Pump-locked random lasers using artificial intelligence

Pump-locked random lasers using artificial intelligence | 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 Pump-locked random lasers using artificial intelligence Tianrui Zhai, Junhua Tong, Xiaoyu Shi, Zhiyang Xu, Naeem Iqbal, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5425695/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 Random lasing in complex disordered systems demonstrates efficient lasing across multiple localized modes. Accurate control of these modes could enable random lasers to function as fast-switching, multifunctional light sources. In this work, an unpredictable random laser was prepared and artificial intelligence (AI) technology was used for taming the lasing action. Lasing modes were precisely controlled by programming the spatial shape of the pump profile. Genetic algorithms were introduced, allowing any mode within the complex system to be extracted by setting a target value associated with that mode. Based on experimental results, the interaction model of pumping cells with lasing modes was locked. This work advances programmable random lasers, enhancing their potential for practical applications in signal processing, spectral sensing, communication, and optical computing. Physical sciences/Optics and photonics/Lasers, LEDs and light sources Physical sciences/Optics and photonics/Optical physics/Micro-optics Random lasers artificial intelligence mode selection pump-locked Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Random lasers, characterized by low spatial coherence, 1 multi-modal operation, 2 intensity fluctuations, 3 and diverse physical phenomena, 4 exhibit potential for applications in speckle-free imaging, 5 biosensing, 6 super-resolution spectroscopy, 7 and cross-disciplinary research platforms. 8 However, limited controllability over lasing modes, such as emission frequency, threshold, and direction, restricts the application of random lasers in spectral sensing, signal processing, optical computing, and communication. Directional random lasing has been achieved by incorporating directional components, such as optical microcavities (fiber, waveguide, distributed Bragg reflectors, spherical microcavities). 9 , 10 , 11 , 12 Control over random lasing frequency and threshold is mainly exerted by manipulating gain materials and scattering paths. 13 , 14 , 15 However, these methods typically yield only an overall shift in lasing frequency. 16 Manually extracting specific modes from the numerous lasing modes is highly challenging due to strong inter-mode correlations in complex disordered systems. With the rapid advancement of AI technology, its interdisciplinary integration has expanded scientific research boundaries. Notably, optics, as a foundational research field, is at the forefront of this integration. Currently, AI and optics converge primarily in two areas: empowering AI with optics, where photons replace traditional electronics for more efficient AI computing, 17 and empowering optics with AI, using intelligent algorithms to accelerate the design and optimization of optical devices. 18 Deep learning models, such as multi-layer perceptron neural networks, enable the rapid design and characterization of micro-nano structures. 19 In summary, AI algorithms now facilitate researchers in bypassing inefficient processes, enabling everything from theoretical model integration to complex simulations of optical phenomena and intelligent light field analysis. Theoretical evidence suggests that shaping the pump profile can precisely extract specific modes from random lasers. 20 , 21 Using genetic algorithms (GA) in AI, we can now demonstrate the mode extraction process from an experimental perspective. Sebbah’s research group first reported this process in precisely structured systems, such as one-dimensional random systems. 12 , 23 , 24 Recently, Saxena et al. applied AI to control lasing performance in a designed network system. 25 However, in a complex system lacking predefined design, it remains challenging to extract specific modes and explore their corresponding optical paths. In this study, we successfully extracted specific lasing modes within a disordered system using genetic algorithms. Notably, the relationship between pump cells and lasing modes could be established by comparing the pump profiles of corresponding modes. This approach enables effective prediction of optical paths for random lasing generation, providing a means to determine the generation principle of random lasing action and advancing the performance customization of random lasers. 2. Design principles Random lasing is boosted by multiple and recurrent scattering of light in gain medium. The pump profile determines the stimulated region within the random system, influencing optical scattering feedback and gain paths, which in turn control the lasing modes. A disordered membrane was fabricated using a simple spin-coating method (see Methods and Fig. S1 ). Typically, a digital micromirror device (DMD) is used to shape pump profiles. However, manually importing pump patterns makes it challenging to extract specific modes from the many lasing modes. Thus, intelligent methods are required. Figure 1 a-b illustrates the principle of mode selection in random lasers using our intelligent control system. The core of the intelligent control system is a self-developed software comprising a genetic algorithm (GA), DMD control program, and spectral acquisition program (Fig. S2 ). The genetic algorithm is an effective method for global target searching. 26 Based on a predefined target mode (λ t ) and initial spectral data under a uniform pump profile, the GA optimization program infers the required pump profile for the next cycle. By accessing the DMD program interface, the DMD control program converts the 2D pump profile image data into 1D data and transmits it to the DMD. Based on the image grayscale, the program autonomously controls the DMD reflectors to adjust the pumping light, thereby controlling the pump pattern. The spectral acquisition program, based on the preset parameters, acquires random lasing spectra excited by the pump pattern through spectrometer control. Through iterative optimization, the target modes are obtained, as shown in Fig. 1 c. The corresponding pump patterns of selected modes can then be locked and analyzed. This enables prediction of possible optical paths for the random lasing modes. 