarization (SOP) of the optical carrier is adjusted by a polarization controller (PC 1 ) to be aligned with the principal polarization axis of the phase modulator (PM, Covega Mach-40). The microwave signal emitted by a vector network analyzer (VNA, Anritsu MS4647B) is modulated onto the optical carrier via PM. An optical bandpass filter (OBPF, Alnair BVF-300CL) is used to eliminate the -1st-order sideband, generating a single-sideband (SSB) signal (see Supplementary Notes IV for more details of the modulated signal). Then, the SSB signal is amplified by an erbium-doped fiber amplifier (EDFA, WZEDFA-SO-L-CW40) and processed by the device under test (DUT) after SOP adjustment via PC 2 . The +1st-order sideband signal processed by the DUT is combined with the remaining 30% CW light in the lower branch via OC 2 . The combined optical signal is then injected into a photodetector (PD, SHF 47100A) for O/E conversion to obtain the microwave signal. The microwave signal is amplified by a broadband electrical amplifier (EA, SHF S804B) before being fed into the VNA for measurement. PC 3 is used to adjust the SOP of the optical carrier in the lower branch to maximize the power of the beat frequency signal, and a variable optical attenuator (VOA) is used to adjust the optical power fed into the PD. Considering the extensive measurement range of the optical spectrum analyzer (OSA), we first conduct preliminary tests in the optical domain, and execute the intelligent configuration of the DUT before configuring the IMPF. Based on the MOMS-HC algorithm, an optical filter with an RR of 18.93 dB and an FWHM bandwidth of 0.018 nm (2.25 GHz) is realized (see Supplementary Notes V for more details). In brief, the intelligent configuration module consists of three components: controller, sensor, and actuator. The VNA functions as a sensor in the system, monitoring the operation state of the controlled device and relaying the data back to the computer, which serves as the central controller. The computer formulates the optimization problem via programming on the MATLAB platform and transmits the required control signals to the voltage source array (VSA, MCVS6400-A), which acts as the actuator. Both the sensor and the actuator are connected and controlled by the same computer. Intelligent configuration of IMPF In traditional manual manipulation of the IMPF, operators usually rely on accumulated experience and subjective judgment, making it difficult to systematically explore the optimization parameter space and prone to getting stuck in local optima (see Supplementary Notes III for more details). In contrast, the MOMS-HC algorithm enables synergy of all properties and fully unleashes the performance potential of its hardware. We first demonstrate an intelligent configuration of the IMPF based on the 4th-order CROW, aiming for a center frequency of 30 GHz, an FWHM bandwidth of 3 GHz, an RR higher than 30 dB, an SF higher than 0.7, and an in-band ripple smaller than 0.6 dB. Fig. 3a shows the normalized FOM as a function of the number of iterations during the configuration process, where the blue and green curves correspond to Stage II and III, respectively (see Supplementary Notes II for FOM definitions). The preprocessing and predefined steps in Stage I are omitted here due to their negligible time cost. Fig. 3b-e and 3f-i show the voltage distribution applied to all microheaters from H 1 to H 9 and the amplitude frequency response of the IMPF after different iterations, corresponding to nodes “I”, “II”, “IV”, and “V”, respectively. Specifically, Fig. 3b and 3f present the results after 8 iterations of the PSO algorithm in Stage II, corresponding to node “I” in Fig. 3a. At this node, the IMPF demonstrates a basic filtering function, but the passband exhibits splitting with poor edge symmetry, and the center frequency is approximately 40 GHz, as shown in Fig. 3f. This degradation is due to the significant temperature fluctuations of the chip caused by the extensive voltage search in this stage. However, the degradation does not affect the subsequent configuration. In Stage III, the normalized FOM increases as the number of iterations increases. However, it does not increase monotonically and exhibits significant degradation and oscillation in early iterations, as shown by the FOM variation between node “I” and node “III” in Fig. 3a. There are four main reasons for this phenomenon. Firstly, the global search for optimal control voltage at the beginning of each iteration leads to fluctuations in chip temperature. This is because abrupt voltage jumps (e.g., sudden high-to-low transitions) induce drastic heating power variations, which consequently lead to chip temperature fluctuations. Secondly, the adjustable parameters of the CROW-based IMPF interact with each other due to the complexity of its inherent structure. For example, adjusting either microheater of the MZI-based coupler not only modifies the coupling coefficient between the two adjacent MRRs but also alters their resonant wavelengths due to simultaneous optical phase shift. Moreover, the thermal crosstalk can further exacerbate the inter-coupling between these thermo-optic adjustable parameters. Thirdly, GA needs to simultaneously handle multiple optimization objectives, including FWHM bandwidth, center frequency, RR, SF, roll-off rate, insertion loss (IL), in-band ripple, response symmetry, and response smoothness associated with the absence of passband splitting, which results in a high degree of complexity in the evolution process (see Supplementary Notes II for more details). As a result, a certain amount of time is required for adaptive selection to balance the weights of different objectives. Fourthly, environmental temperature fluctuations and mechanical vibrations, as well as power supply noise 47 , 48 , also lead to degradation of the IMPF performance. These factors are also the main challenges in achieving high-performance IMPF. After 25 iterations, the FOM reaches 0.953, as shown by node “III” in Fig. 3a. We can see that the frequency response of the IMPF gradually converges, indicating that the overall temperature of the chip is stabilizing. Fig. 3h shows the frequency response of the IMPF after 31 iterations, which corresponds to node “IV” in Fig. 3a. The normalized FOM reaches 0.969. The overall performance is satisfactory except for the in-band ripple, which is 0.85 dB. Subsequent iterations focus on further flattening the passband to approach the target response. Finally, after 42 iterations, the FOM reaches 0.982 corresponding to node “V” in Fig. 3a. The corresponding IMPF achieves an FWHM bandwidth of 2.97 GHz, a center frequency of 29.899 GHz, an RR of 31.21 dB, an in-band ripple of 0.52 dB, an SF of 0.74, and a roll-off rate of 13.08 dB/GHz simultaneously, as shown in Fig. 3i. The results show that all properties of the IMPF perfectly meet the expected targets. Tuning and reconfiguring the response of IMPF The powerful tunability and reconfigurability are essential capabilities of the IMPF. To demonstrate the excellent performance of the intelligently configured IMPF, we first evaluate its consistency while tuning the center frequency from 2 to 48 GHz with different target bandwidths. Fig. 4a illustrates the measured center frequency tuning results of the IMPF with bandwidths of 0.66, 1.85, and 3.15 GHz, respectively. From the measured results, we can draw the following conclusions. Firstly, for the narrow bandwidth operation (0.66 ± 0.02 GHz), the IMPF exhibits a maximum RR of 31.47 dB with a roll-off rate up to 17.50 dB/GHz, as shown in Fig. 4a I. This high roll-off rate improves the rejection of out-of-band interference while preserving precise frequency selectivity, a crucial aspect of high-performance filtering. Secondly, for the medium bandwidth operation (1.85 ± 0.10 GHz), the maximum RR reaches 37.67 dB with a roll-off rate up to 13.21 dB/GHz, delivering balanced selectivity for multi-channel communication environments, as shown in Fig. 4a II. Finally, for the wide bandwidth operation (3.15 ± 0.03 GHz), the IMPF maintains an in-band ripple of less than 0.61 dB, as shown in Fig. 4a III. Meanwhile, the RR, roll-off rate, and SF are as high as 31.42 dB, 13.33 dB/GHz, and 0.75, respectively. Fig. 4b-d present labeled heatmaps illustrating the bandwidth, RR, and SF of the IMPF at different center frequencies, respectively. In Fig. 4b-d, the horizontal axis represents three bandwidth labels, corresponding to respective 0.66 GHz I , 1.85 GHz II , and 3.15 GHz III , while the vertical axis represents the center frequency labels. The color intensity in each heatmap corresponds to the magnitude of the respective property (bandwidth, RR, and SF). We can see that the variations of bandwidth and RR during the center frequency tuning are less than ±0.10 GHz and ±1.00 dB, respectively. The results demonstrate that the performance of the IMPF maintains excellent consistency across the whole operational band, representing a key advantage of IMPFs. Meanwhile, the FWHM bandwidth of the IMPF can be continuously reconfigured from 0.66 to 4.15 GHz by jointly configuring the control voltages via the MOMS-HC algorithm, and the SF varied between 0.45 and 0.78, as shown in Fig. 4e and 4f, respectively. When the bandwidth is reconfigured from 0.66 to 3.68 GHz, the in-band ripple stays below 0.63 dB, and the RR stays above 30.14 dB, as shown in Fig. 4f and 4g, respectively. When the bandwidth of the IMPF is reconfigured to 4.15 GHz, the RR drops to 26.66 dB, and the in-band ripple increases to 0.91 dB, as shown by dark purple curve in Fig. 4e. The degradation in overall performance primarily stems from the limited adjustable range of the equivalent coupling coefficient based on the MZI coupler, resulting in a trade-off between the RR and in-band ripple. The frequency response with a bandwidth of 4.15 GHz indicates that the 4th-order CROW is approaching its hardware-imposed performance limit. The result also demonstrates the capability of the configuration framework based on the MOMS-HC algorithm to fully unleash the hardware potential of the optical device. Self-stabilization and self- recover y The self-stabilization and sustained operation capability of the IMPF are critical for a practical system. Under manual manipulation, the absence of dynamic adjustment and stabilization mechanisms makes it challenging to maintain sustained and stable operation of the IMPF. In contrast, the MOMS-HC algorithm enables dynamic self-configuration of adjustable parameters during IMPF operation to effectively suppress the negative effects of environmental disturbances. The elaborate design of the SA algorithm aims to provide real-time voltage compensation while minimizing temperature fluctuations as much as possible (see Supplementary Notes I for more details). Before evaluating the self-stabilization of the IMPF, we first configure its various properties as follows: the center frequency and FWHM bandwidth are set as 15.000 and 0.90 GHz, and the SF and RR are set to exceed 0.50 and 30.00 dB, respectively. The above state serves as the initial state for all subsequent evaluations in this section. Next, we let the IMPF operate in the laboratory environment for three hours with MOMS-HC off and on, respectively. The basic operating cycle (BOC) is set as 60 seconds. When the MOMS-HC is off, the IMPF operates normally within each BOC without executing additional intelligent configurations. When the MOMS-HC is on, SA obtains the current amplitude frequency response of the IMPF in real-time during each BOC, monitoring degradation of properties such as center frequency, SF, RR, and bandwidth. Any degradation exceeding the threshold will trigger the intelligent configuration process, and the configuration is constrained to be completed within a BOC. The current amplitude frequency response of the IMPF captured at the end of each BOC will be overlapped to generate an “eye-like” diagram. The “eye-like” diagrams without and with MOMS-HC-based configuration are shown as Fig. 5a and 5b, respectively. We can see that the “eye-like” diagram is highly chaotic when the MOMS-HC is off, indicating significant performance degradation of the IMPF over the 3 hours. The degradation is primarily caused by environmental disturbances, such as temperature fluctuations 46 , mechanical vibrations, and accumulated power supply noise 47 , 48 . In contrast, when the MOMS-HC is on, the “eye-like” diagram distinctly exhibits sustained and stable dynamic characteristics of the IMPF. The nine voltages applied to the microheaters are dynamically adjusted in real time during IMPF operation, as shown in Fig. 5c. Table 1 shows the comparison of IMPF with MOMS-HC on and off during the 3-hour stability test. When the MOMS-HC is on, the mean center frequency of the IMPF is 15.002 GHz with a relative standard deviation of 0.085%, and the fluctuation range of center frequency is between 14.970 and 15.051 GHz. The maximum frequency drift is less than ± 0.051 GHz, yielding a 234-fold improvement in stability compared to the maximum frequency drift of up to 11.950 GHz when the MOMS-HC is off. The average SF is 0.52 with a relative standard deviation of 0.667%, and the average bandwidth is 0.90 GHz with a relative standard deviation of 0.682%. During the 3-hour stability test, both the RR and roll-off rate remain above 29.89 dB and 16.57 dB/GHz, respectively. These results demonstrate that the MOMS-HC algorithm effectively suppresses performance degradation caused by environmental disturbances, allowing the IMPF to maintain sustained stability and consistency over an extended period. The detailed time-dependent variations of the SF, center frequency offset, bandwidth, and RR without and with MOMS-HC-based configuration are shown in Fig. 5d-g, respectively. Table 1. Comparison of IMPF with MOMS-HC on and off Property BW (GHz) RR (dB) Center (GHz) SF Target 0.90 >30.00 15.000 >0.50 MOMS-HC off Mean 0.70 24.18 20.508 0.24 Min 0.56 17.59 15.091 0.10 Max 0.99 30.80 26.950 0.53 Rela.Std. Dev. 17.553% 16.367% 15.048% 51.708% MOMS-HC on Mean 0.90 30.52 15.002 0.52 Min 0.88 29.89 14.970 0.51 Max 0.92 31.36 15.051 0.54 Rela.Std. Dev. 0.682% 0.648% 0.085% 0.667% Abbreviations: Rela.Std. Dev.: relative standard deviation, BW: bandwidth. We also evaluate the self-recovery capability of the intelligently configured IMPF under extreme environmental disturbances. In this evaluation, we let the IMPF operate for 30 minutes with MOMS-HC on. The initial state is the same as that of the 3-hour stability test, operating with a center frequency of 15.010 GHz and a bandwidth of 0.90 GHz, as indicated by the green curve in Fig. 5h. To simulate extreme disturbances, random voltage biases of up to 200 mV are simultaneously applied to all nine control ports at the beginning of the 4th BOC. The red curve in Fig. 5h illustrates the degraded frequency response of IMPF at the beginning of applying voltage disturbances to the control ports. We can observe that the center frequency of the IMPF instantaneously drifts toward higher frequencies by 1.637 GHz, and the RR decreases by 10.62 dB, from 30.27 dB to 19.65 dB. Concurrently, the SF drops to 0.14, and the passband is severely split. However, the MOMS-HC algorithm enables the IMPF to recover to its initial operation state within 5 minutes, as demonstrated in Fig. 5i. During the subsequent 22-minute period, the IMPF maintains an optimal operation state with an RR exceeding 30.01 dB and a roll-off rate above 15.00 dB/GHz. Fig. 5j indicates the “eye-like” diagram characterizing the stable performance of the IMPF from the 9th to the 30th minute. Meanwhile, Fig. 5k-n display the temporal variations of SF, frequency offset, bandwidth, and RR over the 30-minute evaluation period, respectively. IMPF-based channel equalization Table 2. Results of Channel Equalization Center (GHz) BW (GHz) Unequalized Ripple (dB) Equalized Ripple (dB) Ripple Reduction (dB) 5.5 3.02 3.66 1.49 2.42 22.0 3.51 4.24 1.87 2.37 27.0 3.22 2.98 0.84 2.14 32.5 4.34 3.44 1.53 1.91 Abbreviations: BW: bandwidth. In high-speed and large-capacity communication systems, signals using advanced modulation formats are prone to distortion induced by electro-optic and photoelectric conversion, electrical amplifiers, and fiber dispersion. Among these challenges, frequency-dependent fading stands as the primary impairment mechanism. This is primarily caused by two factors: the chromatic dispersion in optical fibers and the inherent frequency-dependent amplitude responses of the components (e.g., PM, PD, and EA) in link 49 . Traditionally, this issue can be solved by modifying the architecture of intensity modulation and direct detection systems, including single-sideband or vestigial-sideband modulation, intensity modulation with coherent detection, and Kramers-Kronig