back

AI in nanophotonics

Computer vision and machine learning
for nanophotonic security labels and nanostructure analysis

About

Our team applies computer vision and machine learning methods to research problems. These approaches are used in developing and recognizing physically unclonable security labels, as well as for analyzing and identifying substances.

Because every physically unclonable security label is unique, using them requires protocols that make it possible to compare one label with another. To work with security labels, we develop our own computer vision algorithms based on the OpenCV library. When a label is read with computer vision, the position of each nanostructure in the label and its color are identified. After reading, an authentication protocol compares the information the user has read from the label with the information in the database.

Analyzing substances with Raman spectroscopy requires processing large amounts of spectral data, which can be quite complex because of the variety of molecular bonds in the substances under study. To solve this problem we use machine learning. We have applied various methods for classification and regression tasks, such as Decision Tree, Random Forest, Support Vectors and Logistic Regression. These methods made it possible to carry out quantitative and qualitative analysis with an accuracy above 95%. This approach solved the problems of processing and interpreting complex spectral data and automated the analysis of large amounts of data.

Interested in working together?

Tell us about your research task or project idea, and we will discuss possible areas of collaboration and our team’s expertise

contact us

4

people on the team

5

publications

27

citations

2

h-index

100%

papers in Q1 and Q2

5

papers since 2024

computer visionmachine learningunclonable labelsRaman spectroscopyOpenCVspectral analysisdata processingverification protocolsautomationdecision treesupport vector machineslogistic regressionclassificationregressionauthenticationquantitative analysisdata interpretation

Team

Dmitry Zuev

PhD in Physics and Mathematics, leading researcher

Elena Petrova

3rd-year PhD student, junior researcher

Maria Fedorova

3rd-year bachelor’s student, laboratory assistant

Alexander Loshkarev

bachelor’s student, laboratory assistant

Publications

all publications
Q1 Langmuir

Laser-Induced Stochastic Copper Microstructures for Flexible Physically Unclonable Labels

Elena Petrova, Viktoria Orekhova, Maria Fedorova, Martin Sandomirskii, Pavel Kustov, Kirill Bogdanov, Lev Logunov, Dmitry Zuev

Physically unclonable functions (PUFs) are a promising solution for anticounterfeiting technologies. However, their mass implementation is limited by fabrication complexity and/or special materials demands, as well as specialized authentication. To overcome these issues, a scalable and sustainable approach for fabricating optical PUF labels is proposed in this work. The approach is based on direct laser metallization (DLM) of deep eutectic solvents on flexible polyimide substrates. The laser-induced reduction of copper precursors produces microstructures made by specially designed templates with stochastic edge morphologies serving as unique unclonable identifiers. A computer vision–based authentication algorithm enables robust recognition, achieving an encoding capacity exceeding 10^465. The created labels can be authenticated using standard microscopes or consumer smartphones with macro lenses. Additionally, created labels were tested for stability and unauthorized copying. The replication attempts using 1200 dpi laser printing reproduced macroscopic outlines but failed to mimic microscopic features, confirming their unclonability. The fabricated labels remain functional after exposure to running water and heating up to 200°C, demonstrating excellent environmental durability. Due to its flexibility, scalability, and low-cost eco-friendly processing, the proposed label design is well suited for integration into flexible and wearable microelectronic systems, providing a practical solution to next-generation anticounterfeit technologies.

Keywords

CopperMicrostructurePhysical unclonable functionNoise
Q2 Photonics and Nanostructures - Fundamentals and Applications

Modification of wettability on fused silica surface via laser-induced nanostructures for enhanced capture of polystyrene particles

Chunyu Li, Elena Petrova, Sabina Bikmetova, Denis V. Danilov, Dmitry Zuev, Eduard Ageev

Self-assembly is an effective instrument for organization of nano- and microparticles to specially designed surface patterns. To obtain the uniform distribution of nano-and microparticles in mask-free self-assembly, the “coffee ring” effect should be carefully avoided. In this study, we propose an approach for effective limitation of such an effect on fused silica surfaces through femtosecond generation of laser-induced periodical surface structures (LIPSS). First, we determine optimal irradiation parameters to provide a higher ratio between LIPSS covered (treated) and untreated surface. Then we study a LIPPS formation in a scanning mode to cover the substrate by LIPSS reducing a contact angle. We demonstrate that scanning laser parameters with frequency of 1 kHz, pulse duration of 270 fs, laser peak fluence of 6.5 J/cm2 and scanning period of 9 µm ensure a high rate of surface wettability adjustment (0.23 mm2/min) with a contact angle of ∼ 9 ± 0.5°. Droplet tests with colloidal solution of polystyrene sub-micron (~900 nm) particles demonstrate that “coffee ring” effect is considerably reduced with more than 50 % particles captured inside laser-scanned lines. The demonstrated results are prospective for creation of various photonic structures and systems through mask-free self-assembly.