3. Results and discussions As a well-known intelligent algorithm, GA includes key components: population, fitness selection, and evolutionary operators, as shown in Fig. 2a. These evolutionary operators include selection, crossover, and mutation. A population of possible solutions is employed to search the solution space. Each solution is encoded as a character string, referred to as a genome. In each iteration, the fitness values of genomes are evaluated, and those with higher fitness have an increased probability of survival. Surviving genomes combine to form “child” genomes through a process called crossover. In addition, genomes may undergo mutation. Finally, a new iteration is formed by recombining and replacing the initial population. This process repeats to yield stronger solutions over successive iterations until the target solution is found. In our work, the population comprises image data representing possible pump profiles. The fitness function assigns a value to each genome in the population. The fitness function is defined as follows: $$\:f\left({\lambda\:}_{t}\right)=\frac{Max\left\{{I}_{{\lambda\:}_{s1}},{I}_{{\lambda\:}_{s2}},\dots\:,{I}_{{\lambda\:}_{sn}}\right\}}{Min\left\{{I}_{{\lambda\:}_{t1}},{I}_{{\lambda\:}_{t2}},\dots\:,{I}_{{\lambda\:}_{tm}}\right\}}$$ Here, \(\:{\lambda\:}_{t}\) and \(\:{\lambda\:}_{s}\) represent the modes to be selected and suppressed, respectively. Where m is the number of target modes (m ≥ 1), and n is the number of suppressed modes (n ≥ 0). Max denotes the maximum peak intensity among suppressed modes, while Min denotes the minimum peak intensity among target modes. Thus, a smaller f ( \(\:{\lambda\:}_{t}\) ) value in the optimization process indicates greater dominance of the target modes. The target modes are selected when f ( \(\:{\lambda\:}_{t}\) ) converges to an ideal value (≤ 0.5). Figure 2b shows the evolution of f ( \(\:{\lambda\:}_{t}\) ) across iteration cycles for selecting a specific mode. The value converges to approximately 0.21. The inset displays the optimization process of pump profiles. When excited by a uniform pump profile, random lasing with multiple modes is observed, as shown in the upper panel of Fig. 2c. When excited by the optimized pump profile, the target mode is accurately selected, and the other modes are effectively suppressed, as shown in the lower panel of Fig. 2c. Once the pump profile is determined, the target mode can be accurately selected. As a result, the emission characteristics of a single mode versus multimode operation were explored in this work, as shown in Fig. 2c-d. Random lasing with multiple modes was obtained by exciting the sample with a uniform pump profile, as shown in the upper panel of Fig. 2c. As pump power density increases, the emission spectrum evolves from a typical fluorescent spectrum to a multimode random lasing spectrum. To verify the effect of mode selection, the optimized pump profile was applied to excite the sample. The lower panel of Fig. 2c shows emission spectra at different pump power densities, yielding a stable, pure single mode of random lasing. Furthermore, the threshold behavior for a specific mode, pumped by both uniform and optimized profiles, is shown in Fig. 2d. The threshold at 589 nm with optimized pumping is approximately 0.55 MW/cm², nearly half of that with uniform pumping. Overall, these results demonstrate that our proposed intelligent control system not only accurately extracts modes but also significantly enhances pumping efficiency. To further test the mode extraction capability of our proposed intelligent control system, we attempted to extract arbitrary single modes and multimodes, as shown in Fig. 3. When pumped by a uniform profile at higher pump power density, a random lasing spectrum with multiple modes is obtained, as shown in Fig. 3a. Keeping all other conditions unchanged, we set a single mode (590.5 nm), two modes (587.0 nm and 588.1 nm), and three modes (587.0 nm, 588.1 nm, and 590.5 nm) in the original spectrum as target modes, respectively. We adjusted the search parameters in the GA optimization program to control the pump pattern in the DMD until the target modes were obtained. The results indicate that any target mode within a complex disordered system can be accurately extracted using our proposed intelligent control system, even the adjacent modes and modes that are at a competitive disadvantage, such as those at 587.0 nm and 588.1 nm. For random lasers, the pump pattern determines the excited region of the sample, affecting the optical paths for lasing generation. Thus, locking the role of each pump cell in the lasing mode provides an effective method for exploring optical feedback paths. In our work, two modes (λ 1 , λ 2 ) were selected individually or simultaneously for analysis, as shown in Fig. 4. A video recording of the optimization process is provided in the supporting information. The spectra and corresponding pump patterns for individual modes λ 1 and λ 2 are shown in Fig.4a. In contrast, Fig.4c shows the spectrum and corresponding pump pattern for λ 1 and λ 2 simultaneously. For comparison, the linear superposition of pump patterns for λ 1 and λ 2 was also calculated, as shown in Fig. 4a, where the number of pump cells is significantly higher and can cover the combined pattern (λ 1 +λ 2 ) in Fig.4c. Based on these results, we analyzed the effect of pump cells on λ 1 and λ 2 selection, as shown in Fig. 4b. Cells in gray (SS) simultaneously suppress λ 1 and λ 2 . Green cells (EE) simultaneously enhance λ 1 and λ 2 , while red cells (ES) enhance λ 1 but suppress λ 2 . Orange cells (E-) enhance λ 1 with no effect on λ 2 , while blue cells (SE) suppress λ 1 but enhance λ 2 . Cyan cells (-E) enhance λ 2 with no effect on λ 1 . That is to say, the pump-locked random laser for certain mode is achieved at the first time. The coupling relationship between modes can be predicted based on locked pump cells and mode intensities. When extracting single modes of λ 1 and λ 2 in turn, the illuminated cells counted up to 144 and 136, respectively, as shown in Fig. 4a. Clearly, λ 1 is easier to extract with strong intensity, indicating it have relative dominant optical feedback paths. λ 2 is more difficult to extract alone from tens of modes in the initial spectrum, suggesting weaker competitiveness and significant overlap in optical feedback paths with competing modes. When extracting the two modes simultaneously, the illuminated cells are counted up to 129, as shown in Fig. 4c. Compare with Fig. 4a, it is easier to extract the two modes than any single one with higher intensity, especially for the λ 2 . All the results indicate there is an overlap between their optical feedback paths, which has more influence on λ 2 . This work thus provides an effective method for analyzing the coupling relationships of random lasing modes. 4. Conclusions A powerful tool for mode selection in an unpredictable random system is first demonstrated. By building intelligent pumping control system based on GA, any expected modes can be extracted accurately and efficiently, while the effect of pump cells on specific modes can be locked. Based on the locked pump cells and corresponding mode intensity, the coupling relationship between modes can be deduced. The work breaks through the barrier of quantitatively controlling modes in complex disordered systems and could guide the reverse design of photonic devices. 