receiver 50 , 51 . However, these approaches inevitably increase system complexity and cost. Alternatively, advanced digital signal processing (DSP) techniques are employed to perform pre-equalization based on accurate channel estimation and channel feedback, but these techniques often suffer from high computational complexity and power consumption 54,55 . In this section, the intelligently configured IMPF, based on the MOMS-HC algorithm, is employed to compensate for frequency-dependent fading and achieve channel equalization. This equalization capability is achieved by dynamically reshaping the passband response of the IMPF (see Supplementary Notes II for more details). To demonstrate this capability, we directly utilize the microwave photonic (MWP) link shown in Fig. 2 to simulate the practical scenario with frequency-dependent fading. Fig. 6a shows the inherent amplitude frequency response of the MWP link without the IMPF based on the 4th-order CROW. We can observe significant amplitude ripples and degradation across the operational band from 2 to 50 GHz. These effects are attributed to the combined influence of the inherent responses of each device in the MWP link. Specifically, within the 2–10 GHz range, the filtering edge of the OBPF adopted for SSB modulation induces higher loss for the lower frequency compared to the higher frequency. In the 10–50 GHz range, the combined effects of parasitic capacitance in the EA 54 , 55 and the bandwidth constraints of both the PD and PM lead to degradation in response flatness and insufficient responsivity at the high frequencies. Therefore, we set the channel window centers at 5.5, 22.0, 27.0, and 32.5 GHz and set different target bandwidths for each to evaluate the dynamic equalization capability of the IMPF. The green and red curves in Fig. 6b-e indicate the equalized channel and the configured IMPF amplitude frequency responses, respectively. We can see that multiple equalized MWP links with flat passbands at various center frequencies are established. These responses demonstrate that, owing to the MOMS-HC algorithm, the IMPF passband is adjusted to match the inherent response of the link, and its transition edges are also adjusted to ensure excellent symmetry in the equalized link response. Table 2 shows channel equalization results with different center frequencies and bandwidths. We can see that when the center frequency of the channel window is respective 5.5, 22.0, 27.0, and 32.5 GHz, the corresponding in-band ripple before equalization is 3.66, 4.24, 2.98, and 3.44 dB, respectively. After applying the IMPF-based equalization to the MWP link, the corresponding in-band ripple is reduced to 1.49, 1.87, 0.84, and 1.53 dB, respectively. Therefore, these experimental results demonstrate the effectiveness of IMPF-based equalization. Additionally, we can observe from Fig. 6b-e that the equalized channel response has different excess ILs at different channel windows. This is because the PM used in the experiment is polarization-dependent, and SOP variations during the configuration process can result in gain degradation in the MWP link. To solve this problem, an electrically controlled PC can be used and co-configured by the MOMS-HC algorithm 56 . Discussion Based on the MOMS-HC algorithm, we have successfully implemented an intelligently configured IMPF with continuous tuning of the center frequency from 2 to 48 GHz and a wide bandwidth reconfiguration range from 0.66 to 4.15 GHz, featuring self-stabilization and self-recovery capabilities. The MOMS-HC algorithm also significantly enhances the performance consistency of IMPF during the frequency tuning. In contrast to conventional equalization techniques, the MOMS-HC algorithm enables the IMPF to equalize intra-channel fading through direct passband reshaping, eliminating the need for extra costly devices or digital equalization algorithms. This hardware-instead-of-software approach promises to enable end-to-end integrated microwave photonic system architectures 7 , 42 . Meanwhile, the proposed MOMS-HC algorithm is fundamentally an in situ training process. The “train-as-deploy” paradigm ensures that the IMPF can be directly used for specific wireless communication tasks once the training is complete. Owing to this paradigm, the IMPF also circumvents the reliance on electromagnetic simulations and the challenges of physical alignment faced by traditional methods. In addition to IMPFs, the MOMS-HC algorithm is also universally applicable to IOPUs of arbitrary architectures and can be extended to any scenario where a FOM can be defined to evaluate the state of the system. Notably, bandwidth reconfiguration of IMPFs over a broader range remains challenging, particularly for achieving bandwidths as narrow as the MHz level. Two main factors limit the reconfigurable bandwidth range of the CROW-based IMPFs. Firstly, high waveguide loss results in an inadequate microring Q-factor, which constrains the lower limit of the reconfigurable bandwidth. The utilization of optimized multimode waveguides with Euler bends 18 , 37 or free-form curved 57 microrings can overcome this limitation while improving the narrowband filtering performance of the IMPF. Secondly, the combined effects of the non-ideal DC self-coupling coefficient and thermal crosstalk limit the adjustable range of the equivalent coupling coefficient. This limitation consequently reduces the maximum reconfigurable bandwidth and hinders arbitrary passband shaping. Increasing the interval between the two arms of the MZI coupler and the length of the two arms 23 , 26 , as well as carving grooves between microheaters 15 , can enlarge the adjustable range of the MZI coupling coefficient. Furthermore, the packaged chip in this work lacks TEC integration, leading to potential overvoltage and excessive power consumption. Future integration of TEC modules can address these limitations through active temperature stabilization and thermal management 12 , 26 . Summary In summary, we have proposed an intelligent configuration strategy based on the MOMS-HC algorithm that fully unleashes the hardware potential of IOPU. The MOMS-HC algorithm achieves self-configuration of IOPUs through global-local exploration, robust multi-objective equilibrium, efficient self-optimization, and environmental robustness empowerment. By employing this algorithm, we demonstrate an intelligently configured bandpass IMPF with four key capabilities: wideband center frequency tunability, flexible bandwidth reconfigurability, self-stabilization, and excellent channel equalization. The roll-off rate, RR, and SF of the IMPF reach up to 17.50 dB/GHz, 37.67 dB, and 0.78, respectively. Notably, with the MOMS-HO algorithm, the IMPF can operate stably for 3 hours, and its frequency stability is improved by 234 times. Furthermore, the MOMS-HO algorithm also enables the IMPF to recover to its target operating state quickly when all control ports experience significant disturbances simultaneously, which is rather challenging for manual manipulation. The MOMS-HC algorithm also enables dynamic compensation of intra-channel fading through reshaping the passband response of the IMPF, thus achieving channel equalization. This capability positions the intelligently configured IMPF as a viable alternative to conventional pre-equalization achieved by DSP. Overall, this work makes a solid step forward to applications of the IMPF in practical microwave systems. Methods Design of the MOMS-HC algorithm The MOMS-HC algorithm is a task-driven intelligent configuration framework with four stages. Its detailed procedure is as follows. In Stage I, the MOMS-HC algorithm first calculates and defines various properties of the IOPU programmatically. The FOM is then predefined for the subsequent configuration. Considering the global search capability and fast convergence of the PSO algorithm 43 , the MOMS-HC algorithm selects PSO for initial optimization in Stage II. The PSO algorithm randomly generates an initial state, where each particle is represented as a multi-dimensional vector corresponding to the multiple control voltages one-to-one. The fitness of each particle (i.e., the FOM) is defined as a function of the most important property of the IOPU to achieve rapid global optimization. In the absence of a packaged thermo-electric cooler (TEC), there is a strong correlation between the chip temperature and the magnitude of initial control voltages, which makes the configuration of the IOPU sensitive to the initial point. Therefore, the MOMS-HC algorithm iterates Stage II over multiple random initial points until the convergence criteria are met, and the optimal state is selected at the maximum FOM. The core of Stage III is an intelligent optimizer based on the standard GA 44 . Each individual consists of multiple genes, and the initial genes are obtained from Stage II. The population size should be set to a relatively large value, typically more than 20 times the number of parameters, to balance the global search capability and rapid convergence speed. Notably, the FOM is redefined to consider multiple objectives in this stage. Based on standard genetic operations, such as roulette-wheel selection, gene crossover, mutation, and environmental selection, the MOMS-HC algorithm can balance the weights of different objectives to avoid local optima in multi-objective optimization problems and maintain excellent search capability in uncertain environments. The global search of control voltage significantly exacerbates the overall temperature fluctuation of the chip due to electric power variations and crosstalk, leading to serious degradation of IOPU performance. Therefore, in Stage II and Stage III, the MOMS-HC algorithm incorporates a multi-level constraint mechanism on the voltage search space, effectively suppressing the overall temperature fluctuations. Following the completion of Stage III, Stage IV will be initiated to ensure sustained and stable operation of IOPU. Alternatively, iterative re-execution of Stage III enables dynamically adjusting the operational properties of IOPU, including bandwidth reconfiguration and center frequency tuning. Considering the environmental disturbances in practical operating environments, such as temperature fluctuations 46 , mechanical vibrations, and power supply noise 47 , 48 , the control voltage must be dynamically adjusted to compensate for the performance degradation of the IOPU. Therefore, in Stage IV, the MOMS-HC algorithm performs a local fine-tuning search using the SA algorithm 45 . Customizing the temperature decay mechanism in the SA algorithm can effectively limit the magnitude of each adjustment. Initially, the high annealing temperature permitted a relatively large voltage search range. Later, the annealing temperature rapidly reduces to restrict the voltage search range, thereby suppressing secondary performance fluctuations induced by voltage search and ensuring the performance consistency of the IOPU (see Supplementary Notes I for more details of the sub-algorithm). Throughout the whole execution of the MOMS-HC algorithm, the IOPU frequency response is obtained in real time by the detection module, such as a photodetector. In brief, PSO drives global parameter initialization, leveraging its swarm intelligence to accelerate convergence of IMPF toward the target response. GA executes precision configuration under multi-objective constraints by leveraging crossover and mutation operators. SA suppresses environmental disturbances through a dynamic fine-tuning search to ensure sustained and stable operation of IMPF. Precision and time cost of the configuration In the 3-hour stability test, when the MOMS-HC algorithm is applied to the IMPF, the center frequency of the IMPF drifts between +0.051 GHz and -0.030 GHz relative to the target value of 15.000 GHz, and the bandwidth varies between +0.02 GHz and -0.02 GHz relative to the target value of 0.90 GHz. These deviations are primarily attributed to the combined effects of optimization constraints imposed by the setting of the BOC and measurement errors induced by instrumentation limitations (0.010 GHz resolution of the VNA). In our experiment, the sensors (including OSA and VNA) and actuators (VSA) are accessed via serial and Ethernet ports using the TCP/IP protocol, resulting in a total communication speed of around 2 Hz. Without relying on any prior experience, the first configuration of IMPF can be finished within 2000 s (less than 40 iterations, with 100 calculations per iteration), while the response time of the TiN heater is approximately 10 μs 58 . For center frequency tuning with the same bandwidth or bandwidth reconfiguration at the same center frequency, the MOMS-HC algorithm requires only a single execution of Stage II. For example, after configuring the IMPF to a specific response (e.g., the center frequency is 25 GHz and the bandwidth is 0.66 GHz), the algorithm can directly use the current applied voltages as the starting point and repeat Stage III to reconfigure the IMPF. This mechanism significantly reduces the time cost. Therefore, the single reconfiguration of the IMPF, including bandwidth reconfiguration and center frequency tuning, can be completed within 400 s (less than 8 iterations). If the external driver speed is increased to make the communication time cost negligible, the first configuration and the single reconfiguration of IMPF can be completed in 8 ms and 40 ms, respectively. Therefore, to further reduce the time cost, techniques such as optoelectronic co-integration 59 and frequency-to-time mapping 60 can be used in the future. Declarations Supporting Information Supporting Information is available from the Online Library. Data Availability The data that support the findings of this study are available from the corresponding author upon request. Conflict of interest The authors declare no competing interests. Contributions Y.T.S. and X.L.Z. conceived the idea. Y.T.S., H.L.Z., and Y. Y. designed the MOMS-HCS algorithm and engineered the related experiments. Y.F.L. and Y. Y. designed the device as a demonstration platform. Y.T.S. performed the experiments and recorded and analyzed the data. K.X.C. and M.M.D. assisted with the measurements. Y.T.S. and Y.Y. prepared the manuscript. Y. Y., F.Z. Z. and X.L.Z. supervised the project. All authors participated in revising the manuscript and approved the submission. Acknowledgements This project was supported by the National Key Research and Development Program of China (2024YFF0728000), the Program for HUST Academic Frontier Youth Team (2018QYTD08), and the Project of Key Laboratory of Radar Imaging and Microwave Photonics (Nanjing University of Aeronautics and Astronautics), Ministry of Education (NJ20230001). References Ilderem, V. The technology underpinning 5G. Nat. Electron. 3 , 5–6 (2020). Akyildiz, I. F., Kak, A. & Nie, S. 6G and Beyond: The future of wireless communications systems. IEEE Access 8 , 133995–134030 (2020). Liu, W. et al. A fully reconfigurable photonic integrated signal processor. Nat. Photonics 10 , 190–195 (2016). Marpaung, D., Yao, J. & Capmany, J. Integrated microwave photonics. Nat. Photonics 13 , 80–90 (2019). Yao, J. Microwave photonic systems. J. Lightwave Technol. 40 , 6595–6607 (2022). Capmany, J. & Novak, D. Microwave photonics combines two worlds. Nat. Photonics 1 , 319–330 (2007). Marpaung, D. et al. Integrated microwave photonics. Laser Photonics Rev. 7 , 506–538 (2013). Liu, Y. et al. Integrated microwave photonic filters. Adv. Opt. Photon. 12 , 485 (2020). Fandiño, J. S.et al. A monolithic integrated photonic microwave filter. Nat. Photonics 11 , 124–129 (2017). Wang, L., Wang, X. & Pan, S. Microwave photonics empowered integrated sensing and communication for 6G. IEEE Trans. Microwave Theory Techn. 1–21 (2025). Daulay, O. et al. Ultrahigh dynamic range and low noise figure programmable integrated microwave photonic filter. Nat. Commun. 13 , 7798 (2022). Velamuri, A. et al. Reconfigurable multiband microwave filters using programmable photonic integrated circuit. Optica Open. Preprint (2024). Tao, Z. et al. Highly reconfigurable silicon integrated microwave photonic filter towards next-generation wireless communication. Photon. Res. 11 , 682 (2023). Tao, Y. et al. Hybrid-integrated high-performance microwave photonic filter with switchable response. Photon. Res. 9 , 1569 (2021). Xu, L. et al. Silicon-on-insulator-based microwave photonic filter with widely adjustable bandwidth. Photon. Res. 7 , 110 (2019). Liu, X. et al. Silicon-on-insulator-based microwave photonic filter with narrowband and ultrahigh peak rejection. Opt. Lett. 43 , 1359–1362 (2018). Song, S. et al. Tunable single-passband microwave photonic filter based on integrated optical double notch filter. J. Lightwave Technol. 