Keywords

PolystyreneSuperhydrophilicityWettingFemtosecondLaserFluenceContact angleParticle
Photonics and Nanostructures - Fundamentals and Applications more
Q2 ACS Applied Optical Materials

Polarization-Sensitive Scattering from Tunable Hybrid Au–Si Nanoantennas for Anticounterfeiting Applications

Kustov P., Yaroshenko V., Sandomirskii M., Petrova E., Fedorova M., Ageev E., Mukhin I., Zuev D.A.

Polarization of light provides an additional degree of freedom in nanostructure-light interactions and delivers a wide variety of physical phenomena and applications. While relatively well-described for plasmonics and high-refractive-index nanosystems, the potential for polarization-dependent control of optical properties in hybrid nanostructures composed of two or more materials remains not fully unlocked. Here, we study the polarization-dependent scattering behavior of resonant hybrid gold–silicon nanostructures lacking inversion symmetry. We adjust the geometrical configuration and material phase of the nanostructure components through precise femtosecond laser irradiation to induce polarization-dependent changes of scattering spectra. By evaluating the color change using the color-difference (Δ E *) formula, we demonstrate Δ E * values of approximately 9.7 and 3.9 for TE- and TM-polarized oblique illumination, respectively. Notably, the latter value is close to the just noticeable difference threshold of 2.3 calculated for the studied nanosystems. Numerical simulations and semianalytical multipole decomposition techniques are applied to investigate this polarization-dependent scattering behavior, revealing that for the studied hybrid nanosystem, this phenomenon occurs at illumination angles ranging from 57° to 75°. Finally, we developed an effective defense mechanism against counterfeiting. The achieved results hold significant potential for further research into hybrid nanophotonic systems and for the development of optical chips, sensors, and anticounterfeit systems.

Keywords

PolarizationScatteringPhysical chemistryMaterials scienceOptoelectronicsOpticsPhysicsChemistry
Q1 Nature Communications

Spectral physical unclonable functions: downscaling randomness with multi-resonant hybrid particles

Martin Sandomirskii, Elena Petrova, Pavel Kustov, Lev Chizhov, Artem Larin, Stéphanie Bruyère, Vitaly Yaroshenko, Eduard Ageev, Pavel Belov, Dmitry Zuev

Optical physical unclonable functions (PUFs) are state-of-the-art in advanced security applications. Fabricated with inherent randomness, they generate fingerprint-like responses, serving as trust anchors for material assets. However, the existing PUFs, typically reliant on microscopic spatial features, face increasing threats from rapidly advancing microscale manipulation techniques. Here, we present novel PUFs based on random nanoscale variations within multi-resonant gold-silicon particles. These inevitable structural differences, coupled with strong optical resonances, provide unique spectral features in particles' photoluminescence (PL), which we encode as unclonable keys. Our approach surpasses the shortcomings of diffraction-limited designs, additionally offering a multi-functional platform for robust authentication of goods and verification of individuals. We demonstrate two security label models based on PL mapping and direct PL imaging, as well as a concept for the first all-optical one-time password verification token with an exceptionally high storage density of unique information. This work paves the way toward nanoscale-enabled unclonability, bringing enhanced security for hardware-based cryptography, personalized access control, and cutting-edge anti-counterfeiting.

Keywords

RandomnessDownscalingStatistical physicsMeteorologyStatisticsPhysicsComputer scienceMathematics
Q1 The Journal of Physical Chemistry Letters

IR Hidden Patterns for Security Labels

Vitaly Yaroshenko, Artem Larin, Sergey Syubaev, Ivan Vazhenin, Pavel Kustov, Dmitry Dolgintsev, Eduard Ageev, Stanislav Gurbatov, Alina Maksimova, Kristina Novikova, Sergey Babin, Aleksey Kozlov, Alexandr Dostovalov, Aleksandr Kuchmizhak, Dmitry Zuev

Authentication of a product's originality by anticounterfeiting labels represents a crucial point toward protection against forgery. Fast and scalable fabrication methods of original labels with a high degree of protection are in high demand for the protection of valuable goods. Here, we propose a simple strategy for fabrication of hidden security tags with IR luminescent readout by the direct femtosecond laser patterning of silicon-erbium-silicon sandwiched thin films. The choice of laser processing parameters makes possible the creation of random or quasi-regular self-organized surface nanotextures. The controlled laser-driven oxidation accompanying this process provides simultaneous regulation of the film's optical properties and spontaneous emission yield of the embedded Er atoms. The regimes are detected when optically similar patterned areas demonstrate different Er emission intensities, allowing us to create hidden security tags with facile readout at the C-band telecommunication wavelengths. The obtained results take another step toward the application of IR-luminescent erbium-based anticounterfeiting labels for covert and/or forensic security levels.

Keywords

Computer securityBusinessInternet privacyComputer science

Showing 5 of 5