5. Methods Sample preparation . The disordered system is a polymer film fabricated by dye-doped PMMA with scattering nanoparticles. Here, the typical laser dye RhB is used as the gain material, while Au nanorods (NRs) with diameters and lengths of approximately 25 and 50 nm, respectively, act as scatterers. The Au NRs provide coherent feedback in the polymer film and enhance RhB emission due to localized surface plasmon resonance in the local electric field. The fabrication process is illustrated in Fig. S1 . First, the Au NRs (0.02 mg/mL) were spin-coated onto the flexible PET substrate at 1800 rpm for 40 seconds. Second, RhB (6 mg/mL) and PMMA (200 mg/mL) were mixed at a 1:1 volume ratio and magnetically stirred for 20 minutes. Then, the mixture was spin-coated onto the Au NR layer at 1800 rpm for 40 seconds. Third, the structure was heated at 70°C for 20 minutes to achieve solidification. Finally, the disordered system was obtained. Dichloromethane was used as the solvent in all steps. Intelligent control system. The core of the intelligent control system is a self-developed software including GA optimization program, DMD control program, and spectral acquisition program. The model-view-controller (MVC) architecture is used in our self-developed software, which can separate the interface from the model. The modular design lays a good foundation for the expansion of the software functions, including debug, optimize, and replace the algorithm. The software interface mainly contains parts of spectrum display, parameter setting, regulation and control. The spectrum display area displays the current spectrum data. The parameter setting area for spectrum acquisition can set the smoothing points of the spectrometer, the spectrum acquisition time and the spectrum average number. The parameter setting area for spectrum frequency can set the frequency need to be enhanced and suppressed. The regulation and control area is used to start or stop the control function. Optical measurement. The diagram of test setup based on intelligent control system is illustrated in Fig. S2 . A doubled Q-switched Nd:YAG laser (532 nm, 5–7 ns, 10 Hz) is employed as the pump source. The lens 1 and lens 2 are used for expanding and collimating of pump beam, respectively. The DMD is used to shape the pump profile based on the feedback of grayscale of image. The modulated pump light is reflected to lens 3. The lens 3 is used to adjust the size of pump profile. The modulated pump pattern incident on the sample and the reflected emission light was collected by an optical fiber spectrometer (Ocean Optics HR4000) with a spectral resolution of 0.02 nm. The pump power was controlled by pump voltage. The pump source, digital micromirror device (DMD) and high-resolution spectrometer are linked using the self-development software. Declarations CRediT authorship contribution statement Junhua Tong: Experimental investigations, Device fabrication and characterization, Intelligent control system establishment, Software development, Writing-original draft. Xiaoyu Shi: Methodology & Writing-review. Zhiyang Xu: Supervision & optimization. Naeem Iqbal: Writing-review. Kun Ge: Assist sample preparation. Tianrui Zhai: Conceptualization, Investigation, Methodology, Supervision, Writing-review & editing. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships. Data availability The data that support the findings of this study are presented in Supplementary Information. This paper or available from the corresponding author on request. Code availability The software was written in Python 3.11 using the Python package ecosystem (PySide6, numpy, scipy, scikit-opt). The computer code or available from the corresponding author on request. Acknowledgements This work was financially supported by the National Natural Science Foundation of China (62375007) and the Beijing Natural Science Foundation (1232024). References Cao H, Chriki R, Bittner S, Friesem A (2019) N. Davidson. Complex lasers with controllable coherence. Nat Rev Phys 1:156 Sapienza R (2019) Determining random lasing action. Nat Rev Phys 1:690–695 Sapienza R (2022) Controlling random lasing action. Nat Phys 18:976 Du W, Hu L, Xia J, Zhang L, Li S, Kuai Y, Cao Z, Xu F, Liu Y, Zhou K, Xie K, Yu B, Raposo E, Gomes A, Hu Z (2024) Observation of the photonic Hall effect and photonic magnetoresistance in random lasers. Nat Commun 15(1):4589 Redding B, Choma M, Cao H (2012) Speckle-free laser imaging using random laser illumination. Nat Photonics 6(6):355–359 Xu Z, Hong Q, Ge K, Shi X, Wang X, Deng J, Zhou Z, Zhai T (2022) Random lasing from label-free living cells for rapid cytometry of apoptosis. Nano Lett 22(1):172–178 Boschetti A, Taschin A, Bartolini P, Tiwari AK, Pattelli L, Torre R (2020) Wiersma. Spectral super-resolution spectroscopy using a random laser. Nat Photonics 14:177 Ghofraniha N, Viola I, Di Maria F, Kumar Tiwari A, Pattelli L, Torre R, Diederik S (2015) Wiersma. Experimental evidence of replica symmetry breaking in random lasers. Nat Commun 6:6058 Shi X, Shen K, Bian Y, Song W, Ruan J, Wang Z, Zhai T (2023) Programmable complex pumping field induced color on demand random lasing in fiber-integrated microbelts for speckle free imaging. Sci China Inf Sci 66:222401 Zhai T, Zhang X, Pang Z, Su X, Liu H, Feng S, Wang L (2011) Random laser based on waveguided plasmonic gain channels. Nano Lett 11(10):4295–4298 Liu Y, Yang W, Xiao S, Zhang N, Fan Y, Qu G, Song Q (2019) Surface-emitting perovskite random lasers for speckle-free imaging. ACS Nano 13:10653 Lee K, Ma H, Rotermund F et al (2021) Non-resonant power-efficient directional Nd:YAG ceramic laser using a scattering cavity. Nat Commun 12:8 Adl HP, Sánchez-Díaz J, Vescio G, Cirera A, Garrido B, Pacheco FAV, Żuraw W, Przypis Ł, Öz S, Mora-Seró I, Martínez-Pastor JP (2024) Suárez. Tailoring single-mode random lasing of tin halide perovskites integrated in a vertical cavity. Adv Mater 36:2313252 Li X, Wang Y, Sun H, Zeng H (2017) Amino-mediated anchoring perovskite quantum dots for stable and low-threshold random lasing. Adv Mater 29:1701185 Bian Y, Xue H, Wang Z (2021) Programmable random lasing pluses based on waveguide-assisted random scattering feedback. Laser Photonics Rev 15(4):2000506 Gomes ASL, Moura AL, de Araújo CB et al (2021) Recent advances and applications of random lasers and random fiber lasers. Prog Quantum Electron 78:100343 Wetzstein G, Ozcan A, Gigan S, Fan S, Englund D, Soljačić M, Denz C, Miller D (2020) Psaltis. Inference in artificial intelligence with deep optics and photonics. Nature 588(7836):39–47 Wang N, Yan W, Qu Y, Ma S, Li S (2021) Qiu. Intelligent designs in nanophotonics: from optimization towards inverse creation. PhotoniX 2:1–35 Jiang J, Chen M, Fan J (2021) Deep neural networks for the evaluation and design of photonic devices. Nat Rev Mater 6(8):679–700 Vanneste C, Sebbah P (2001) Selective excitation of localized modes in active random media. Phys Rev Lett 87:183903 Cao H, Mosk A, Rotter S (2022) Shaping the propagation of light in complex media. Nat Phys 18:994 Bachelard N, Andreasen J, Gigan S, Sebbah P (2012) Taming random lasers through active spatial control of the pump. Phys Rev Lett 109:033903 Bachelard N, Gigan S, Noblin X (2014) Sebbah. Adaptive pumping for spectral control of random lasers. Nat Phys 10:426 Kumar B, Homri R, Maurya S, Lebental M (2021) Sebbah. Localized modes revealed in random lasers. Optica 8:1033 Saxena D, Arnaudon A, Cipolato O, Gaio M, Quentel A, Yaliraki S, Pisignano D, Camposeo A, Barahona M, Sapienza R (2022) Sensitivity and spectral control of network lasers. Nat Commun 13:6493 Katoch S, Chauhan S, Kumar V (2021) A review on genetic algorithm: past, present, and future. Multimed Tools Appl 80:8091–8126 Additional Declarations There is NO Competing Interest. Supplementary Files 591.5nmX594nmoptimizationprocess.mp4 The video of optimization process for mode at 591.5 nm&594 nm 594nmoptimizationprocess.mp4 The video of optimization process for mode at 594 nm 591.5nmoptimizationprocess.mp4 The video of optimization process for mode at 591.5 nm maincode.pdf