36 , 4557–4564 (2018). Zhang, L. et al. Ultralow-loss silicon photonics beyond the singlemode regime. Laser Photonics Rev. 16 , 2100292 (2022). Porzi, C. et al. Silicon-on-insulator microwave photonic filter with widely tunable and reconfigurable flat-top bandpass functionality. J. Lightwave Technol. 40 , 6666–6675 (2022). Zhuang, L., Roeloffzen, C. G. H., Hoekman, M., Boller, K.-J. & Lowery, A. J. Programmable photonic signal processor chip for radiofrequency applications. Optica , 2 , 854–859 (2015). Liu, Y. et al. All-optimized integrated RF photonic notch filter. Opt. Lett. 42 , 4631–4634 (2017). Jiang, H., Yan, L. & Marpaung, D. Chip-based arbitrary radio-frequency photonic filter with algorithm-driven reconfigurable resolution. Opt. Lett. 43 , 415 (2018). Cheng, W. et al. Tunable bandpass microwave photonic filter with largely reconfigurable bandwidth and steep shape factor based on cascaded silicon nitride micro-ring resonators. Opt. Express 31 , 25648 (2023). Guzzon, R. S. et al. Integrated InP-InGaAsP tunable coupled ring optical bandpass filters with zero insertion loss. Opt. Express 19 , 7816–7826 (2011). Norberg, E. et al. A. Programmable photonic microwave filters monolithically integrated in InP–InGaAsP. J. Lightwave Technol. 29 , 1611–1619 (2011). Wei, C. et al. Programmable multifunctional integrated microwave photonic circuit on thin-film lithium niobate. Nat. Commun. 16 , 2281 (2025). Marpaung, D. et al. Low-power, chip-based stimulated Brillouin scattering microwave photonic filter with ultrahigh selectivity. Optica 2 , 76–83 (2015). Liu, Y. et al. Integration of Brillouin and passive circuits for enhanced radio-frequency photonic filtering. APL Photonics 4 , 106103 (2019). Garrett, M. et al. Multi-band and frequency-agile chip-based RF photonic filter for ultra-deep interference rejection. J. Lightwave Technol. 40 , 1672–1680 (2022). Shi, F. et al. A microwave photonic channelized receiver with self-interference cancellation. J. Lightwave Technol. 41 , 627–636 (2023). Yu, Y. et al. ANN-assisted high-resolution and large dynamic range temperature sensor based on simultaneous microwave photonic and optical measurement. IEEE Sensors Journal 23 , 1105–1114 (2023). Wang, L. et al. Parity–time-symmetric optoelectronic oscillator based on stimulated brillouin scattering. ACS Photonics 10 , 2308–2315 (2023). Pan, Z. et al. Frequency-stabilized optoelectronic oscillator with simultaneous wideband tunability. IEEE Trans. Microwave Theory Techn. 1–8 (2024). Tripathi, S. et al. Millimeter-wave and terahertz spectrum for 6g wireless. 6G Mobile Wireless Networks. 83–121 (2021). Ge, Z. et al. Millimeter-wave wide-band bandpass filter in CMOS technology using a two-layered highpass-type approach with embedded upper stopband. IEEE Trans. Circuits Syst. II: Exp.s Briefs 68 , 1586–1590 (2021). Liu, L. et al. Narrow passband tunable optical filter based on silicon high-Q rings assisted mzi structure. J. Lightwave Technol. 42 , 2049–2056 (2024). Yan, H. et al. Wideband-tunable on-chip microwave photonic filter with ultrahigh-Q U-bend-Mach–Zehnder-interferometer-coupled microring resonators. Laser Photonics Rev. 17 , 2300347 (2023). Liu, Y. et al. Tunable and reconfigurable microwave photonic bandpass filter based on cascaded silicon microring resonators. J. Lightwave Technol. 40 , 4655–4662 (2022). Liu, Y. et al. Reconfigurable microwave photonic bandpass filter based on CROW. J. Lightwave Technol. 42 , 1597–1604 (2024). Liu, Y. et al. Reconfigurable bandpass microwave photonic filter based on cascaded second-order CROWs. J. Lightwave Technol. 42 , 5599-5605 (2024). Zhang, L. et al. Cascading second-order microring resonators for a box-like filter response. J. Lightwave Technol. 35 , 5347–5360 (2017). Pérez-López et al. Multipurpose self-configuration of programmable photonic circuits. Nat. Commun. 11 , 6359 (2020). Kennedy, J. & Eberhart, R. Particle swarm optimization. In Proceedings of ICNN’95 - International Conference on Neural Networks 4 , 1942–1948 (IEEE, 1995). Holland, J. H. Genetic algorithms. Scientific American 267 , 66–73 (1992). S. Kirkpatrick et al. Optimization by simulated annealing. Science 220 ,671-680(1983). Padmaraju, K. & Bergman, K. Resolving the thermal challenges for silicon microring resonator devices. Nanophotonics 3 , 269–281 (2014). Sheikholeslami, A. Power supply noise [Circuit Intuitions]. IEEE Solid-State Circuits Mag. 12 , 15–17 (2020). Bohannon, E. et al. Passive and active reduction techniques for on-chip high-frequency digital power supply noise. IEEE Transactions on Very Large Scale Integration (VLSI) Systems 18 , 157–161 (2010). Zhang, L. et al. Beyond 100-Gb/s transmission over 80-km SMF using direct-detection SSB-DMT at C-band. J. Lightwave Technol. 34 , 723–729 (2016). Bo, T. et al. Optical single-sideband transmitters. J. Lightwave Technol. 41 , 1163–1174 (2023). Chen, X. et al. Kramers–Kronig receivers for 100-km datacenter interconnects. J. Lightwave Technol. 36 , 79–89 (2018). Tang, X. et al. Digital pre- and post-equalization for C-band 112-Gb/s PAM4 short-reach transport systems. J. Lightwave Technol. 38 , 4683–4690 (2020). Ni, W. et al. LUT enabled low-complexity electrical dispersion pre-compensation scheme for IM-DD systems. Opt. Lett. 49 , 6417–6420 (2024). Choi, H.-W., Choi, S. & Kim, C.-Y. A CMOS band-pass low noise amplifier with excellent gain flatness for mm-Wave 5G communications. In IEEE/MTT-S International Microwave Symposium (IMS) 329–332 (IEEE, 2020). Sewiolo, B., Fischer, G. & Weigel, R. A 30 GHz variable gain amplifier with high output voltage swing for ultra-wideband radar. IEEE Microw. Wireless Compon. Lett . 19 , 590–592 (2009). Zhao, W. et al. Ultracompact silicon on-chip polarization controller. Photon. Res. 12 , 183 (2024). Sun, S. et al. Inverse design of ultra-compact multimode waveguide bends based on the free-form curves. Laser Photonics Rev. 15 , 2100162 (2021). Cheng, J. et al. Multimodal deep learning using on-chip diffractive optics with in situ training capability. Nat. Commun. 15 , 6189 (2024). Zhang, W. et al. A system-on-chip microwave photonic processor solves dynamic RF interference in real time with picosecond latency. Light Sci. Appl. 13 , 14 (2024). Shi, T., Chen, Y. & Yao, J. Seamlessly merging radar ranging and imaging, wireless communications, and spectrum sensing for 6G empowered by microwave photonics. Commun. Eng. 3 , 130 (2024). Additional Declarations There is no conflict of interest Supplementary Files Supplementaryinformation.docx Supplementary Information: Intelligent configuration of integrated microwave photonic filter with programmable response and self-stabilization Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-6669261\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":true,\"archivedVersions\":[],\"articleType\":\"Article\",\"associatedPublications\":[],\"authors\":[{\"id\":456873835,\"identity\":\"84bcdac3-aa67-448a-a5fa-88fc7e9dc9c8\",\"order_by\":0,\"name\":\"Yuan Yu\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzklEQVRIiWNgGAWjYBADOQMIzUy8FmOgFsYGkrQkbiBai8GN5GcPGNsOp2+XSH/+gKHCOrGB/ewBAlrSzA2AWnJ3zsgxbGA4k57YwJOXQEBLgpkESMuGGzmMDUBGYoMEjwEBLenfQFrSgYyHDYz/iNKSA7YlAWidYQNjAxFaJM+8KZNgOJduuOHMG8MZCcfSjdt4cvBr4Tuevk2Cocxa3uB4+oMPH2qsZfvZz+DXonAAGBd/2aC8BCBmw60YAuQbQOQfQspGwSgYBaNgRAMAnwJIOgECWv8AAAAASUVORK5CYII=\",\"orcid\":\"https://orcid.org/0000-0001-7400-5064\",\"institution\":\"Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Yuan\",\"middleName\":\"\",\"lastName\":\"Yu\",\"suffix\":\"\"},{\"id\":456873836,\"identity\":\"2eed6785-c433-436e-885e-a0939c3f628c\",\"order_by\":1,\"name\":\"Yutong Shi\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Huazhong University of Science and Technology\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yutong\",\"middleName\":\"\",\"lastName\":\"Shi\",\"suffix\":\"\"},{\"id\":456873837,\"identity\":\"c297245d-c6f8-46b7-a2b3-11718c4d750f\",\"order_by\":2,\"name\":\"Yifan Liu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yifan\",\"middleName\":\"\",\"lastName\":\"Liu\",\"suffix\":\"\"},{\"id\":456873838,\"identity\":\"63e2c9a2-59b7-4c76-91da-21e0e29b9050\",\"order_by\":3,\"name\":\"Kaixiang Cao\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Kaixiang\",\"middleName\":\"\",\"lastName\":\"Cao\",\"suffix\":\"\"},{\"id\":456873839,\"identity\":\"c70eb8f7-036c-45f3-b9ef-f19a664c2ee9\",\"order_by\":4,\"name\":\"Mengmeng Deng\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Mengmeng\",\"middleName\":\"\",\"lastName\":\"Deng\",\"suffix\":\"\"},{\"id\":456873840,\"identity\":\"5588e5fd-5706-48c6-86a4-94dd42ee01a9\",\"order_by\":5,\"name\":\"Fangzheng Zhang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"National Key Laboratory of Microwave Photonics, Nanjing University of Aeronautics and Astronautics\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Fangzheng\",\"middleName\":\"\",\"lastName\":\"Zhang\",\"suffix\":\"\"},{\"id\":456873841,\"identity\":\"ba044f9c-3dc6-4e45-af01-0d52363c5f8b\",\"order_by\":6,\"name\":\"Hailong Zhou\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0002-1878-4291\",\"institution\":\"Huazhong University of Science and Technology\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Hailong\",\"middleName\":\"\",\"lastName\":\"Zhou\",\"suffix\":\"\"},{\"id\":456873842,\"identity\":\"8719d849-511e-4a67-9393-34bb95f3d7e4\",\"order_by\":7,\"name\":\"Xinliang Zhang\",\"email\":\"\",\"orcid\":\"https://orcid.org/0000-0001-8513-3328\",\"institution\":\"Huazhong University of Science and Technology\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xinliang\",\"middleName\":\"\",\"lastName\":\"Zhang\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-05-15 06:10:38\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-6669261/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-6669261/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":83355356,\"identity\":\"6bcedd3a-adbc-4fe7-a0c2-2249c3317b35\",\"added_by\":\"auto\",\"created_at\":\"2025-05-23 15:04:06\",\"extension\":\"jpeg\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":3362237,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003ePrinciple of our proposed intelligent configuration approach and details of the SOI-based device.\\u003c/strong\\u003e \\u003cstrong\\u003ea \\u003c/strong\\u003eThe flowchart of the proposed MOMS-HO algorithm. \\u003cstrong\\u003eb\\u003c/strong\\u003e The principle of the integrated microwave photonic signal processing based on intelligent configuration. \\u003cstrong\\u003ec\\u003c/strong\\u003e Schematic diagram of the designed 4th-order CROW to be implemented as the IMPF. \\u003cstrong\\u003ed\\u003c/strong\\u003e Schematic diagram of the MRR, in which the 500 nm-wide single-mode waveguide is connected to the 2-μm-wide multimode waveguide via a long linear adiabatic taper. \\u003cstrong\\u003ee\\u003c/strong\\u003e Schematic diagram of the MZI coupler, where r and k are the self-coupling and cross-coupling coefficients of DC, respectively. \\u003cstrong\\u003ef\\u003c/strong\\u003e Schematic diagram of the ridge waveguide with a ridge height of 130 nm. \\u003cstrong\\u003eg\\u003c/strong\\u003e The microscope false-color image of the fabricated device. \\u003cstrong\\u003eh\\u003c/strong\\u003e Photograph of the packaged chip. The inset shows the zoomed-in view of the fabricated chip.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage1.jpeg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/c7f8db02e4f12ff29b7b93f1.jpeg\"},{\"id\":83355382,\"identity\":\"4c00079c-d32a-4e17-ac46-4ef3921aa834\",\"added_by\":\"auto\",\"created_at\":\"2025-05-23 15:04:08\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":268769,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eSchematic diagram of the experimental setups.\\u003c/strong\\u003eExperimental setup for configuring the IMPF based on our proposed MOMS-HC algorithm. LS laser source, OC optical coupler, PC polarization controller, VNA vector network analyzer, EA electrical amplifier, PM phase modulator, OBPF optical bandpass filter, EDFA erbium-doped fiber amplifier, DUT device under test, VSA voltage source array, PD photodetector.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/3d77c712d207b83387d6f800.png\"},{\"id\":83356481,\"identity\":\"30c15144-c9a7-4363-8410-ae29109243a0\",\"added_by\":\"auto\",\"created_at\":\"2025-05-23 15:12:07\",\"extension\":\"jpeg\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":2243782,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eIntelligent configuration of the IMPF. a\\u003c/strong\\u003e The variation of the normalized FOM as the number of iterations increases during the intelligent configuration process. \\u003cstrong\\u003eb-e\\u003c/strong\\u003e Voltage distribution applied to all the microheaters in the DUT at the \\u003cstrong\\u003eb\\u003c/strong\\u003e 8th, \\u003cstrong\\u003ec\\u003c/strong\\u003e 19th, \\u003cstrong\\u003ed\\u003c/strong\\u003e31st, and \\u003cstrong\\u003ee \\u003c/strong\\u003e42nd iterations, respectively. \\u003cstrong\\u003ef-i\\u003c/strong\\u003e Amplitude frequency response of the IMPF during the intelligent configuration process at the \\u003cstrong\\u003ef\\u003c/strong\\u003e 8th, \\u003cstrong\\u003eg\\u003c/strong\\u003e 19th, \\u003cstrong\\u003eh\\u003c/strong\\u003e31st, and \\u003cstrong\\u003ei\\u003c/strong\\u003e 42nd iterations, respectively.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage3.jpeg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/490e17d2472579b7f758a07a.jpeg\"},{\"id\":83355360,\"identity\":\"1dcc725d-6090-49ab-8fd3-f9f726d7f834\",\"added_by\":\"auto\",\"created_at\":\"2025-05-23 15:04:07\",\"extension\":\"jpeg\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1700169,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eTuning and reconfiguring the IMPF based on the proposed MOMS-HC algorithm.\\u003c/strong\\u003e \\u003cstrong\\u003ea\\u003c/strong\\u003e Tuning the center frequency of IMPF with different bandwidths, \\u003cstrong\\u003eI\\u003c/strong\\u003e 0.66GHz, \\u003cstrong\\u003eII\\u003c/strong\\u003e 1.85GHz, and \\u003cstrong\\u003eIII\\u003c/strong\\u003e 3.15 GHz. \\u003cstrong\\u003eb-d\\u003c/strong\\u003e Characterizing the consistency of the various properties during the frequency tuning process, \\u003cstrong\\u003eb\\u003c/strong\\u003e Bandwidth, \\u003cstrong\\u003ec\\u003c/strong\\u003e RR, and \\u003cstrong\\u003ed\\u003c/strong\\u003e SF. The horizontal axis represents three bandwidth labels, and the vertical axis represents the center frequency labels. \\u003cstrong\\u003ee\\u003c/strong\\u003e Reconfiguring the bandwidth of IMPF from 0.66 to 4.15 GHz. \\u003cstrong\\u003ef\\u003c/strong\\u003e Measured SF and RR with different FWHM bandwidths. \\u003cstrong\\u003eg\\u003c/strong\\u003e Measured in-band ripple with different FWHM bandwidths. The curves show polynomial fitting results. The dark shaded area indicates the 95% confidence interval, while the light shaded area represents the 95% prediction interval.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage4.jpeg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/807add00e2d5f1c1ebce8114.jpeg\"},{\"id\":83355361,\"identity\":\"5ba42d74-7055-4523-adcc-c8c9e73489a1\",\"added_by\":\"auto\",\"created_at\":\"2025-05-23 15:04:07\",\"extension\":\"jpeg\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1963577,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eSelf-stabilization and self-recovery of the intelligently configured IMPF.\\u003c/strong\\u003e \\u003cstrong\\u003ea\\u003c/strong\\u003eThe “eye-like” diagram showing the amplitude frequency response of the IMPF without MOMS-HC-based configuration over 3 hours. The gradient color scale from light green to dark green indicates the direction of the center frequency drift. \\u003cstrong\\u003eb\\u003c/strong\\u003e The “eye-like” diagram showing the amplitude frequency response of the IMPF with MOMS-HC-based configuration over 3 hours. \\u003cstrong\\u003ec\\u003c/strong\\u003e Real-time dynamic adjustment of the nine voltages applied to the microheaters during IMPF operation with MOMS-HC on. \\u003cstrong\\u003ed–g\\u003c/strong\\u003e Time-dependentvariations of \\u003cstrong\\u003ed\\u003c/strong\\u003e SF, \\u003cstrong\\u003ee\\u003c/strong\\u003e center frequency offset relative to the target value of 15.000 GHz, \\u003cstrong\\u003ef\\u003c/strong\\u003e bandwidth, and \\u003cstrong\\u003eg\\u003c/strong\\u003e RR during the 3-hour stability test. \\u003cstrong\\u003eh\\u003c/strong\\u003e Frequency response degradation of the IMPF when all nine control ports are disturbed simultaneously by random voltage biases. The amplitude of the random voltage biases is 200 mV. \\u003cstrong\\u003ei\\u003c/strong\\u003e The recorded amplitude frequency responses of the IMPF every minute within 5 minutes after experiencing the disturbance. \\u003cstrong\\u003ej\\u003c/strong\\u003e The “eye-like” diagram showing the amplitude frequency responses of the IMPF in the subsequent 22 minutes after recovery, with the RR and roll-off rate above 30 dB and 15 dB/GHz, respectively. \\u003cstrong\\u003ek-n\\u003c/strong\\u003eTime-dependent variations of \\u003cstrong\\u003ek\\u003c/strong\\u003e SF, \\u003cstrong\\u003el\\u003c/strong\\u003e center frequency offset, \\u003cstrong\\u003em\\u003c/strong\\u003ebandwidth, and \\u003cstrong\\u003en\\u003c/strong\\u003e RR. The red dots and curves in the figure indicate the recovery process.