The main code for our proposed software SupplementaryInformation.docx Supplementary Information 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-5425695","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":376529792,"identity":"ed9226fe-f8bc-4b15-9bcf-086945c57be2","order_by":0,"name":"Tianrui Zhai","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzUlEQVRIiWNgGAWjYHACNgYGA5sExgYomyDggWhJA2phJkkLw+EEBgZitdhLH3/24EfB+TzmGfkHGD6UHWbgn91AwBa+hHTDHoPbxYwzkhkYZ5w7zCBx5wABLTwMxyR4DG4nNgK1MPO2HWYwkEggpIWxTfKPwTmIlr/EaWFmk+YxOADRwkiUljNsbNIyBsnFjD2PDQ72nEvnkbhBQAt7D/szyTd/7PIM2xMfPvhRZi3HP4OAFjgwbGBgOMAAjihigTzxSkfBKBgFo2CkAQBGBTxOy5ewxwAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-1922-2386","institution":"Beijing University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Tianrui","middleName":"","lastName":"Zhai","suffix":""},{"id":376529793,"identity":"529479dc-0d52-4866-8485-59e3b20079ca","order_by":1,"name":"Junhua Tong","email":"","orcid":"","institution":"Beijing University of Chemical Technology","correspondingAuthor":false,"prefix":"","firstName":"Junhua","middleName":"","lastName":"Tong","suffix":""},{"id":376529794,"identity":"afefafc2-f629-4851-84d3-e3e7233cbcd3","order_by":2,"name":"Xiaoyu Shi","email":"","orcid":"","institution":"Beijing University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Xiaoyu","middleName":"","lastName":"Shi","suffix":""},{"id":376529795,"identity":"46d59d3d-628f-4255-925a-1014c4df7baa","order_by":3,"name":"Zhiyang Xu","email":"","orcid":"","institution":"Beijing University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Zhiyang","middleName":"","lastName":"Xu","suffix":""},{"id":376529796,"identity":"770f1197-d80a-4838-ab3c-e37e3fbccd87","order_by":4,"name":"Naeem Iqbal","email":"","orcid":"","institution":"Beijing University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Naeem","middleName":"","lastName":"Iqbal","suffix":""},{"id":376529797,"identity":"4a3b49fe-a6e8-4e01-bba9-3d10353753f5","order_by":5,"name":"Kun Ge","email":"","orcid":"","institution":"Beijing University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Kun","middleName":"","lastName":"Ge","suffix":""}],"badges":[],"createdAt":"2024-11-10 11:40:38","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5425695/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5425695/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":70660297,"identity":"bf2ecbba-2d45-4bab-8f8c-192878293339","added_by":"auto","created_at":"2024-12-05 10:40:26","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":589512,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrinciple of mode selection in random lasers using an intelligent control system.\u003c/strong\u003e (a) Diagram of pump control with a uniform pump profile to obtain the initial spectrum. (b) Diagram of intelligent pump control by the DMD to obtain the target mode, represented by λ\u003csub\u003et\u003c/sub\u003e. (c) Selected random lasing modes with their corresponding pump patterns.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/190b9b59f7fc102fdfc885e9.png"},{"id":70659322,"identity":"7129c74a-7308-4cbe-8044-6fceb5a94e2a","added_by":"auto","created_at":"2024-12-05 10:32:26","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":573980,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle-mode selection operation.\u003c/strong\u003e (a) Genetic algorithm applied for pump pattern optimization within our self-developed software. (b) Convergence of the optimization process: score evolution across iteration cycles for single-mode optimization. Here, 50 score values are selected per iteration cycle, and the pump profile with the optimal score is used for the subsequent cycle. Gray values are set to either 0 (black) or 255 (green). (c) Emission spectra under varying pump power densities, excited by a uniform pump profile (upper panel) and an optimized pump profile (lower panel). (d) Emission intensity of random lasing as a function of pumping fluence for both the uniform and optimized pump profiles. The corresponding uniform and optimized pump profiles, as displayed on the DMD, are shown in the inset of part (c).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/d389dbd699d55b4c8e6c52f2.png"},{"id":70659330,"identity":"14ae0fcd-0f79-421b-9881-dd374318f5b1","added_by":"auto","created_at":"2024-12-05 10:32:27","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":617424,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExtraction of arbitrary single-mode or multimode lasing.\u003c/strong\u003e (a) Random lasing spectra with multiple modes excited by a uniform pump profile. (b) Random lasing spectrum with a single mode at 590.5 nm. (c) Random lasing spectrum with two modes at 587.0 nm and 588.1 nm. (d) Random lasing spectrum with three modes at 587.0 nm, 588.1 nm, and 590.5 nm. Insets show the corresponding pump profiles. Symbols ①, ②, and ③ indicate the modes at 590.5 nm, 588.1 nm, and 587.0 nm, respectively.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/b0919d7bb68d9a806f5a8a0b.png"},{"id":70660559,"identity":"830c3ad5-c753-4886-8197-3807cfbec12d","added_by":"auto","created_at":"2024-12-05 10:48:27","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":310182,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePump-locked analysis for random lasers.\u003c/strong\u003e (a) Random lasing spectra and corresponding pump patterns for (a) single modes at 591.5 nm (λ\u003csub\u003e1\u003c/sub\u003e), 594.0 nm (λ\u003csub\u003e2\u003c/sub\u003e), and (c) dual-mode lasing (λ\u003csub\u003e1 \u003c/sub\u003e\u0026amp; λ\u003csub\u003e2\u003c/sub\u003e) simultaneously. (b) Effect of pump cells on the selection of λ\u003csub\u003e1 \u003c/sub\u003eand λ\u003csub\u003e2\u003c/sub\u003e. Here, “S” indicates suppression and “E” indicates enhancement. The symbol “-” represents no effect on mode selection. For example, “ES” indicates that red pump cells enhance λ\u003csub\u003e1 \u003c/sub\u003ewhile suppressing λ\u003csub\u003e2\u003c/sub\u003e simultaneously.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/ed72cfbccaa0e2e2b3e39fd6.png"},{"id":77809054,"identity":"2abd1224-b714-4ff6-846c-54dd3a0d0602","added_by":"auto","created_at":"2025-03-05 17:44:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2928052,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/ce62d095-2e7b-482e-bf32-7b56b03362d1.pdf"},{"id":70659324,"identity":"e33fd677-db0b-44d7-acd6-a05f20da69a6","added_by":"auto","created_at":"2024-12-05 10:32:26","extension":"mp4","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":112345,"visible":true,"origin":"","legend":"The video of optimization process for mode at 591.5 nm\u0026594 nm","description":"","filename":"591.5nmX594nmoptimizationprocess.mp4","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/8f0e3820de3e134d7bcb6cb5.mp4"},{"id":70660298,"identity":"ae47d5fd-88b9-4919-a95b-5d634589f2e1","added_by":"auto","created_at":"2024-12-05 10:40:27","extension":"mp4","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":121479,"visible":true,"origin":"","legend":"\u003cp\u003eThe video of optimization process for mode at 594 nm\u003c/p\u003e","description":"","filename":"594nmoptimizationprocess.mp4","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/29fa0e5c3bae19f2dc2b166e.mp4"},{"id":70660299,"identity":"d58f50cf-bad9-4653-8d4b-82680482a696","added_by":"auto","created_at":"2024-12-05 10:40:27","extension":"mp4","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":118006,"visible":true,"origin":"","legend":"\u003cp\u003eThe video of optimization process for mode at 591.5 nm\u003c/p\u003e","description":"","filename":"591.5nmoptimizationprocess.mp4","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/12fbc2c82555fb4cdeaf9427.mp4"},{"id":70659325,"identity":"bfb1ef33-9bba-47ec-8a38-38ebc589a215","added_by":"auto","created_at":"2024-12-05 10:32:27","extension":"pdf","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":61918,"visible":true,"origin":"","legend":"\u003cp\u003eThe main code for our proposed software\u003c/p\u003e","description":"","filename":"maincode.