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage5.jpeg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/19a256a314156ac720697c54.jpeg\"},{\"id\":83355367,\"identity\":\"ae43efaa-d1f2-4af2-965b-8f8b1eb934c4\",\"added_by\":\"auto\",\"created_at\":\"2025-05-23 15:04:07\",\"extension\":\"jpeg\",\"order_by\":6,\"title\":\"Figure 6\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1768206,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e\\u003cstrong\\u003eChannel equalization performance.\\u003c/strong\\u003e \\u003cstrong\\u003ea\\u003c/strong\\u003e The amplitude frequency responses of a general microwave photonic link without the IMPF based on the 4th-order CROW. \\u003cstrong\\u003eb-e\\u003c/strong\\u003e Channel equalization results when the center frequency is \\u003cstrong\\u003eb \\u003c/strong\\u003e5.5, \\u003cstrong\\u003ec \\u003c/strong\\u003e22.0, \\u003cstrong\\u003ed\\u003c/strong\\u003e 27.0, and \\u003cstrong\\u003ee\\u003c/strong\\u003e 32.5 GHz, respectively. Green and red shaded areas denote the passband and stopband, respectively.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"floatimage6.jpeg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/3b4b3ca2a0890e0b035871bd.jpeg\"},{\"id\":83697230,\"identity\":\"55b64328-6fcb-4ba5-b176-326f326484fc\",\"added_by\":\"auto\",\"created_at\":\"2025-05-31 04:53:39\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":12272448,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/a98680f7-14fd-4b7b-9968-41cb6a4e3aee.pdf\"},{\"id\":83355358,\"identity\":\"3b0c1ded-95a5-48eb-bffb-a09fa80c348d\",\"added_by\":\"auto\",\"created_at\":\"2025-05-23 15:04:07\",\"extension\":\"docx\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":1006847,\"visible\":true,\"origin\":\"\",\"legend\":\"Supplementary Information: Intelligent configuration of integrated microwave photonic filter with programmable response and self-stabilization\",\"description\":\"\",\"filename\":\"Supplementaryinformation.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-6669261/v1/ba3f5fde9d0fd3b1bd4dcb2e.docx\"}],\"financialInterests\":\"There is no conflict of interest\",\"formattedTitle\":\"Intelligent configuration of integrated microwave photonic filter with programmable response and self-stabilization\",\"fulltext\":[{\"header\":\"Introduction\",\"content\":\"\\u003cp\\u003eTraditional radio frequency (RF) electronic technologies, constrained by limited sampling rates and component bandwidths, increasingly fail to meet the demands of next-generation communication systems, which require large bandwidth, flexible reconfigurability, excellent programmability, low cost, low consumption, and electromagnetic immunity\\u003csup\\u003e\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e\\u003c/sup\\u003e. In contrast, microwave photonics offers a breakthrough pathway by leveraging its inherent advantages, including ultra-wide bandwidth, high flexibility, low loss, and electromagnetic immunity\\u003csup\\u003e\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e\\u003c/sup\\u003e. As fundamental components, microwave photonic filters are used to select target signals from background noise or reject disturbing noise from signals. Advances in photonic integrated circuit (PIC) technology have driven the growing adoption of integrated microwave photonic filters (IMPFs)\\u003csup\\u003e\\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e11\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e\\u003c/sup\\u003e, which offer significant advantages in terms of size, weight, power consumption, and cost\\u003csup\\u003e\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e\\u003c/sup\\u003e. Currently, IMPFs have been demonstrated on various platforms, including silicon-on-insulator (SOI)\\u003csup\\u003e\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e16\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e18\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e\\u003c/sup\\u003e, silicon nitride\\u003csup\\u003e\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e22\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e23\\u003c/span\\u003e\\u003c/sup\\u003e, indium phosphide\\u003csup\\u003e\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e24\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e25\\u003c/span\\u003e\\u003c/sup\\u003e, lithium niobate\\u003csup\\u003e\\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e26\\u003c/span\\u003e\\u003c/sup\\u003e, and chalcogenide glass\\u003csup\\u003e\\u003cspan citationid=\\\"CR27\\\" class=\\\"CitationRef\\\"\\u003e27\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR28\\\" class=\\\"CitationRef\\\"\\u003e28\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e29\\u003c/span\\u003e\\u003c/sup\\u003e. Compared with electronic filters, IMPFs exhibit superior tunability and reconfigurability, which drive their broad application prospects in wireless communications\\u003csup\\u003e\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e29\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR30\\\" class=\\\"CitationRef\\\"\\u003e30\\u003c/span\\u003e\\u003c/sup\\u003e, sensor networks\\u003csup\\u003e\\u003cspan citationid=\\\"CR31\\\" class=\\\"CitationRef\\\"\\u003e31\\u003c/span\\u003e\\u003c/sup\\u003e, and radar systems\\u003csup\\u003e\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e\\u003c/sup\\u003e. IMPFs require tailored configurations for various scenarios. For example, optoelectronic oscillators (OEOs) require IMPFs with a narrow bandwidth\\u003csup\\u003e\\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e32\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR33\\\" class=\\\"CitationRef\\\"\\u003e33\\u003c/span\\u003e\\u003c/sup\\u003e, and radar systems demand IMPFs with a high rejection ratio (RR), wideband operation, and flexible tunability\\u003csup\\u003e\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e34\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR35\\\" class=\\\"CitationRef\\\"\\u003e35\\u003c/span\\u003e\\u003c/sup\\u003e. With the development of integrating multidimensional functions into a single system, reconfiguring IMPF responses according to diverse requirements significantly enhances economic efficiency.\\u003c/p\\u003e \\u003cp\\u003eAlthough the flexibility of IMPFs has been demonstrated\\u003csup\\u003e\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e26\\u003c/span\\u003e\\u003c/sup\\u003e, most reported reconfigurable IMPFs primarily depend on cumbersome manual manipulation for frequency tuning and bandwidth reconfiguration, which presents significant challenges for achieving real-time, dynamic, and self-adaptive configuration for complex applications. Many reconfigurable IMPFs utilizing various techniques have been demonstrated, such as Mach-Zehnder interferometer (MZI)-assisted microring resonators (MRRs)\\u003csup\\u003e\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR36\\\" class=\\\"CitationRef\\\"\\u003e36\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR37\\\" class=\\\"CitationRef\\\"\\u003e37\\u003c/span\\u003e\\u003c/sup\\u003e, cascaded MRRs\\u003csup\\u003e\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e26\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e29\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e38\\u003c/span\\u003e\\u003c/sup\\u003e, coupled resonator optical waveguides (CROWs)\\u003csup\\u003e\\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR39\\\" class=\\\"CitationRef\\\"\\u003e39\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR40\\\" class=\\\"CitationRef\\\"\\u003e40\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR41\\\" class=\\\"CitationRef\\\"\\u003e41\\u003c/span\\u003e\\u003c/sup\\u003e, and programmable two-dimensional (2D) mesh network topologies\\u003csup\\u003e\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e,\\u003cspan citationid=\\\"CR42\\\" class=\\\"CitationRef\\\"\\u003e42\\u003c/span\\u003e\\u003c/sup\\u003e. These IMPFs exhibit superior performance in key properties, including RR, tunable range of center frequency, reconfigurable range of full width at half-maximum (FWHM) bandwidth, shape factor (SF), and multifunctional integration. Nonetheless, these designs often face performance compromises due to insufficient configuration, making it difficult to achieve synergy of all properties. Furthermore, their sustained and stable operation and resistance to random environmental disturbances have not been sufficiently addressed. Overall, traditional manual manipulation and offline design methods fail to meet the requirements for synergy of all properties, dynamic configuration, and sustained and stable operation of IMPFs, which significantly hinder their application prospects. Therefore, developing a supporting intelligent configuration framework is essential for advancing IMPF toward practical wireless applications.\\u003c/p\\u003e \\u003cp\\u003eIn this work, we propose a universal multi-objective and multi-stage hybrid collaboration (MOMS-HC) algorithm for the intelligent configuration of IMPFs. The algorithm innovatively leverages the complementary strengths of particle swarm optimization (PSO)\\u003csup\\u003e\\u003cspan citationid=\\\"CR43\\\" class=\\\"CitationRef\\\"\\u003e43\\u003c/span\\u003e\\u003c/sup\\u003e, genetic algorithm (GA)\\u003csup\\u003e\\u003cspan citationid=\\\"CR44\\\" class=\\\"CitationRef\\\"\\u003e44\\u003c/span\\u003e\\u003c/sup\\u003e, and simulated annealing (SA)\\u003csup\\u003e\\u003cspan citationid=\\\"CR45\\\" class=\\\"CitationRef\\\"\\u003e45\\u003c/span\\u003e\\u003c/sup\\u003e through a staged collaboration mechanism. Specifically, the task-driven specialization of each sub-algorithm at different stages governs the collaboration, enabling multi-algorithm hybrid execution and complementary strengths integration. The algorithm integrates global-local exploration, robust multi-objective equilibrium, efficient self-optimization, and environmental robustness empowerment, and fully unleashes the hardware potential of IMPF. Utilizing the MOMS-HC algorithm, we implement, to the best of our knowledge, the first intelligently configured bandpass IMPF, whose hardware is based on a 4th-order coupled resonator optical waveguide (CROW) on a commercial SOI wafer. Experimental results demonstrate that the IMPF achieves unprecedented synergy of almost all properties, including a tuning range of 2\\u0026ndash;48 GHz, an FWHM bandwidth reconfiguration range of 0.66\\u0026ndash;4.15 GHz, a roll-off rate of up to 17.50 dB/GHz, an in-band ripple of below 0.91 dB, an RR of up to 37.67 dB, and an SF of up to 0.78. Notably, with the MOMS-HO algorithm, the maximum center frequency drift of IMPF is only 0.051 GHz in 3 hours. Compared with the free-running case, the frequency stability is enhanced by 234 times. Results also show that the MOMS-HO algorithm enables the IMPF to quickly recover to its target operating state when significant disturbances occur. Additionally, IMPF-based equalization of the frequency-dependent fading in a microwave photonic link based on the MOMS-HO algorithm is demonstrated, and up to 2.42 dB improvement for intra-channel fading is obtained by reshaping the passband of IMPF. The proposed intelligent configuration approach provides critical support for practical applications of IMPFs and is expected to accelerate the deployment of high-performance integrated optical devices.\\u003c/p\\u003e\"},{\"header\":\"Results\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003ePrinciple of the MOMS-HC algorithm and device\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe integrated microwave photonic signal processing is fundamentally implemented based on the optical-to-electrical mapping technique. Therefore, the integrated optical processing unit (IOPU) must be carefully designed and configured to achieve high-performance signal processing. As a result, we first consider the general problem of configuring numerous electrically controlled adjustable parameters to ensure that the IOPU converges to the desired response. The challenges of this problem include the vast parameter space resulting from complex structures and the inter-coupling between multiple adjustable parameters. Therefore, we propose to consider the utilization of advanced machine learning algorithms to transform the configuration problem of IOPU into an optimization problem. The optimization objective and variables are the required frequency response and voltage applied to the adjustable parameters. However, to realize an IOPU with superior overall performance, we must prioritize the multi-objective equilibrium of the configuration algorithm.\\u003c/p\\u003e\\n\\u003cp\\u003eBased on these ideas, we propose the MOMS-HC algorithm. The execution flowchart of the MOMS-HC algorithm, which consists of four stages, is shown in Fig.\\u0026nbsp;1a.\\u0026nbsp;The \\u0026ldquo;Preprocessing Stage\\u0026rdquo; (Stage I) first programmatically defines various properties of the IOPU and a figure of merit (FOM) to guide subsequent optimization. The \\u0026ldquo;Global Search Stage\\u0026rdquo; (Stage II) executes PSO optimization, where swarm intelligence drives rapid global optimization using a single critical property of the IOPU as the FOM. The \\u0026ldquo;Dynamic Configuration Stage\\u0026rdquo; (Stage III) employs GA, which\\u0026nbsp;specializes in precision\\u0026nbsp;optimization\\u0026nbsp;under multi-objective constraints. The GA performs collaborative multi-objective optimization\\u0026nbsp;with a redefinable FOM. It enables IMPF to achieve synergy of multiple properties due to\\u0026nbsp;leverage roulette selection, gene crossover, and mutation operations, while a flexible population management mechanism balances search efficiency and robustness. In the \\u0026ldquo;Self-stabilization Stage\\u0026rdquo; (Stage IV), the SA algorithm is employed to suppress performance degradation caused by environmental disturbances, such as\\u0026nbsp;temperature fluctuations\\u003csup\\u003e46\\u003c/sup\\u003e,\\u0026nbsp;mechanical vibrations,\\u0026nbsp;and\\u0026nbsp;power supply noise\\u003csup\\u003e47\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e48\\u003c/sup\\u003e. Through a customized temperature decay mechanism, SA enables rapid local optimization while progressively constraining the amplitude of voltage adjustments. This mechanism suppresses secondary performance fluctuations induced by voltage searches, ensuring sustained consistency and self-stabilization of the IOPU under complex operating conditions (see Methods for more details of the MOMS-HC algorithm).\\u003c/p\\u003e\\n\\u003cp\\u003eThe principle of the integrated microwave photonic signal processing based on intelligent configuration is shown in Fig.\\u0026nbsp;1b. The continuous wave (CW) light emitted by a laser source (LS) is modulated by microwave signals via a modulator. Then, the modulated microwave photonic signal is processed by IOPU. Following processing, the microwave signal is extracted by a photodetector (PD) through optical-electrical (O/E) conversion. The resulting microwave signal is fed back to the intelligent configuration module, which subsequently calculates the control signals based on the configuration targets. After calculation, the output control signals are applied to the IOPU, thereby enabling dynamic reconfiguration of the IOPU.\\u003c/p\\u003e\\n\\u003cp\\u003eConventionally, dynamic configuration of high-order CROW is quite challenging because of inherent structural complexity and strong inter-coupling between adjustable parameters in compact size\\u003csup\\u003e15\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e41\\u003c/sup\\u003e. Meanwhile, filtering is one of the most classical and commonly used signal processing techniques. Therefore, the implementation of intelligent configuration of IMPF based on 4th-order CROW\\u003csup\\u003e39\\u003c/sup\\u003e is utilized to demonstrate the superior performance of our proposed MOMS-HC algorithm. The 4th-order CROW is composed of four identical MRRs integrated with 14 titanium nitride (TiN) microheaters, as shown in Fig. 1c. The detailed parameters of each MRR are shown in Fig. 1d. The MRR coupling regions are implemented as MZI couplers consisting of two identical directional couplers (DCs) and two phase shifters. These couplers are designed as\\u003cem\\u003e\\u0026nbsp;\\u003c/em\\u003etwo single-mode waveguides with three 60\\u0026deg; arcs and a bending radius of 20 \\u0026mu;m, as illustrated in Fig. 1e. The device employs ridge waveguides with a ridge height of 130 nm to reduce the waveguide loss, as shown in Fig. 1f. The values of r and k should satisfy \\u0026nbsp; in order to obtain the maximum tuning range of the equivalent coupling coefficients\\u003csup\\u003e38\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e40\\u003c/sup\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003eFig.\\u0026nbsp;1g shows the false-color image of the fabricated 4th-order CROW serving as the verification device. The device is fabricated based on an SOI wafer with a 220 nm-thick silicon layer and a 2 \\u0026mu;m-thick buried oxide (BOX) layer. The four microheaters (H\\u003csub\\u003e1\\u003c/sub\\u003e-H\\u003csub\\u003e4\\u003c/sub\\u003e) deposited over individual MRR waveguides are used for tuning the resonant wavelength of each MRR. The remaining microheaters (H\\u003csub\\u003e5\\u003c/sub\\u003e-H\\u003csub\\u003e14\\u003c/sub\\u003e) deposited over the upper and lower arms of MZI couplers are used for adjusting equivalent coupling coefficients between coupled MRRs, as shown in Fig.\\u0026nbsp;1c and\\u0026nbsp;1g. In practical operation, only one microheater (H\\u003csub\\u003e5\\u003c/sub\\u003e-H\\u003csub\\u003e9\\u003c/sub\\u003e) of the adjustable MZI coupler actively controls coupling coefficients, and the others (H\\u003csub\\u003e10\\u003c/sub\\u003e-H\\u003csub\\u003e14\\u003c/sub\\u003e) are used for backup. These microheaters enable refractive index modulation in waveguide structures by changing local temperature\\u003csup\\u003e20\\u003c/sup\\u003e. The measured linear resistances of the nine microheaters H\\u003csub\\u003e1\\u003c/sub\\u003e-H\\u003csub\\u003e9\\u003c/sub\\u003e are 1.5491, 1.5532, 1.6022, 1.5848, 0.2109, 0.2001, 0.2184, 0.2361, and 0.2222 k\\u0026Omega;, respectively. To ensure the stability of the optical and electrical ports during experimental tests, the on-chip device is optically and electrically packaged, as shown in Fig. 1h. The inset shows the zoomed-in view of the fabricated device. All pads fabricated on the chip are connected to DuPont connectors on a printed circuit board (PCB) via gold wires. The four grating couplers (GCs) are coupled and packaged with a standard fiber array (FA). When the microwave photonic signal is coupled into the device through GC\\u003csub\\u003e1\\u003c/sub\\u003e, the signal will be processed and then coupled out of the device through GC\\u003csub\\u003e3\\u003c/sub\\u003e.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eExperimental setups\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eFig.\\u0026nbsp;2\\u0026nbsp;illustrates the experimental setup for dynamically\\u0026nbsp;configuring the frequency response of the IMPF\\u0026nbsp;based on our proposed MOMS-HC algorithm.\\u0026nbsp;A\\u0026nbsp;CW light at 1550 nm generated by a laser source (LS, Koheras BasiK E15) is divided\\u0026nbsp;into two branches by\\u0026nbsp;an optical coupler (OC\\u003csub\\u003e1\\u003c/sub\\u003e) with a 7:3 splitting ratio. In the upper branch, 70% of the\\u0026nbsp;CW light\\u0026nbsp;is utilized for microwave signal modulation.\\u0026nbsp;The state\\u0026nbsp;of\\u0026nbsp;polarization\\u0026nbsp;(SOP)\\u0026nbsp;of the optical carrier is adjusted by a polarization controller (PC\\u003csub\\u003e1\\u003c/sub\\u003e)\\u0026nbsp;to be aligned with the principal\\u0026nbsp;polarization axis of the phase modulator (PM, Covega Mach-40). The microwave signal emitted by a vector network analyzer (VNA, Anritsu MS4647B) is modulated onto the optical carrier via PM. An optical bandpass filter (OBPF, Alnair BVF-300CL) is used to eliminate the\\u0026nbsp;-1st-order sideband, generating a\\u0026nbsp;single-sideband (SSB)\\u0026nbsp;signal\\u0026nbsp;(see Supplementary Notes\\u0026nbsp;IV\\u0026nbsp;for more details\\u0026nbsp;of the modulated signal). Then, the\\u0026nbsp;SSB signal is amplified by an erbium-doped fiber amplifier (EDFA, WZEDFA-SO-L-CW40) and processed by\\u0026nbsp;the device under test (DUT)\\u0026nbsp;after SOP adjustment via PC\\u003csub\\u003e2\\u003c/sub\\u003e. The\\u0026nbsp;+1st-order sideband signal processed by the\\u0026nbsp;DUT\\u0026nbsp;is combined with the remaining 30%\\u0026nbsp;CW light\\u0026nbsp;in the lower branch via OC\\u003csub\\u003e2\\u003c/sub\\u003e. The combined optical signal is then injected into a photodetector (PD, SHF 47100A) for\\u0026nbsp;O/E\\u0026nbsp;conversion to obtain the microwave signal. The microwave signal is amplified by a broadband electrical amplifier (EA, SHF S804B) before being fed into the VNA for measurement. PC\\u003csub\\u003e3\\u003c/sub\\u003e is used to adjust the SOP of the optical carrier in the lower branch to maximize the power of the beat frequency signal, and a variable optical attenuator (VOA) is used to adjust the optical power fed into the PD.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eConsidering the extensive measurement\\u0026nbsp;range\\u0026nbsp;of the optical spectrum analyzer (OSA),\\u0026nbsp;we\\u0026nbsp;first\\u0026nbsp;conduct preliminary tests\\u0026nbsp;in the\\u0026nbsp;optical domain,\\u0026nbsp;and execute the intelligent configuration of the DUT\\u0026nbsp;before configuring the IMPF. Based on the MOMS-HC\\u0026nbsp;algorithm,\\u0026nbsp;an optical filter with an RR of 18.93 dB and an FWHM bandwidth of 0.018 nm (2.25 GHz) is realized\\u0026nbsp;(see Supplementary Notes\\u0026nbsp;V\\u0026nbsp;for more details).\\u003c/p\\u003e\\n\\u003cp\\u003eIn brief, the intelligent configuration module consists of three\\u0026nbsp;components: controller, sensor, and actuator. The VNA functions as a sensor in the system, monitoring the operation state of the controlled device and relaying the data back to the computer, which serves as the central controller. The computer formulates the optimization problem via programming on the MATLAB platform and transmits the required control signals to the voltage source array (VSA, MCVS6400-A), which acts as the actuator. Both the sensor and the actuator are connected and controlled by the same computer.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eIntelligent configuration of IMPF\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eIn traditional manual manipulation of the IMPF, operators usually rely on accumulated experience and subjective judgment, making it difficult to systematically explore the optimization parameter space and prone to getting stuck in local optima (see Supplementary Notes III for more details). In contrast, the MOMS-HC algorithm enables synergy of all properties and fully unleashes the performance potential of its hardware.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eWe\\u0026nbsp;first demonstrate\\u0026nbsp;an\\u0026nbsp;intelligent configuration\\u0026nbsp;of\\u0026nbsp;the IMPF\\u0026nbsp;based on the 4th-order CROW, aiming for a center frequency of 30 GHz, an FWHM bandwidth of 3 GHz, an\\u0026nbsp;RR\\u0026nbsp;higher than 30 dB, an\\u0026nbsp;SF\\u0026nbsp;higher than\\u0026nbsp;0.7,\\u0026nbsp;and an in-band ripple\\u0026nbsp;smaller\\u0026nbsp;than 0.6\\u0026nbsp;dB. Fig.\\u0026nbsp;3a\\u0026nbsp;shows the\\u0026nbsp;normalized\\u0026nbsp;FOM as a function of the number of iterations during the configuration process, where the blue and green curves correspond to Stage II\\u0026nbsp;and III, respectively\\u0026nbsp;(see Supplementary Notes II for FOM definitions).\\u0026nbsp;The preprocessing and predefined steps in Stage I are omitted here due to their negligible\\u0026nbsp;time\\u0026nbsp;cost. Fig.\\u0026nbsp;3b-e and\\u0026nbsp;3f-i show the voltage\\u0026nbsp;distribution applied to all microheaters from\\u0026nbsp;H\\u003csub\\u003e1\\u003c/sub\\u003e to H\\u003csub\\u003e9\\u003c/sub\\u003e and the amplitude frequency response of the IMPF after different iterations, corresponding to nodes \\u0026ldquo;I\\u0026rdquo;, \\u0026ldquo;II\\u0026rdquo;, \\u0026ldquo;IV\\u0026rdquo;, and \\u0026ldquo;V\\u0026rdquo;, respectively. Specifically, Fig. 3b and 3f present the results after 8 iterations of the PSO algorithm in Stage II, corresponding to node \\u0026ldquo;I\\u0026rdquo; in Fig. 3a. At this node, the IMPF demonstrates a basic filtering function, but the passband exhibits splitting with poor edge symmetry, and the center frequency is approximately 40 GHz, as shown in Fig. 3f. This degradation is due to the significant temperature fluctuations of the chip caused by the extensive voltage search in this stage. However, the degradation does not affect the subsequent configuration.\\u003c/p\\u003e\\n\\u003cp\\u003eIn Stage III, the normalized FOM increases as the number of iterations increases. However,\\u0026nbsp;it\\u0026nbsp;does not increase monotonically and exhibits significant degradation and oscillation in early iterations, as shown by the FOM variation\\u0026nbsp;between\\u0026nbsp;node \\u0026ldquo;I\\u0026rdquo;\\u0026nbsp;and\\u0026nbsp;node \\u0026ldquo;III\\u0026rdquo;\\u0026nbsp;in\\u0026nbsp;Fig.\\u0026nbsp;3a. There are\\u0026nbsp;four\\u0026nbsp;main reasons for this phenomenon.\\u0026nbsp;Firstly,\\u0026nbsp;the global search for optimal control voltage\\u0026nbsp;at the beginning of\\u0026nbsp;each iteration leads\\u0026nbsp;to fluctuations\\u0026nbsp;in chip\\u0026nbsp;temperature. This is because abrupt voltage jumps (e.g., sudden high-to-low transitions) induce drastic heating power variations, which consequently lead to chip temperature fluctuations. Secondly,\\u0026nbsp;the\\u0026nbsp;adjustable\\u0026nbsp;parameters of the CROW-based IMPF interact with each other due to\\u0026nbsp;the complexity of\\u0026nbsp;its\\u0026nbsp;inherent structure. For example, adjusting either microheater of the MZI-based coupler not only modifies the coupling coefficient between the two adjacent MRRs but also alters their resonant wavelengths due to simultaneous optical phase shift. Moreover, the thermal crosstalk can further exacerbate the inter-coupling between\\u0026nbsp;these thermo-optic adjustable\\u0026nbsp;parameters. Thirdly,\\u0026nbsp;GA\\u0026nbsp;needs\\u0026nbsp;to simultaneously handle multiple optimization objectives,\\u0026nbsp;including FWHM bandwidth, center frequency, RR, SF, roll-off rate, insertion loss (IL), in-band ripple, response symmetry, and response smoothness associated with the absence of passband splitting, which results in a high degree of complexity in the evolution process\\u0026nbsp;(see Supplementary Notes\\u0026nbsp;II\\u0026nbsp;for more details). As a result, a certain amount of time is required for adaptive selection to balance the weights of\\u0026nbsp;different objectives.\\u0026nbsp;Fourthly,\\u0026nbsp;environmental\\u0026nbsp;temperature fluctuations\\u0026nbsp;and\\u0026nbsp;mechanical vibrations, as well as power supply noise\\u003csup\\u003e47\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e48\\u003c/sup\\u003e,\\u0026nbsp;also lead to\\u0026nbsp;degradation\\u0026nbsp;of\\u0026nbsp;the IMPF performance. These factors are\\u0026nbsp;also\\u0026nbsp;the main challenges in achieving high-performance IMPF.\\u003c/p\\u003e\\n\\u003cp\\u003eAfter 25\\u0026nbsp;iterations, the FOM reaches\\u0026nbsp;0.953,\\u0026nbsp;as shown by\\u0026nbsp;node \\u0026ldquo;III\\u0026rdquo;\\u0026nbsp;in\\u0026nbsp;Fig.\\u0026nbsp;3a.\\u0026nbsp;We can see that the frequency response of the IMPF gradually converges, indicating that the overall temperature of the chip is stabilizing.\\u0026nbsp;Fig.\\u0026nbsp;3h\\u0026nbsp;shows the frequency response of the IMPF after 31 iterations, which corresponds to\\u0026nbsp;node \\u0026ldquo;IV\\u0026rdquo;\\u0026nbsp;in Fig.\\u0026nbsp;3a.\\u0026nbsp;The normalized FOM reaches\\u0026nbsp;0.969.\\u0026nbsp;The overall performance is satisfactory\\u0026nbsp;except for\\u0026nbsp;the in-band ripple, which\\u0026nbsp;is 0.85\\u0026nbsp;dB. Subsequent iterations focus on further\\u0026nbsp;flattening the passband\\u0026nbsp;to approach the target response. Finally, after 42 iterations,\\u0026nbsp;the FOM reaches\\u0026nbsp;0.982\\u0026nbsp;corresponding to\\u0026nbsp;node \\u0026ldquo;V\\u0026rdquo;\\u0026nbsp;in Fig. 3a. The corresponding\\u0026nbsp;IMPF achieves\\u0026nbsp;an FWHM\\u0026nbsp;bandwidth of 2.97\\u0026nbsp;GHz, a center frequency of 29.899\\u0026nbsp;GHz, an\\u0026nbsp;RR of 31.21\\u0026nbsp;dB,\\u0026nbsp;an\\u0026nbsp;in-band ripple\\u0026nbsp;of\\u0026nbsp;0.52\\u0026nbsp;dB, an\\u0026nbsp;SF of 0.74, and a roll-off rate of\\u0026nbsp;13.08\\u0026nbsp;dB/GHz\\u0026nbsp;simultaneously, as shown in\\u0026nbsp;Fig.\\u0026nbsp;3i.\\u0026nbsp;The results show that all properties\\u0026nbsp;of the IMPF\\u0026nbsp;perfectly meet the expected targets.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eTuning and\\u003c/strong\\u003e\\u003cstrong\\u003e\\u0026nbsp;\\u003c/strong\\u003e\\u003cstrong\\u003ereconfiguring\\u003c/strong\\u003e\\u003cstrong\\u003e\\u0026nbsp;\\u003c/strong\\u003e\\u003cstrong\\u003ethe response of IMPF\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe powerful tunability and reconfigurability are essential capabilities of\\u0026nbsp;the\\u0026nbsp;IMPF. To demonstrate the excellent performance of the\\u0026nbsp;intelligently configured\\u0026nbsp;IMPF, we first evaluate\\u0026nbsp;its\\u0026nbsp;consistency while tuning\\u0026nbsp;the\\u0026nbsp;center frequency\\u0026nbsp;from\\u0026nbsp;2 to 48 GHz\\u0026nbsp;with different\\u0026nbsp;target bandwidths.\\u003c/p\\u003e\\n\\u003cp\\u003eFig. 4a illustrates the measured center frequency tuning results of the IMPF with bandwidths of 0.66, 1.85, and 3.15 GHz, respectively. From the measured results, we can draw the following conclusions. Firstly, for the narrow bandwidth operation (0.66 \\u0026plusmn; 0.02 GHz), the IMPF exhibits a maximum RR of 31.47 dB with a roll-off rate up to 17.50 dB/GHz, as shown in Fig. 4a I. This high roll-off rate improves the rejection of out-of-band interference while preserving precise frequency selectivity, a crucial aspect of high-performance filtering. Secondly, for the medium bandwidth operation (1.85 \\u0026plusmn; 0.10 GHz), the maximum RR reaches 37.67 dB with a roll-off rate up to 13.21 dB/GHz, delivering balanced selectivity for multi-channel communication environments, as shown in Fig. 4a II. Finally, for the wide bandwidth operation (3.15 \\u0026plusmn; 0.03 GHz), the IMPF maintains an in-band ripple of less than 0.61 dB, as shown in Fig. 4a III. Meanwhile, the RR, roll-off rate, and SF are as high as 31.42 dB, 13.33 dB/GHz, and 0.75, respectively.\\u003c/p\\u003e\\n\\u003cp\\u003eFig. 4b-d present labeled heatmaps illustrating the bandwidth, RR, and SF of the IMPF at different center frequencies, respectively. In Fig. 4b-d, the horizontal axis represents three bandwidth labels, corresponding to respective 0.66 GHz \\u003cstrong\\u003eI\\u003c/strong\\u003e, 1.85 GHz \\u003cstrong\\u003eII\\u003c/strong\\u003e, and 3.15 GHz \\u003cstrong\\u003eIII\\u003c/strong\\u003e, while the vertical axis represents the center frequency labels. The color intensity in each heatmap corresponds to the magnitude of the respective property (bandwidth, RR, and SF). We can see that the variations of bandwidth and RR during the center frequency tuning are less than \\u0026plusmn;0.10 GHz and \\u0026plusmn;1.00 dB, respectively. The results demonstrate that the performance of the IMPF maintains excellent consistency across the whole operational band, representing a key advantage of IMPFs.\\u003c/p\\u003e\\n\\u003cp\\u003eMeanwhile,\\u0026nbsp;the\\u0026nbsp;FWHM\\u0026nbsp;bandwidth of the IMPF can be continuously reconfigured\\u0026nbsp;from 0.66 to 4.15 GHz\\u0026nbsp;by jointly configuring the control voltages via the MOMS-HC algorithm, and the SF varied between\\u0026nbsp;0.45\\u0026nbsp;and 0.78, as shown in Fig.\\u0026nbsp;4e and\\u0026nbsp;4f, respectively.\\u0026nbsp;When the bandwidth is reconfigured from 0.66 to 3.68 GHz, the\\u0026nbsp;in-band ripple\\u0026nbsp;stays\\u0026nbsp;below 0.63\\u0026nbsp;dB, and the RR\\u0026nbsp;stays above 30.14\\u0026nbsp;dB, as shown in Fig.\\u0026nbsp;4f and\\u0026nbsp;4g, respectively.\\u0026nbsp;When the bandwidth\\u0026nbsp;of the IMPF\\u0026nbsp;is reconfigured to 4.15 GHz, the RR drops\\u0026nbsp;to 26.66\\u0026nbsp;dB, and the in-band ripple increases\\u0026nbsp;to 0.91\\u0026nbsp;dB, as shown by dark purple curve in Fig.