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/c03c957761f8101c337b12e9.pdf"},{"id":70660307,"identity":"2fdd138d-1b66-4318-b9b5-c08884d717de","added_by":"auto","created_at":"2024-12-05 10:40:27","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":790790,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-5425695/v1/0768a01d7ba61a78c21a737a.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Pump-locked random lasers using artificial intelligence","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eRandom lasers, characterized by low spatial coherence,\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e multi-modal operation,\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e intensity fluctuations,\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e and diverse physical phenomena,\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e exhibit potential for applications in speckle-free imaging, \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e biosensing,\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e super-resolution spectroscopy, \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e and cross-disciplinary research platforms.\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e However, limited controllability over lasing modes, such as emission frequency, threshold, and direction, restricts the application of random lasers in spectral sensing, signal processing, optical computing, and communication. Directional random lasing has been achieved by incorporating directional components, such as optical microcavities (fiber, waveguide, distributed Bragg reflectors, spherical microcavities).\u003csup\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 Control over random lasing frequency and threshold is mainly exerted by manipulating gain materials and scattering paths.\u003csup\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\u003c/sup\u003e However, these methods typically yield only an overall shift in lasing frequency.\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e Manually extracting specific modes from the numerous lasing modes is highly challenging due to strong inter-mode correlations in complex disordered systems.\u003c/p\u003e \u003cp\u003eWith the rapid advancement of AI technology, its interdisciplinary integration has expanded scientific research boundaries. Notably, optics, as a foundational research field, is at the forefront of this integration. Currently, AI and optics converge primarily in two areas: empowering AI with optics, where photons replace traditional electronics for more efficient AI computing,\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e and empowering optics with AI, using intelligent algorithms to accelerate the design and optimization of optical devices.\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e Deep learning models, such as multi-layer perceptron neural networks, enable the rapid design and characterization of micro-nano structures.\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e In summary, AI algorithms now facilitate researchers in bypassing inefficient processes, enabling everything from theoretical model integration to complex simulations of optical phenomena and intelligent light field analysis.\u003c/p\u003e \u003cp\u003eTheoretical evidence suggests that shaping the pump profile can precisely extract specific modes from random lasers.\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e Using genetic algorithms (GA) in AI, we can now demonstrate the mode extraction process from an experimental perspective. Sebbah\u0026rsquo;s research group first reported this process in precisely structured systems, such as one-dimensional random systems.\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e Recently, Saxena et al. applied AI to control lasing performance in a designed network system.\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e However, in a complex system lacking predefined design, it remains challenging to extract specific modes and explore their corresponding optical paths. In this study, we successfully extracted specific lasing modes within a disordered system using genetic algorithms. Notably, the relationship between pump cells and lasing modes could be established by comparing the pump profiles of corresponding modes. This approach enables effective prediction of optical paths for random lasing generation, providing a means to determine the generation principle of random lasing action and advancing the performance customization of random lasers.\u003c/p\u003e"},{"header":"2. Design principles","content":"\u003cp\u003eRandom lasing is boosted by multiple and recurrent scattering of light in gain medium. The pump profile determines the stimulated region within the random system, influencing optical scattering feedback and gain paths, which in turn control the lasing modes. A disordered membrane was fabricated using a simple spin-coating method (see Methods and Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). Typically, a digital micromirror device (DMD) is used to shape pump profiles. However, manually importing pump patterns makes it challenging to extract specific modes from the many lasing modes. Thus, intelligent methods are required. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea-b illustrates the principle of mode selection in random lasers using our intelligent control system. The core of the intelligent control system is a self-developed software comprising a genetic algorithm (GA), DMD control program, and spectral acquisition program (Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). The genetic algorithm is an effective method for global target searching.\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e Based on a predefined target mode (λ\u003csub\u003et\u003c/sub\u003e) and initial spectral data under a uniform pump profile, the GA optimization program infers the required pump profile for the next cycle. By accessing the DMD program interface, the DMD control program converts the 2D pump profile image data into 1D data and transmits it to the DMD. Based on the image grayscale, the program autonomously controls the DMD reflectors to adjust the pumping light, thereby controlling the pump pattern. The spectral acquisition program, based on the preset parameters, acquires random lasing spectra excited by the pump pattern through spectrometer control. Through iterative optimization, the target modes are obtained, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec. The corresponding pump patterns of selected modes can then be locked and analyzed. This enables prediction of possible optical paths for the random lasing modes.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"3. Results and discussions","content":"\u003cp\u003eAs a well-known intelligent algorithm, GA includes key components: population, fitness selection, and evolutionary operators, as shown in Fig.\u0026nbsp;2a. These evolutionary operators include selection, crossover, and mutation. A population of possible solutions is employed to search the solution space. Each solution is encoded as a character string, referred to as a genome. In each iteration, the fitness values of genomes are evaluated, and those with higher fitness have an increased probability of survival. Surviving genomes combine to form \u0026ldquo;child\u0026rdquo; genomes through a process called crossover. In addition, genomes may undergo mutation. Finally, a new iteration is formed by recombining and replacing the initial population. This process repeats to yield stronger solutions over successive iterations until the target solution is found.\u003c/p\u003e\n\u003cp\u003eIn our work, the population comprises image data representing possible pump profiles. The fitness function assigns a value to each genome in the population. The fitness function is defined as follows:\u003c/p\u003e\n\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\n \u003cdiv class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e$$\\:f\\left({\\lambda\\:}_{t}\\right)=\\frac{Max\\left\\{{I}_{{\\lambda\\:}_{s1}},{I}_{{\\lambda\\:}_{s2}},\\dots\\:,{I}_{{\\lambda\\:}_{sn}}\\right\\}}{Min\\left\\{{I}_{{\\lambda\\:}_{t1}},{I}_{{\\lambda\\:}_{t2}},\\dots\\:,{I}_{{\\lambda\\:}_{tm}}\\right\\}}$$\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003eHere, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{t}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{s}\\)\u003c/span\u003e\u003c/span\u003e represent the modes to be selected and suppressed, respectively. Where m is the number of target modes (m\u0026thinsp;\u0026ge;\u0026thinsp;1), and n is the number of suppressed modes (n\u0026thinsp;\u0026ge;\u0026thinsp;0). Max denotes the maximum peak intensity among suppressed modes, while Min denotes the minimum peak intensity among target modes. Thus, a smaller \u003cem\u003ef\u003c/em\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{t}\\)\u003c/span\u003e\u003c/span\u003e) value in the optimization process indicates greater dominance of the target modes. The target modes are selected when \u003cem\u003ef\u003c/em\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{t}\\)\u003c/span\u003e\u003c/span\u003e) converges to an ideal value (\u0026le;\u0026thinsp;0.5). Figure 2b shows the evolution of \u003cem\u003ef\u003c/em\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{t}\\)\u003c/span\u003e\u003c/span\u003e) across iteration cycles for selecting a specific mode. The value converges to approximately 0.21. The inset displays the optimization process of pump profiles. When excited by a uniform pump profile, random lasing with multiple modes is observed, as shown in the upper panel of Fig. 2c. When excited by the optimized pump profile, the target mode is accurately selected, and the other modes are effectively suppressed, as shown in the lower panel of Fig. 2c.\u003c/p\u003e\n\u003cp\u003eOnce the pump profile is determined, the target mode can be accurately selected. As a result, the emission characteristics of a single mode versus multimode operation were explored in this work, as shown in Fig. 2c-d. Random lasing with multiple modes was obtained by exciting the sample with a uniform pump profile, as shown in the upper panel of Fig. 2c. As pump power density increases, the emission spectrum evolves from a typical fluorescent spectrum to a multimode random lasing spectrum. To verify the effect of mode selection, the optimized pump profile was applied to excite the sample. The lower panel of Fig. 2c shows emission spectra at different pump power densities, yielding a stable, pure single mode of random lasing. Furthermore, the threshold behavior for a specific mode, pumped by both uniform and optimized profiles, is shown in Fig. 2d. The threshold at 589 nm with optimized pumping is approximately 0.55 MW/cm\u0026sup2;, nearly half of that with uniform pumping. Overall, these results demonstrate that our proposed intelligent control system not only accurately extracts modes but also significantly enhances pumping efficiency.\u003c/p\u003e\n\u003cp\u003eTo further test the mode extraction capability of our proposed intelligent control system, we attempted to extract arbitrary single modes and multimodes, as shown in Fig. 3. When pumped by a uniform profile at higher pump power density, a random lasing spectrum with multiple modes is obtained, as shown in Fig. 3a. Keeping all other conditions unchanged, we set a single mode (590.5 nm), two modes (587.0 nm and 588.1 nm), and three modes (587.0 nm, 588.1 nm, and 590.5 nm) in the original spectrum as target modes, respectively. We adjusted the search parameters in the GA optimization program to control the pump pattern in the DMD until the target modes were obtained. The results indicate that any target mode within a complex disordered system can be accurately extracted using our proposed intelligent control system, even the adjacent modes and modes that are at a competitive disadvantage, such as those at 587.0 nm and 588.1 nm.\u003c/p\u003e\n\u003cp\u003eFor random lasers, the pump pattern determines the excited region of the sample, affecting the optical paths for lasing generation. Thus, locking the role of each pump cell in the lasing mode provides an effective method for exploring optical feedback paths. In our work, two modes (\u0026lambda;\u003csub\u003e1\u003c/sub\u003e, \u0026lambda;\u003csub\u003e2\u003c/sub\u003e) were selected individually or simultaneously for analysis, as shown in Fig. 4. A video recording of the optimization process is provided in the supporting information. The spectra and corresponding pump patterns for individual modes \u0026lambda;\u003csub\u003e1\u003c/sub\u003e and \u0026lambda;\u003csub\u003e2\u003c/sub\u003e are shown in Fig.4a. In contrast, Fig.4c shows the spectrum and corresponding pump pattern for \u0026lambda;\u003csub\u003e1\u0026nbsp;\u003c/sub\u003eand \u0026lambda;\u003csub\u003e2\u003c/sub\u003e simultaneously. For comparison, the linear superposition of pump patterns for \u0026lambda;\u003csub\u003e1\u003c/sub\u003e and \u0026lambda;\u003csub\u003e2\u003c/sub\u003e was also calculated, as shown in Fig. 4a, where the number of pump cells is significantly higher and can cover the combined pattern (\u0026lambda;\u003csub\u003e1\u003c/sub\u003e+\u0026lambda;\u003csub\u003e2\u003c/sub\u003e) in Fig.4c. Based on these results, we analyzed the effect of pump cells on \u0026lambda;\u003csub\u003e1\u003c/sub\u003e and \u0026lambda;\u003csub\u003e2\u003c/sub\u003e selection, as shown in Fig. 4b. Cells in gray (SS) simultaneously suppress \u0026lambda;\u003csub\u003e1\u003c/sub\u003e and \u0026lambda;\u003csub\u003e2\u003c/sub\u003e. Green cells (EE) simultaneously enhance \u0026lambda;\u003csub\u003e1\u003c/sub\u003e and \u0026lambda;\u003csub\u003e2\u003c/sub\u003e, while red cells (ES) enhance \u0026lambda;\u003csub\u003e1\u003c/sub\u003e but suppress \u0026lambda;\u003csub\u003e2\u003c/sub\u003e. Orange cells (E-) enhance \u0026lambda;\u003csub\u003e1\u0026nbsp;\u003c/sub\u003ewith no effect on \u0026lambda;\u003csub\u003e2\u003c/sub\u003e, while blue cells (SE) suppress \u0026lambda;\u003csub\u003e1\u003c/sub\u003e but enhance \u0026lambda;\u003csub\u003e2\u003c/sub\u003e. Cyan cells (-E) enhance \u0026lambda;\u003csub\u003e2\u003c/sub\u003e with no effect on \u0026lambda;\u003csub\u003e1\u003c/sub\u003e. That is to say, the pump-locked random laser for certain mode is achieved at the first time.