\\u0026nbsp;4e. The degradation\\u0026nbsp;in overall performance\\u0026nbsp;primarily stems from\\u0026nbsp;the\\u0026nbsp;limited\\u0026nbsp;adjustable range of the equivalent coupling coefficient\\u0026nbsp;based on the MZI coupler, resulting in a trade-off between the RR and in-band ripple. The frequency response\\u0026nbsp;with\\u0026nbsp;a bandwidth\\u0026nbsp;of\\u0026nbsp;4.15 GHz indicates that the\\u0026nbsp;4th-order CROW\\u0026nbsp;is approaching its hardware-imposed performance limit. The result also demonstrates the capability of the configuration framework based on the MOMS-HC\\u0026nbsp;algorithm\\u0026nbsp;to fully unleash the hardware potential of the optical device.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eSelf-stabilization\\u003c/strong\\u003e\\u003cstrong\\u003e\\u0026nbsp;and self-\\u003c/strong\\u003e\\u003cstrong\\u003erecover\\u003c/strong\\u003e\\u003cstrong\\u003ey\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe self-stabilization and sustained operation capability\\u0026nbsp;of the IMPF are critical for a practical system.\\u0026nbsp;Under\\u0026nbsp;manual\\u0026nbsp;manipulation, the absence of dynamic adjustment and stabilization mechanisms makes it challenging to maintain sustained and stable operation\\u0026nbsp;of the IMPF.\\u0026nbsp;In contrast,\\u0026nbsp;the\\u0026nbsp;MOMS-HC\\u0026nbsp;algorithm\\u0026nbsp;enables\\u0026nbsp;dynamic\\u0026nbsp;self-configuration of\\u0026nbsp;adjustable parameters\\u0026nbsp;during\\u0026nbsp;IMPF\\u0026nbsp;operation\\u0026nbsp;to effectively suppress the negative effects of\\u0026nbsp;environmental disturbances.\\u0026nbsp;The elaborate design of the SA algorithm aims to provide real-time voltage compensation while minimizing temperature fluctuations as much as possible\\u0026nbsp;(see Supplementary Notes\\u0026nbsp;I\\u0026nbsp;for more details).\\u003c/p\\u003e\\n\\u003cp\\u003eBefore evaluating the self-stabilization of the IMPF, we first configure its various properties as follows: the center frequency and FWHM bandwidth are set as 15.000 and 0.90 GHz, and the SF and RR are set to exceed 0.50 and 30.00 dB, respectively. The above state serves as the initial state for all subsequent evaluations in this section. Next, we let the IMPF operate in the laboratory environment for three hours with MOMS-HC off and on, respectively. The basic operating cycle (BOC) is set as 60 seconds. When the MOMS-HC is off, the IMPF operates normally within each BOC without executing additional intelligent configurations. When the MOMS-HC is on, SA obtains the current amplitude frequency response of the IMPF in real-time during each BOC, monitoring degradation of properties such as center frequency, SF, RR, and bandwidth. Any degradation exceeding the threshold will trigger the intelligent configuration process, and the configuration is constrained to be completed within a BOC.\\u0026nbsp;The current amplitude frequency response of the IMPF captured at the end of each BOC will be overlapped to generate an \\u0026ldquo;eye-like\\u0026rdquo; diagram. The \\u0026ldquo;eye-like\\u0026rdquo; diagrams without and with MOMS-HC-based\\u0026nbsp;configuration are\\u0026nbsp;shown as Fig.\\u0026nbsp;5a and\\u0026nbsp;5b, respectively. We can see that the \\u0026ldquo;eye-like\\u0026rdquo; diagram is highly chaotic when the MOMS-HC is off, indicating significant performance degradation of the IMPF over the 3 hours. The degradation is primarily caused by environmental disturbances, such as\\u0026nbsp;temperature fluctuations\\u003csup\\u003e46\\u003c/sup\\u003e, mechanical vibrations, and accumulated\\u0026nbsp;power supply noise\\u003csup\\u003e47\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e48\\u003c/sup\\u003e. In contrast, when the MOMS-HC is on, the \\u0026ldquo;eye-like\\u0026rdquo; diagram distinctly exhibits sustained and stable dynamic characteristics of the IMPF. The nine voltages applied to the microheaters are dynamically adjusted in real time during IMPF operation, as shown in Fig. 5c.\\u003c/p\\u003e\\n\\u003cp\\u003eTable\\u0026nbsp;1\\u0026nbsp;shows the comparison of IMPF with MOMS-HC on and off during the 3-hour stability test. When the MOMS-HC is on, the mean center frequency of the IMPF is 15.002 GHz with a relative standard deviation of 0.085%, and the fluctuation range of center frequency is between 14.970 and 15.051 GHz. The maximum frequency drift is less than \\u0026plusmn; 0.051 GHz, yielding a 234-fold improvement in stability compared to the maximum frequency drift of up to 11.950 GHz when the MOMS-HC is off. The average SF is 0.52 with a relative standard deviation of 0.667%, and the average bandwidth is 0.90 GHz with a relative standard deviation of 0.682%. During the 3-hour stability test, both the RR and roll-off rate remain above 29.89 dB and 16.57 dB/GHz, respectively. These results demonstrate that the MOMS-HC algorithm effectively suppresses\\u0026nbsp;performance degradation caused by environmental disturbances, allowing the IMPF to maintain\\u0026nbsp;sustained\\u0026nbsp;stability and consistency over an extended period. The detailed time-dependent variations of the SF, center frequency offset, bandwidth, and RR without and with MOMS-HC-based\\u0026nbsp;configuration\\u0026nbsp;are shown in Fig. 5d-g, respectively.\\u003c/p\\u003e\\n\\u003cp\\u003eTable 1. Comparison of IMPF with MOMS-HC on and off\\u003c/p\\u003e\\n\\u003cdiv align=\\\"\\\"\\u003e\\n \\u003ctable border=\\\"0\\\" cellspacing=\\\"0\\\" cellpadding=\\\"0\\\" width=\\\"637\\\"\\u003e\\n \\u003ctbody\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd colspan=\\\"2\\\" style=\\\"width: 198px;\\\"\\u003e\\n \\u003cp\\u003eProperty\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003eBW (GHz)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003eRR (dB)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003eCenter (GHz)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003eSF\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd colspan=\\\"2\\\" style=\\\"width: 198px;\\\"\\u003e\\n \\u003cp\\u003eTarget\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.90\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e\\u0026gt;30.00\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e15.000\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e\\u0026gt;0.50\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd rowspan=\\\"4\\\" style=\\\"width: 76px;\\\"\\u003e\\n \\u003cp\\u003eMOMS-HC off\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eMean\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.70\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e24.18\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e20.508\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.24\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eMin\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.56\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e17.59\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e15.091\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.10\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eMax\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.99\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e30.80\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e26.950\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.53\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eRela.Std. Dev.\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e17.553%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e16.367%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e15.048%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e51.708%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd rowspan=\\\"4\\\" style=\\\"width: 76px;\\\"\\u003e\\n \\u003cp\\u003eMOMS-HC on\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eMean\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.90\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e30.52\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e15.002\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.52\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eMin\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.88\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e29.89\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e14.970\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.51\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eMax\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.92\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e31.36\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e15.051\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.54\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 123px;\\\"\\u003e\\n \\u003cp\\u003eRela.Std. Dev.\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 87px;\\\"\\u003e\\n \\u003cp\\u003e0.682%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 116px;\\\"\\u003e\\n \\u003cp\\u003e0.648%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.085%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e0.667%\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003c/tbody\\u003e\\n \\u003c/table\\u003e\\n\\u003c/div\\u003e\\n\\u003cp\\u003eAbbreviations: Rela.Std. Dev.: relative standard deviation, BW: bandwidth.\\u003c/p\\u003e\\n\\u003cp\\u003eWe also evaluate the self-recovery capability of the intelligently configured IMPF under extreme environmental disturbances. In this evaluation, we let the IMPF operate for 30 minutes with MOMS-HC on. The initial state is the same as that of the 3-hour stability test, operating with a center frequency of 15.010 GHz and a bandwidth of 0.90 GHz, as indicated by the green curve in Fig.\\u0026nbsp;5h. To simulate extreme disturbances, random voltage biases of up to 200 mV are simultaneously applied to all nine control ports at the beginning of the 4th BOC. The red curve in Fig.\\u0026nbsp;5h illustrates the degraded frequency response of IMPF at the beginning of applying voltage disturbances to the control ports. We can observe that the center frequency of the IMPF instantaneously drifts toward higher frequencies by 1.637 GHz, and the RR decreases by 10.62 dB, from 30.27 dB to 19.65 dB. Concurrently, the SF drops to 0.14, and the passband is severely split. However, the MOMS-HC algorithm enables the IMPF to recover to its initial operation state within 5 minutes, as demonstrated in Fig.\\u0026nbsp;5i. During the subsequent 22-minute period, the IMPF maintains an optimal operation state with an RR exceeding 30.01 dB and a roll-off rate above 15.00 dB/GHz. Fig.\\u0026nbsp;5j indicates the \\u0026ldquo;eye-like\\u0026rdquo; diagram characterizing the stable performance of the IMPF from the 9th to the 30th minute. Meanwhile, Fig.\\u0026nbsp;5k-n display the temporal variations of SF, frequency offset, bandwidth, and RR over the 30-minute evaluation period, respectively.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eIMPF-based channel equalization\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eTable\\u0026nbsp;2. Results of Channel Equalization\\u003c/p\\u003e\\n\\u003cdiv align=\\\"\\\"\\u003e\\n \\u003ctable border=\\\"0\\\" cellspacing=\\\"0\\\" cellpadding=\\\"0\\\" width=\\\"593\\\" class=\\\"fr-table-selection-hover\\\"\\u003e\\n \\u003ctbody\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 134px;\\\"\\u003e\\n \\u003cp\\u003eCenter (GHz)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 102px;\\\"\\u003e\\n \\u003cp\\u003eBW (GHz)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003eUnequalized Ripple (dB)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 111px;\\\"\\u003e\\n \\u003cp\\u003eEqualized Ripple (dB)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 129px;\\\"\\u003e\\n \\u003cp\\u003eRipple Reduction (dB)\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 134px;\\\"\\u003e\\n \\u003cp\\u003e5.5\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 102px;\\\"\\u003e\\n \\u003cp\\u003e3.02\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e3.66\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 111px;\\\"\\u003e\\n \\u003cp\\u003e1.49\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 129px;\\\"\\u003e\\n \\u003cp\\u003e2.42\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 134px;\\\"\\u003e\\n \\u003cp\\u003e22.0\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 102px;\\\"\\u003e\\n \\u003cp\\u003e3.51\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e4.24\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 111px;\\\"\\u003e\\n \\u003cp\\u003e1.87\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 129px;\\\"\\u003e\\n \\u003cp\\u003e2.37\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 134px;\\\"\\u003e\\n \\u003cp\\u003e27.0\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 102px;\\\"\\u003e\\n \\u003cp\\u003e3.22\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e2.98\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 111px;\\\"\\u003e\\n \\u003cp\\u003e0.84\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 129px;\\\"\\u003e\\n \\u003cp\\u003e2.14\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd style=\\\"width: 134px;\\\"\\u003e\\n \\u003cp\\u003e32.5\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 102px;\\\"\\u003e\\n \\u003cp\\u003e4.34\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 118px;\\\"\\u003e\\n \\u003cp\\u003e3.44\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 111px;\\\"\\u003e\\n \\u003cp\\u003e1.53\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd style=\\\"width: 129px;\\\"\\u003e\\n \\u003cp\\u003e1.91\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003c/tbody\\u003e\\n \\u003c/table\\u003e\\n\\u003c/div\\u003e\\n\\u003cp\\u003eAbbreviations: BW: bandwidth.\\u003c/p\\u003e\\n\\u003cp\\u003eIn high-speed and large-capacity communication systems, signals using advanced modulation formats are prone to distortion induced by electro-optic and photoelectric conversion, electrical amplifiers, and fiber dispersion. Among these challenges, frequency-dependent fading stands as the primary impairment mechanism. This is primarily caused by two factors: the chromatic dispersion in optical fibers and the inherent frequency-dependent amplitude responses of the components (e.g., PM, PD, and EA) in link\\u003csup\\u003e49\\u003c/sup\\u003e. Traditionally, this issue can be solved by modifying the architecture of intensity modulation and direct detection systems, including single-sideband or vestigial-sideband modulation, intensity modulation with coherent detection, and Kramers-Kronig receiver\\u003csup\\u003e50\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e51\\u003c/sup\\u003e. However, these approaches inevitably increase system complexity and cost. Alternatively, advanced digital signal processing (DSP) techniques are employed to perform pre-equalization based on accurate channel estimation and channel feedback, but these techniques often suffer from high computational complexity and power consumption\\u003csup\\u003e54,55\\u003c/sup\\u003e.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eIn this section, the intelligently configured IMPF, based on the MOMS-HC algorithm, is employed to compensate for frequency-dependent fading and achieve channel equalization. This equalization capability is achieved by dynamically reshaping the passband response of the IMPF (see Supplementary Notes\\u0026nbsp;II\\u0026nbsp;for more details). To demonstrate this capability, we directly utilize the microwave photonic (MWP) link shown in Fig.\\u0026nbsp;2\\u0026nbsp;to simulate the practical scenario with frequency-dependent fading. Fig.\\u0026nbsp;6a shows the inherent amplitude frequency response of the MWP link\\u0026nbsp;without\\u0026nbsp;the IMPF based on the\\u0026nbsp;4th-order CROW. We can observe significant amplitude ripples and degradation across the operational band from 2 to 50 GHz. These effects are attributed to the combined influence of the inherent responses of each device in the MWP link. Specifically, within the 2\\u0026ndash;10 GHz range, the filtering edge of the OBPF adopted for SSB modulation induces higher loss for the lower frequency compared to the higher frequency. In the 10\\u0026ndash;50 GHz range, the combined effects of parasitic capacitance in the EA\\u003csup\\u003e54\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e55\\u003c/sup\\u003e\\u003csup\\u003e\\u0026nbsp;\\u003c/sup\\u003e and the bandwidth constraints of both the PD and PM lead to degradation in response flatness and insufficient responsivity at the high frequencies.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eTherefore, we set the channel window centers at 5.5, 22.0, 27.0, and 32.5 GHz and set different target bandwidths for each to evaluate the dynamic equalization capability of the IMPF. The green and red curves in Fig.