\u003c/p\u003e\n\u003cp\u003eThe coupling relationship between modes can be predicted based on locked pump cells and mode intensities. When extracting single modes of \u0026lambda;\u003csub\u003e1\u003c/sub\u003e and \u0026lambda;\u003csub\u003e2\u003c/sub\u003e in turn, the illuminated cells counted up to 144 and 136, respectively, as shown in Fig. 4a. Clearly, \u0026lambda;\u003csub\u003e1\u003c/sub\u003e is easier to extract with strong intensity, indicating it have relative dominant optical feedback paths. \u0026lambda;\u003csub\u003e2\u003c/sub\u003e is more difficult to extract alone from tens of modes in the initial spectrum, suggesting weaker competitiveness and significant overlap in optical feedback paths with competing modes. When extracting the two modes simultaneously, the illuminated cells are counted up to 129, as shown in Fig. 4c. Compare with Fig. 4a, it is easier to extract the two modes than any single one with higher intensity, especially for the \u0026lambda;\u003csub\u003e2\u003c/sub\u003e. All the results indicate there is an overlap between their optical feedback paths, which has more influence on \u0026lambda;\u003csub\u003e2\u003c/sub\u003e. This work thus provides an effective method for analyzing the coupling relationships of random lasing modes.\u003c/p\u003e"},{"header":"4. Conclusions","content":"\u003cp\u003eA powerful tool for mode selection in an unpredictable random system is first demonstrated. By building intelligent pumping control system based on GA, any expected modes can be extracted accurately and efficiently, while the effect of pump cells on specific modes can be locked. Based on the locked pump cells and corresponding mode intensity, the coupling relationship between modes can be deduced. The work breaks through the barrier of quantitatively controlling modes in complex disordered systems and could guide the reverse design of photonic devices.\u003c/p\u003e"},{"header":"5. Methods","content":"\u003cp\u003e \u003cb\u003eSample preparation\u003c/b\u003e.\u003c/p\u003e \u003cp\u003eThe disordered system is a polymer film fabricated by dye-doped PMMA with scattering nanoparticles. Here, the typical laser dye RhB is used as the gain material, while Au nanorods (NRs) with diameters and lengths of approximately 25 and 50 nm, respectively, act as scatterers. The Au NRs provide coherent feedback in the polymer film and enhance RhB emission due to localized surface plasmon resonance in the local electric field. The fabrication process is illustrated in Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. First, the Au NRs (0.02 mg/mL) were spin-coated onto the flexible PET substrate at 1800 rpm for 40 seconds. Second, RhB (6 mg/mL) and PMMA (200 mg/mL) were mixed at a 1:1 volume ratio and magnetically stirred for 20 minutes. Then, the mixture was spin-coated onto the Au NR layer at 1800 rpm for 40 seconds. Third, the structure was heated at 70\u0026deg;C for 20 minutes to achieve solidification. Finally, the disordered system was obtained. Dichloromethane was used as the solvent in all steps.\u003c/p\u003e \u003cp\u003e \u003cb\u003eIntelligent control system.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe core of the intelligent control system is a self-developed software including GA optimization program, DMD control program, and spectral acquisition program. The model-view-controller (MVC) architecture is used in our self-developed software, which can separate the interface from the model. The modular design lays a good foundation for the expansion of the software functions, including debug, optimize, and replace the algorithm. The software interface mainly contains parts of spectrum display, parameter setting, regulation and control. The spectrum display area displays the current spectrum data. The parameter setting area for spectrum acquisition can set the smoothing points of the spectrometer, the spectrum acquisition time and the spectrum average number. The parameter setting area for spectrum frequency can set the frequency need to be enhanced and suppressed. The regulation and control area is used to start or stop the control function.\u003c/p\u003e \u003cp\u003e \u003cb\u003eOptical measurement.\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe diagram of test setup based on intelligent control system is illustrated in Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e. A doubled Q-switched Nd:YAG laser (532 nm, 5\u0026ndash;7 ns, 10 Hz) is employed as the pump source. The lens 1 and lens 2 are used for expanding and collimating of pump beam, respectively. The DMD is used to shape the pump profile based on the feedback of grayscale of image. The modulated pump light is reflected to lens 3. The lens 3 is used to adjust the size of pump profile. The modulated pump pattern incident on the sample and the reflected emission light was collected by an optical fiber spectrometer (Ocean Optics HR4000) with a spectral resolution of 0.02 nm. The pump power was controlled by pump voltage. The pump source, digital micromirror device (DMD) and high-resolution spectrometer are linked using the self-development software.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eCRediT authorship contribution statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJunhua Tong: Experimental investigations, Device fabrication and characterization,\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIntelligent control system establishment, Software development, Writing-original draft.\u0026nbsp;Xiaoyu Shi: Methodology \u0026amp; Writing-review.\u0026nbsp;Zhiyang Xu: Supervision \u0026amp; optimization.\u0026nbsp;Naeem Iqbal: Writing-review.\u0026nbsp;Kun Ge: Assist sample preparation.\u0026nbsp;Tianrui Zhai: Conceptualization, Investigation, Methodology, Supervision, Writing-review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of competing interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships.\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 presented in Supplementary Information. This paper or available from the corresponding author on request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe software was written in Python\u0026thinsp;3.11 using the Python package ecosystem (PySide6, numpy, scipy, scikit-opt). The computer code or available from the corresponding author on request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was financially supported by the National Natural Science Foundation of China (62375007) and the Beijing Natural Science Foundation (1232024).