\\u0026nbsp;6b-e indicate the equalized channel and the configured IMPF amplitude\\u0026nbsp;frequency\\u0026nbsp;responses, respectively. We can see that multiple equalized MWP links with flat passbands at various center frequencies are established. These responses demonstrate that, owing to the MOMS-HC algorithm, the IMPF passband is adjusted to match the inherent response of the link, and its transition edges are also adjusted to ensure excellent symmetry in the equalized link response. Table\\u0026nbsp;2\\u0026nbsp;shows channel equalization results with different center frequencies and bandwidths. We can see that when the center frequency of the channel window is respective 5.5, 22.0, 27.0, and 32.5 GHz, the corresponding in-band ripple before equalization is 3.66, 4.24, 2.98, and 3.44 dB, respectively. After applying the IMPF-based equalization to the MWP link, the corresponding in-band ripple is reduced to 1.49, 1.87, 0.84, and 1.53 dB, respectively. Therefore, these experimental results demonstrate the effectiveness of IMPF-based equalization. Additionally, we can observe from Fig.\\u0026nbsp;6b-e that the equalized channel response has different excess ILs at different channel windows. This is because the PM used in the experiment is polarization-dependent, and SOP variations during the configuration process can result in gain degradation in the MWP link. To solve this problem, an electrically controlled PC can be used and co-configured by the MOMS-HC algorithm\\u003csup\\u003e56\\u003c/sup\\u003e.\\u003c/p\\u003e\"},{\"header\":\"Discussion\",\"content\":\"\\u003cp\\u003eBased on the MOMS-HC algorithm, we have successfully implemented an intelligently configured IMPF with continuous tuning of the center frequency from 2 to 48 GHz and a wide bandwidth reconfiguration range from 0.66 to 4.15 GHz, featuring self-stabilization and self-recovery capabilities. The MOMS-HC algorithm also significantly enhances the performance consistency of IMPF during the frequency tuning. In contrast to conventional equalization techniques, the MOMS-HC algorithm enables the IMPF to equalize intra-channel\\u0026nbsp;fading\\u0026nbsp;through direct passband reshaping, eliminating the need for extra costly devices or digital equalization algorithms. This hardware-instead-of-software approach promises to enable end-to-end integrated microwave photonic system architectures\\u003csup\\u003e7\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e42\\u003c/sup\\u003e.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eMeanwhile, the proposed MOMS-HC algorithm is fundamentally an in situ training process. The\\u0026nbsp;\\u0026ldquo;train-as-deploy\\u0026rdquo;\\u0026nbsp;paradigm\\u0026nbsp;ensures that the IMPF can be directly used for specific wireless communication tasks once the training is complete. Owing to\\u0026nbsp;this paradigm, the IMPF\\u0026nbsp;also circumvents the reliance on\\u0026nbsp;electromagnetic\\u0026nbsp;simulations\\u0026nbsp;and the challenges of physical alignment faced by traditional methods. In addition to IMPFs, the MOMS-HC algorithm is also universally applicable to IOPUs of arbitrary architectures and can be extended to any scenario where a FOM can be defined to evaluate the state of the system.\\u003c/p\\u003e\\n\\u003cp\\u003eNotably, bandwidth reconfiguration of IMPFs over a broader range remains challenging, particularly for achieving bandwidths as narrow as the MHz level. Two main factors limit the reconfigurable bandwidth range of the CROW-based IMPFs. Firstly, high waveguide loss results in an inadequate microring Q-factor, which constrains the lower limit of the reconfigurable bandwidth. The utilization of optimized multimode waveguides with Euler bends\\u003csup\\u003e18\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e37\\u003c/sup\\u003e or free-form curved\\u003csup\\u003e57\\u003c/sup\\u003e microrings can overcome this limitation while improving the narrowband filtering performance of the IMPF. Secondly, the combined effects of the non-ideal DC self-coupling coefficient and thermal crosstalk limit the adjustable range of the equivalent coupling coefficient. This limitation consequently reduces the maximum reconfigurable bandwidth and hinders arbitrary passband shaping. Increasing the interval between the two arms of the MZI coupler and the length of the two arms\\u003csup\\u003e23\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e26\\u003c/sup\\u003e, as well as carving grooves between microheaters\\u003csup\\u003e15\\u003c/sup\\u003e, can enlarge the adjustable range of the MZI coupling coefficient. Furthermore, the packaged chip in this work lacks TEC integration, leading to potential overvoltage and excessive power consumption. Future integration of TEC modules can address these limitations through active temperature stabilization and thermal management\\u003csup\\u003e12\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e26\\u003c/sup\\u003e.\\u003c/p\\u003e\"},{\"header\":\"Summary\",\"content\":\"\\u003cp\\u003eIn summary, we have proposed an intelligent configuration strategy based on the MOMS-HC algorithm that fully unleashes the hardware potential of IOPU. The MOMS-HC algorithm achieves self-configuration of IOPUs through global-local exploration, robust multi-objective equilibrium, efficient self-optimization, and environmental robustness empowerment. By employing this algorithm, we demonstrate an intelligently configured bandpass IMPF with four key capabilities: wideband center frequency tunability, flexible bandwidth reconfigurability, self-stabilization, and excellent channel equalization. The roll-off rate, RR, and SF of the IMPF reach up to 17.50 dB/GHz, 37.67 dB, and 0.78, respectively. Notably, with the MOMS-HO algorithm, the IMPF can operate stably for 3 hours, and its frequency stability is improved by 234 times. Furthermore, the MOMS-HO algorithm also enables the IMPF to recover to its target operating state quickly when all control ports experience significant disturbances simultaneously, which is rather challenging for manual manipulation. The MOMS-HC algorithm also enables dynamic compensation of intra-channel fading through reshaping the passband response of the IMPF, thus achieving channel equalization. This capability positions the intelligently configured IMPF as a viable alternative to conventional pre-equalization achieved by DSP. Overall, this work makes a solid step forward to applications of the IMPF in practical microwave systems.\\u003c/p\\u003e\"},{\"header\":\"Methods\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eDesign of the MOMS-HC algorithm\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe MOMS-HC algorithm is a task-driven intelligent configuration framework with four stages. Its detailed procedure is as follows. In Stage I, the MOMS-HC algorithm first calculates and defines various properties of the IOPU programmatically. The FOM is then predefined for the subsequent configuration. Considering the global search capability and fast convergence of the PSO algorithm\\u003csup\\u003e43\\u003c/sup\\u003e, the MOMS-HC algorithm selects PSO for initial optimization in Stage II. The PSO algorithm randomly generates an initial state, where each particle is represented as a multi-dimensional vector corresponding to the multiple control voltages one-to-one. The fitness of each particle (i.e., the FOM) is defined as a function of the most important property of the IOPU to achieve rapid global optimization. In the absence of a packaged thermo-electric cooler (TEC), there is a strong correlation between the chip temperature and the magnitude of initial control voltages, which makes the configuration of the IOPU sensitive to the initial point. Therefore, the MOMS-HC algorithm iterates Stage II over multiple random initial points until the convergence criteria are met, and the optimal state is selected at the maximum FOM.\\u003c/p\\u003e\\n\\u003cp\\u003eThe core of Stage III is an intelligent optimizer based on the standard GA\\u003csup\\u003e44\\u003c/sup\\u003e. Each individual consists of multiple genes, and the initial genes are obtained from Stage II. The population size should be set to a relatively large value, typically more than 20 times the number of parameters, to balance the global search capability and rapid convergence speed. Notably, the FOM is redefined to consider multiple objectives in this stage. Based on standard genetic operations, such as roulette-wheel selection, gene crossover, mutation, and environmental selection, the MOMS-HC algorithm can balance the weights of different objectives to avoid local optima in multi-objective optimization problems and maintain excellent search capability in uncertain environments. The global search of control voltage significantly exacerbates the overall temperature fluctuation of the chip due to electric power variations and crosstalk, leading to serious degradation of IOPU performance. Therefore, in Stage II and Stage III, the MOMS-HC algorithm incorporates a multi-level constraint mechanism on the voltage search space, effectively suppressing the overall temperature fluctuations.\\u003c/p\\u003e\\n\\u003cp\\u003eFollowing the completion of Stage III, Stage IV will be initiated to ensure sustained and stable operation of IOPU. Alternatively, iterative re-execution of Stage III enables dynamically adjusting the operational properties of IOPU, including bandwidth reconfiguration and center frequency tuning. Considering the environmental disturbances in practical operating environments, such as temperature fluctuations\\u003csup\\u003e46\\u003c/sup\\u003e, mechanical vibrations, and power supply noise\\u003csup\\u003e47\\u003c/sup\\u003e\\u003csup\\u003e,\\u003c/sup\\u003e\\u003csup\\u003e48\\u003c/sup\\u003e, the control voltage must be dynamically adjusted to compensate for the performance degradation of the IOPU. Therefore, in Stage IV, the MOMS-HC algorithm performs a local fine-tuning search using the SA algorithm\\u003csup\\u003e45\\u003c/sup\\u003e. Customizing the temperature decay mechanism in the SA algorithm can effectively limit the magnitude of each adjustment. Initially, the high annealing temperature permitted a relatively large voltage search range. Later, the annealing temperature rapidly reduces to restrict the voltage search range, thereby suppressing secondary performance fluctuations induced by voltage search and ensuring the performance consistency of the IOPU (see Supplementary Notes I for more details of the sub-algorithm). Throughout the whole execution of the MOMS-HC algorithm, the IOPU frequency response is obtained in real time by the detection module, such as a photodetector.\\u003c/p\\u003e\\n\\u003cp\\u003eIn brief, PSO drives global parameter initialization, leveraging its swarm intelligence to accelerate convergence of IMPF toward the target response. GA executes precision configuration under multi-objective constraints by leveraging crossover and mutation operators. SA suppresses environmental disturbances through a dynamic fine-tuning search to ensure sustained and stable operation of IMPF.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003ePrecision\\u003c/strong\\u003e\\u003cstrong\\u003e\\u0026nbsp;and time cost of the configuration\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eIn the 3-hour stability test, when the MOMS-HC algorithm is applied to the IMPF, the center frequency of the IMPF drifts between +0.051 GHz and -0.030 GHz relative to the target value of 15.000 GHz, and the bandwidth varies between +0.02 GHz and -0.02 GHz relative to the target value of 0.90 GHz. These deviations are primarily attributed to the combined effects of optimization constraints imposed by the setting of the BOC and measurement errors induced by instrumentation limitations (0.010 GHz resolution of the VNA).\\u003c/p\\u003e\\n\\u003cp\\u003eIn our experiment, the sensors (including OSA and VNA) and actuators (VSA) are accessed via serial and Ethernet ports using the TCP/IP protocol, resulting in a total communication speed of around 2 Hz. Without relying on any prior experience, the first configuration of IMPF can be finished within 2000 s (less than 40 iterations, with 100 calculations per iteration), while the response time of the TiN heater is approximately 10 \\u0026mu;s\\u003csup\\u003e58\\u003c/sup\\u003e. For center frequency tuning with the same bandwidth or bandwidth reconfiguration at the same center frequency, the MOMS-HC algorithm requires only a single execution of Stage II. For example, after configuring the IMPF to a specific response (e.g., the center frequency is 25 GHz and the bandwidth is 0.66 GHz), the algorithm can directly use the current applied voltages as the starting point and repeat Stage III to reconfigure the IMPF. This mechanism significantly reduces the time cost. Therefore, the single reconfiguration of the IMPF, including bandwidth reconfiguration and center frequency tuning, can be completed within 400 s (less than 8 iterations). If the external driver speed is increased to make the communication time cost negligible, the first configuration and the single reconfiguration of IMPF can be completed in 8 ms and 40 ms, respectively. Therefore, to further reduce the time cost, techniques such as optoelectronic co-integration\\u003csup\\u003e59\\u003c/sup\\u003e and frequency-to-time mapping\\u0026nbsp;\\u003csup\\u003e60\\u003c/sup\\u003e can be used in the future.\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eSupporting Information\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eSupporting Information is available from the Online Library.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eData Availability\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe data that support the findings of this study are available from the corresponding author upon request.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eConflict of interest\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe authors declare no competing interests.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eContributions\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eY.T.S. and X.L.Z. conceived the idea. Y.T.S., H.L.Z., and Y. Y. designed the MOMS-HCS algorithm and engineered the related experiments. Y.F.L. and Y. Y. designed the device as a demonstration platform. Y.T.S. performed the experiments and recorded and analyzed the data. K.X.C. and M.M.D. assisted with the measurements. Y.T.S. and Y.Y. prepared the manuscript. Y. Y., F.Z. Z. and X.L.Z. supervised the project. All authors participated in revising the manuscript and approved the submission.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAcknowledgements\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis project was supported by the National Key Research and Development Program of China (2024YFF0728000), the Program for HUST Academic Frontier Youth Team (2018QYTD08), and the Project of Key Laboratory of Radar Imaging and Microwave Photonics (Nanjing University of Aeronautics and Astronautics), Ministry of Education (NJ20230001).\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\n \\u003cli\\u003eIlderem, V. The technology underpinning 5G. \\u003cem\\u003eNat. Electron.\\u003c/em\\u003e \\u003cstrong\\u003e3\\u003c/strong\\u003e, 5\\u0026ndash;6 (2020).\\u003c/li\\u003e\\n \\u003cli\\u003eAkyildiz, I. F., Kak, A. \\u0026amp; Nie, S. 6G and Beyond: The future of wireless communications systems. \\u003cem\\u003eIEEE Access\\u003c/em\\u003e \\u003cstrong\\u003e8\\u003c/strong\\u003e, 133995\\u0026ndash;134030 (2020).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, W. et al. A fully reconfigurable photonic integrated signal processor.\\u0026nbsp;\\u003cem\\u003eNat. Photonics\\u003c/em\\u003e\\u003cem\\u003e\\u0026nbsp;\\u003c/em\\u003e\\u003cstrong\\u003e10\\u003c/strong\\u003e, 190\\u0026ndash;195 (2016).\\u003c/li\\u003e\\n \\u003cli\\u003eMarpaung, D., Yao, J. \\u0026amp; Capmany, J. Integrated microwave photonics. \\u003cem\\u003eNat. Photonics\\u003c/em\\u003e \\u003cstrong\\u003e13\\u003c/strong\\u003e, 80\\u0026ndash;90 (2019).\\u003c/li\\u003e\\n \\u003cli\\u003eYao, J. Microwave photonic systems. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e40\\u003c/strong\\u003e, 6595\\u0026ndash;6607 (2022).\\u003c/li\\u003e\\n \\u003cli\\u003eCapmany, J. \\u0026amp; Novak, D. Microwave photonics combines two worlds. \\u003cem\\u003eNat. Photonics\\u003c/em\\u003e \\u003cstrong\\u003e1\\u003c/strong\\u003e, 319\\u0026ndash;330 (2007).\\u003c/li\\u003e\\n \\u003cli\\u003eMarpaung, D. et al. Integrated microwave photonics. \\u003cem\\u003eLaser Photonics Rev.\\u003c/em\\u003e \\u003cstrong\\u003e7\\u003c/strong\\u003e, 506\\u0026ndash;538 (2013).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, Y. et al. Integrated microwave photonic filters. \\u003cem\\u003eAdv. Opt. Photon.\\u003c/em\\u003e \\u003cstrong\\u003e12\\u003c/strong\\u003e, 485 (2020).