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eCao H, Chriki R, Bittner S, Friesem A (2019) N. Davidson. Complex lasers with controllable coherence. Nat Rev Phys 1:156\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSapienza R (2019) Determining random lasing action. Nat Rev Phys 1:690\u0026ndash;695\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSapienza R (2022) Controlling random lasing action. Nat Phys 18:976\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDu W, Hu L, Xia J, Zhang L, Li S, Kuai Y, Cao Z, Xu F, Liu Y, Zhou K, Xie K, Yu B, Raposo E, Gomes A, Hu Z (2024) Observation of the photonic Hall effect and photonic magnetoresistance in random lasers. Nat Commun 15(1):4589\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRedding B, Choma M, Cao H (2012) Speckle-free laser imaging using random laser illumination. Nat Photonics 6(6):355\u0026ndash;359\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu Z, Hong Q, Ge K, Shi X, Wang X, Deng J, Zhou Z, Zhai T (2022) Random lasing from label-free living cells for rapid cytometry of apoptosis. Nano Lett 22(1):172\u0026ndash;178\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoschetti A, Taschin A, Bartolini P, Tiwari AK, Pattelli L, Torre R (2020) Wiersma. Spectral super-resolution spectroscopy using a random laser. Nat Photonics 14:177\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGhofraniha N, Viola I, Di Maria F, Kumar Tiwari A, Pattelli L, Torre R, Diederik S (2015) Wiersma. Experimental evidence of replica symmetry breaking in random lasers. Nat Commun 6:6058\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi X, Shen K, Bian Y, Song W, Ruan J, Wang Z, Zhai T (2023) Programmable complex pumping field induced color on demand random lasing in fiber-integrated microbelts for speckle free imaging. Sci China Inf Sci 66:222401\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhai T, Zhang X, Pang Z, Su X, Liu H, Feng S, Wang L (2011) Random laser based on waveguided plasmonic gain channels. Nano Lett 11(10):4295\u0026ndash;4298\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Y, Yang W, Xiao S, Zhang N, Fan Y, Qu G, Song Q (2019) Surface-emitting perovskite random lasers for speckle-free imaging. ACS Nano 13:10653\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee K, Ma H, Rotermund F et al (2021) Non-resonant power-efficient directional Nd:YAG ceramic laser using a scattering cavity. Nat Commun 12:8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdl HP, S\u0026aacute;nchez-D\u0026iacute;az J, Vescio G, Cirera A, Garrido B, Pacheco FAV, Żuraw W, Przypis Ł, \u0026Ouml;z S, Mora-Ser\u0026oacute; I, Mart\u0026iacute;nez-Pastor JP (2024) Su\u0026aacute;rez. Tailoring single-mode random lasing of tin halide perovskites integrated in a vertical cavity. Adv Mater 36:2313252\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi X, Wang Y, Sun H, Zeng H (2017) Amino-mediated anchoring perovskite quantum dots for stable and low-threshold random lasing. Adv Mater 29:1701185\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBian Y, Xue H, Wang Z (2021) Programmable random lasing pluses based on waveguide-assisted random scattering feedback. Laser Photonics Rev 15(4):2000506\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGomes ASL, Moura AL, de Ara\u0026uacute;jo CB et al (2021) Recent advances and applications of random lasers and random fiber lasers. Prog Quantum Electron 78:100343\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWetzstein G, Ozcan A, Gigan S, Fan S, Englund D, Soljačić M, Denz C, Miller D (2020) Psaltis. Inference in artificial intelligence with deep optics and photonics. Nature 588(7836):39\u0026ndash;47\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang N, Yan W, Qu Y, Ma S, Li S (2021) Qiu. Intelligent designs in nanophotonics: from optimization towards inverse creation. PhotoniX 2:1\u0026ndash;35\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang J, Chen M, Fan J (2021) Deep neural networks for the evaluation and design of photonic devices. Nat Rev Mater 6(8):679\u0026ndash;700\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanneste C, Sebbah P (2001) Selective excitation of localized modes in active random media. Phys Rev Lett 87:183903\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCao H, Mosk A, Rotter S (2022) Shaping the propagation of light in complex media. Nat Phys 18:994\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBachelard N, Andreasen J, Gigan S, Sebbah P (2012) Taming random lasers through active spatial control of the pump. Phys Rev Lett 109:033903\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBachelard N, Gigan S, Noblin X (2014) Sebbah. Adaptive pumping for spectral control of random lasers. Nat Phys 10:426\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKumar B, Homri R, Maurya S, Lebental M (2021) Sebbah. Localized modes revealed in random lasers. Optica 8:1033\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaxena D, Arnaudon A, Cipolato O, Gaio M, Quentel A, Yaliraki S, Pisignano D, Camposeo A, Barahona M, Sapienza R (2022) Sensitivity and spectral control of network lasers. Nat Commun 13:6493\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKatoch S, Chauhan S, Kumar V (2021) A review on genetic algorithm: past, present, and future. Multimed Tools Appl 80:8091\u0026ndash;8126\u003c/span\u003e\u003c/li\u003e\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":"[email protected]","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":"Random lasers, artificial intelligence, mode selection, pump-locked","lastPublishedDoi":"10.21203/rs.3.rs-5425695/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5425695/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRandom lasing in complex disordered systems demonstrates efficient lasing across multiple localized modes. Accurate control of these modes could enable random lasers to function as fast-switching, multifunctional light sources. In this work, an unpredictable random laser was prepared and artificial intelligence (AI) technology was used for taming the lasing action. Lasing modes were precisely controlled by programming the spatial shape of the pump profile. Genetic algorithms were introduced, allowing any mode within the complex system to be extracted by setting a target value associated with that mode. Based on experimental results, the interaction model of pumping cells with lasing modes was locked. This work advances programmable random lasers, enhancing their potential for practical applications in signal processing, spectral sensing, communication, and optical computing.\u003c/p\u003e","manuscriptTitle":"Pump-locked random lasers using artificial intelligence","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-12-05 10:32:22","doi":"10.21203/rs.3.rs-5425695/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","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":"75d789bb-ac22-47a9-a223-bdbf0e5386db","owner":[],"postedDate":"December 5th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":40077604,"name":"Physical sciences/Optics and photonics/Lasers, LEDs and light sources"},{"id":40077605,"name":"Physical sciences/Optics and photonics/Optical physics/Micro-optics"}],"tags":[],"updatedAt":"2025-03-05T17:36:14+00:00","versionOfRecord":[],"versionCreatedAt":"2024-12-05 10:32:22","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5425695","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5425695","identity":"rs-5425695","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}