\\u0026nbsp;\\u003c/li\\u003e\\n \\u003cli\\u003eFandi\\u0026ntilde;o, J. S.et al.\\u0026nbsp;A monolithic integrated photonic microwave filter. \\u003cem\\u003eNat. Photonics\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, 124\\u0026ndash;129 (2017).\\u0026nbsp;\\u003c/li\\u003e\\n \\u003cli\\u003eWang, L., Wang, X. \\u0026amp; Pan, S. Microwave photonics empowered integrated sensing and communication for 6G. \\u003cem\\u003eIEEE Trans. Microwave Theory Techn.\\u003c/em\\u003e 1\\u0026ndash;21 (2025).\\u0026nbsp;\\u003c/li\\u003e\\n \\u003cli\\u003eDaulay, O. et al. Ultrahigh dynamic range and low noise figure programmable integrated microwave photonic filter. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e13\\u003c/strong\\u003e, 7798 (2022).\\u003c/li\\u003e\\n \\u003cli\\u003eVelamuri, A. et al. Reconfigurable multiband microwave filters using programmable photonic integrated circuit. \\u003cem\\u003eOptica Open.\\u003c/em\\u003e Preprint (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eTao, Z. et al. Highly reconfigurable silicon integrated microwave photonic filter towards next-generation wireless communication. \\u003cem\\u003ePhoton. Res.\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, 682 (2023).\\u003c/li\\u003e\\n \\u003cli\\u003eTao, Y. et al. Hybrid-integrated high-performance microwave photonic filter with switchable response. \\u003cem\\u003ePhoton. Res.\\u003c/em\\u003e \\u003cstrong\\u003e9\\u003c/strong\\u003e, 1569 (2021).\\u003c/li\\u003e\\n \\u003cli\\u003eXu, L. et al. Silicon-on-insulator-based microwave photonic filter with widely adjustable bandwidth. \\u003cem\\u003ePhoton. Res.\\u003c/em\\u003e \\u003cstrong\\u003e7\\u003c/strong\\u003e, 110 (2019).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, X. et al. Silicon-on-insulator-based microwave photonic filter with narrowband and ultrahigh peak rejection. \\u003cem\\u003eOpt. Lett.\\u003c/em\\u003e \\u003cstrong\\u003e43\\u003c/strong\\u003e, 1359\\u0026ndash;1362 (2018).\\u003c/li\\u003e\\n \\u003cli\\u003eSong, S. et al. Tunable single-passband microwave photonic filter based on integrated optical double notch filter. \\u003cem\\u003eJ. Lightwave Technol.\\u0026nbsp;\\u003c/em\\u003e\\u003cstrong\\u003e36\\u003c/strong\\u003e, 4557\\u0026ndash;4564 (2018).\\u003c/li\\u003e\\n \\u003cli\\u003eZhang, L. et al. Ultralow-loss silicon photonics beyond the singlemode regime. \\u003cem\\u003eLaser Photonics Rev.\\u003c/em\\u003e \\u003cstrong\\u003e16\\u003c/strong\\u003e, 2100292 (2022).\\u003c/li\\u003e\\n \\u003cli\\u003ePorzi, C. et al. Silicon-on-insulator microwave photonic filter with widely tunable and reconfigurable flat-top bandpass functionality. \\u003cem\\u003eJ. Lightwave Technol.\\u0026nbsp;\\u003c/em\\u003e\\u003cstrong\\u003e40\\u003c/strong\\u003e, 6666\\u0026ndash;6675 (2022).\\u003c/li\\u003e\\n \\u003cli\\u003eZhuang, L., Roeloffzen, C. G. H., Hoekman, M., Boller, K.-J. \\u0026amp; Lowery, A. J. Programmable photonic signal processor chip for radiofrequency applications. \\u003cem\\u003eOptica\\u003c/em\\u003e,\\u0026nbsp;\\u003cstrong\\u003e2\\u003c/strong\\u003e, 854\\u0026ndash;859 (2015).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, Y. et al. All-optimized integrated RF photonic notch filter. \\u003cem\\u003eOpt. Lett.\\u003c/em\\u003e \\u003cstrong\\u003e42\\u003c/strong\\u003e, 4631\\u0026ndash;4634 (2017).\\u003c/li\\u003e\\n \\u003cli\\u003eJiang, H., Yan, L. \\u0026amp; Marpaung, D. Chip-based arbitrary radio-frequency photonic filter with algorithm-driven reconfigurable resolution. \\u003cem\\u003eOpt. Lett.\\u003c/em\\u003e \\u003cstrong\\u003e43\\u003c/strong\\u003e, 415 (2018).\\u003c/li\\u003e\\n \\u003cli\\u003eCheng, W. et al. Tunable bandpass microwave photonic filter with largely reconfigurable bandwidth and steep shape factor based on cascaded silicon nitride micro-ring resonators. \\u003cem\\u003eOpt. Express\\u003c/em\\u003e \\u003cstrong\\u003e31\\u003c/strong\\u003e, 25648 (2023).\\u003c/li\\u003e\\n \\u003cli\\u003eGuzzon, R. S. et al.\\u0026nbsp;Integrated InP-InGaAsP tunable coupled ring optical bandpass filters with zero insertion loss. \\u003cem\\u003eOpt. Express\\u003c/em\\u003e \\u003cstrong\\u003e19\\u003c/strong\\u003e, 7816\\u0026ndash;7826 (2011).\\u003c/li\\u003e\\n \\u003cli\\u003eNorberg, E. et al. A. Programmable photonic microwave filters monolithically integrated in InP\\u0026ndash;InGaAsP. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e29\\u003c/strong\\u003e, 1611\\u0026ndash;1619 (2011).\\u003c/li\\u003e\\n \\u003cli\\u003eWei, C. et al. Programmable multifunctional integrated microwave photonic circuit on thin-film lithium niobate. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e16\\u003c/strong\\u003e, 2281 (2025).\\u003c/li\\u003e\\n \\u003cli\\u003eMarpaung, D. et al. Low-power, chip-based stimulated Brillouin scattering microwave photonic filter with ultrahigh selectivity. \\u003cem\\u003eOptica\\u003c/em\\u003e \\u003cstrong\\u003e2\\u003c/strong\\u003e, 76\\u0026ndash;83 (2015).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, Y. et al. Integration of Brillouin and passive circuits for enhanced radio-frequency photonic filtering. APL Photonics\\u0026nbsp;\\u003cstrong\\u003e4\\u003c/strong\\u003e, 106103 (2019).\\u003c/li\\u003e\\n \\u003cli\\u003eGarrett, M. et al. Multi-band and frequency-agile chip-based RF photonic filter for ultra-deep interference rejection. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e40\\u003c/strong\\u003e, 1672\\u0026ndash;1680 (2022).\\u003c/li\\u003e\\n \\u003cli\\u003eShi, F. et al. A microwave photonic channelized receiver with self-interference cancellation. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e41\\u003c/strong\\u003e, 627\\u0026ndash;636 (2023).\\u003c/li\\u003e\\n \\u003cli\\u003eYu, Y. et al. ANN-assisted high-resolution and large dynamic range temperature sensor based on simultaneous microwave photonic and optical measurement. \\u003cem\\u003eIEEE Sensors Journal\\u003c/em\\u003e \\u003cstrong\\u003e23\\u003c/strong\\u003e, 1105\\u0026ndash;1114 (2023).\\u003c/li\\u003e\\n \\u003cli\\u003eWang, L. et al. Parity\\u0026ndash;time-symmetric optoelectronic oscillator based on stimulated brillouin scattering. \\u003cem\\u003eACS Photonics\\u003c/em\\u003e \\u003cstrong\\u003e10\\u003c/strong\\u003e, 2308\\u0026ndash;2315 (2023).\\u003c/li\\u003e\\n \\u003cli\\u003ePan, Z. et al. Frequency-stabilized optoelectronic oscillator with simultaneous wideband tunability. \\u003cem\\u003eIEEE Trans. Microwave Theory Techn.\\u003c/em\\u003e 1\\u0026ndash;8 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eTripathi, S. et al. Millimeter-wave and terahertz spectrum for 6g wireless. \\u003cem\\u003e6G Mobile Wireless Networks.\\u003c/em\\u003e 83\\u0026ndash;121 (2021).\\u003c/li\\u003e\\n \\u003cli\\u003eGe, Z. et al. Millimeter-wave wide-band bandpass filter in CMOS technology using a two-layered highpass-type approach with embedded upper stopband. \\u003cem\\u003eIEEE Trans. Circuits Syst. II: Exp.s Briefs\\u003c/em\\u003e \\u003cstrong\\u003e68\\u003c/strong\\u003e, 1586\\u0026ndash;1590 (2021).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, L. et al. Narrow passband tunable optical filter based on silicon high-Q rings assisted mzi structure. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e42\\u003c/strong\\u003e, 2049\\u0026ndash;2056 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eYan, H. et al. Wideband-tunable on-chip microwave photonic filter with ultrahigh-Q U-bend-Mach\\u0026ndash;Zehnder-interferometer-coupled microring resonators. \\u003cem\\u003eLaser Photonics Rev.\\u003c/em\\u003e \\u003cstrong\\u003e17\\u003c/strong\\u003e, 2300347 (2023).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, Y. et al. Tunable and reconfigurable microwave photonic bandpass filter based on cascaded silicon microring resonators. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e40\\u003c/strong\\u003e, 4655\\u0026ndash;4662 (2022).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, Y. et al. Reconfigurable microwave photonic bandpass filter based on CROW. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e42\\u003c/strong\\u003e, 1597\\u0026ndash;1604 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eLiu, Y. et al. Reconfigurable bandpass microwave photonic filter based on cascaded second-order CROWs. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e42\\u003c/strong\\u003e, 5599-5605 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eZhang, L. et al. Cascading second-order microring resonators for a box-like filter response. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e35\\u003c/strong\\u003e, 5347\\u0026ndash;5360 (2017).\\u003c/li\\u003e\\n \\u003cli\\u003eP\\u0026eacute;rez-L\\u0026oacute;pez et al.\\u0026nbsp;Multipurpose self-configuration of programmable photonic circuits. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e11\\u003c/strong\\u003e, 6359 (2020).\\u003c/li\\u003e\\n \\u003cli\\u003eKennedy, J. \\u0026amp; Eberhart, R. Particle swarm optimization.\\u0026nbsp;In \\u003cem\\u003eProceedings of ICNN\\u0026rsquo;95 - International Conference on Neural Networks\\u003c/em\\u003e \\u003cstrong\\u003e4\\u003c/strong\\u003e, 1942\\u0026ndash;1948 (IEEE, 1995).\\u003c/li\\u003e\\n \\u003cli\\u003eHolland, J. H. Genetic algorithms. \\u003cem\\u003eScientific American\\u003c/em\\u003e \\u003cstrong\\u003e267\\u003c/strong\\u003e, 66\\u0026ndash;73 (1992).\\u003c/li\\u003e\\n \\u003cli\\u003eS. Kirkpatrick et al. Optimization by simulated annealing. \\u003cem\\u003eScience\\u003c/em\\u003e \\u003cstrong\\u003e220\\u003c/strong\\u003e,671-680(1983).\\u003c/li\\u003e\\n \\u003cli\\u003ePadmaraju, K. \\u0026amp; Bergman, K. Resolving the thermal challenges for silicon microring resonator devices. \\u003cem\\u003eNanophotonics\\u003c/em\\u003e \\u003cstrong\\u003e3\\u003c/strong\\u003e, 269\\u0026ndash;281 (2014).\\u003c/li\\u003e\\n \\u003cli\\u003eSheikholeslami, A. Power supply noise [Circuit Intuitions]. \\u003cem\\u003eIEEE Solid-State Circuits Mag.\\u003c/em\\u003e \\u003cstrong\\u003e12\\u003c/strong\\u003e, 15\\u0026ndash;17 (2020).\\u003c/li\\u003e\\n \\u003cli\\u003eBohannon, E. et al. Passive and active reduction techniques for on-chip high-frequency digital power supply noise. \\u003cem\\u003eIEEE Transactions on Very Large Scale Integration (VLSI) Systems\\u003c/em\\u003e \\u003cstrong\\u003e18\\u003c/strong\\u003e, 157\\u0026ndash;161 (2010).\\u003c/li\\u003e\\n \\u003cli\\u003eZhang, L. et al. Beyond 100-Gb/s transmission over 80-km SMF using direct-detection SSB-DMT at C-band. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e34\\u003c/strong\\u003e, 723\\u0026ndash;729 (2016).\\u003c/li\\u003e\\n \\u003cli\\u003eBo, T. et al. Optical single-sideband transmitters. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e41\\u003c/strong\\u003e, 1163\\u0026ndash;1174 (2023).\\u003c/li\\u003e\\n \\u003cli\\u003eChen, X. et al. Kramers\\u0026ndash;Kronig receivers for 100-km datacenter interconnects. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e36\\u003c/strong\\u003e, 79\\u0026ndash;89 (2018).\\u003c/li\\u003e\\n \\u003cli\\u003eTang, X. et al. Digital pre- and post-equalization for C-band 112-Gb/s PAM4 short-reach transport systems. \\u003cem\\u003eJ. Lightwave Technol.\\u003c/em\\u003e \\u003cstrong\\u003e38\\u003c/strong\\u003e, 4683\\u0026ndash;4690 (2020).\\u003c/li\\u003e\\n \\u003cli\\u003eNi, W. et al. LUT enabled low-complexity electrical dispersion pre-compensation scheme for IM-DD systems. \\u003cem\\u003eOpt. Lett.\\u003c/em\\u003e \\u003cstrong\\u003e49\\u003c/strong\\u003e, 6417\\u0026ndash;6420 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eChoi, H.-W., Choi, S. \\u0026amp; Kim, C.-Y. A CMOS band-pass low noise amplifier with excellent gain flatness for mm-Wave 5G communications.\\u0026nbsp;In \\u003cem\\u003eIEEE/MTT-S International Microwave Symposium (IMS)\\u003c/em\\u003e 329\\u0026ndash;332 (IEEE, 2020).\\u003c/li\\u003e\\n \\u003cli\\u003eSewiolo, B., Fischer, G. \\u0026amp; Weigel, R. A 30 GHz variable gain amplifier with high output voltage swing for ultra-wideband radar. \\u003cem\\u003eIEEE Microw. Wireless Compon. Lett\\u003c/em\\u003e.\\u0026nbsp;\\u003cstrong\\u003e19\\u003c/strong\\u003e, 590\\u0026ndash;592 (2009).\\u003c/li\\u003e\\n \\u003cli\\u003eZhao, W. et al. Ultracompact silicon on-chip polarization controller. \\u003cem\\u003ePhoton. Res.\\u003c/em\\u003e \\u003cstrong\\u003e12\\u003c/strong\\u003e, 183 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eSun, S. et al. Inverse design of ultra-compact multimode waveguide bends based on the free-form curves. \\u003cem\\u003eLaser Photonics Rev.\\u003c/em\\u003e \\u003cstrong\\u003e15\\u003c/strong\\u003e, 2100162 (2021).\\u003c/li\\u003e\\n \\u003cli\\u003eCheng, J. et al. Multimodal deep learning using on-chip diffractive optics with in situ training capability. \\u003cem\\u003eNat. Commun.\\u003c/em\\u003e \\u003cstrong\\u003e15\\u003c/strong\\u003e, 6189 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eZhang, W. et al. A system-on-chip microwave photonic processor solves dynamic RF interference in real time with picosecond latency.\\u0026nbsp;\\u003cem\\u003eLight Sci. Appl.\\u003c/em\\u003e \\u003cstrong\\u003e13\\u003c/strong\\u003e, 14 (2024).\\u003c/li\\u003e\\n \\u003cli\\u003eShi, T., Chen, Y. \\u0026amp; Yao, J. Seamlessly merging radar ranging and imaging, wireless communications, and spectrum sensing for 6G empowered by microwave photonics. \\u003cem\\u003eCommun. Eng.\\u003c/em\\u003e \\u003cstrong\\u003e3\\u003c/strong\\u003e, 130 (2024).\\u003c/li\\u003e\\n\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":true,\"hideJournal\":true,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":false,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true},\"keywords\":\"\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-6669261/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-6669261/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003eIntegrated microwave photonic filters (IMPFs) have emerged as promising candidates for advanced microwave systems owing to their distinctive combination of wide operational bandwidth, reconfigurable functionality, and compact size. Nevertheless, the complex and time-consuming manual manipulation of IMPFs remains a significant impediment to their widespread applications. Here, to the best of our knowledge, we experimentally demonstrate the first intelligent configuration of IMPF featuring wideband center frequency tunability, flexible bandwidth reconfigurability, self-stabilization, and excellent channel equalization simultaneously. The configuration strategy is enabled by our proposed universal hybrid collaboration algorithm, which can fully unleash the hardware potential of the optical device, thus enabling comprehensive synergy of multiple properties. Results show that the center frequency of IMPF is tuned from 2 to 48 GHz, and the bandwidth is reconfigured from 0.66 to 4.15 GHz, with a rejection ratio of up to 37.67 dB. The roll-off rate and shape factor reach as high as 17.50 dB/GHz and 0.78, respectively. Meanwhile, the maximum center frequency drift of IMPF in 3 hours is reduced from 11.950 to 0.051 GHz even without a thermo-electric cooler, indicating that the center frequency stability is enhanced by 234 times. The passband shape of the IMPF can also be dynamically adjusted to equalize frequency-dependent fading, achieving up to 2.42 dB compensation of intra-channel fading. Our work highlights the potential of IMPFs based on intelligent configuration, unlocking new avenues for practical applications of microwave photonic signal processing.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Intelligent configuration of integrated microwave photonic filter with programmable response and self-stabilization\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-05-23 15:04:02\",\"doi\":\"10.21203/rs.3.rs-6669261/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true}}],\"origin\":\"\",\"ownerIdentity\":\"a1c6cf70-6a83-4ff9-b6ab-f2eaeecddacb\",\"owner\":[],\"postedDate\":\"May 23rd, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"posted\",\"subjectAreas\":[{\"id\":48558134,\"name\":\"Physical sciences/Optics and photonics/Applied optics/Microwave photonics\"},{\"id\":48558135,\"name\":\"Physical sciences/Optics and photonics/Applied optics/Integrated optics\"}],\"tags\":[],\"updatedAt\":\"2025-05-31T04:45:28+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2025-05-23 15:04:02\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-6669261\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-6669261\",\"identity\":\"rs-6669261\",\"version\":[\"v1\"]},\"buildId\":\"8U1c8b4HqxoKbykW_rLl7\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}