One-click information capture, is the high-dimensional light field detector so powerful?

Produced by: Science Popularization China

Author: Cai Mingxuan (Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences)

Producer: China Science Expo

Editor's note: In order to unveil the mystery of scientific work, the China Science Popularization Frontier Science Project launched a series of articles called "Me and My Research", inviting scientists to write articles themselves, share their scientific research experiences, and create a scientific world. Let us follow the explorers at the forefront of science and technology and embark on a journey full of passion, challenges, and surprises.

With the rapid development of information technology, the importance of light field sensing technology in various fields has become increasingly prominent. The light field contains not only light intensity, but also multi-dimensional information such as polarization, frequency and phase. This information is crucial for revealing the composition and surface morphology of matter, and has a wide range of application value in optical communications, remote sensing, industrial detection, medical diagnosis, chemical analysis and environmental protection.

Spectral and polarization detection

Spectral detection and polarization detection are the main means of obtaining key information. Spectral detection obtains information by analyzing the different wavelength components of light. It is a distribution diagram of light after it is decomposed according to wavelength or frequency. Since each substance has different absorption and reflection characteristics of light at different wavelengths, the spectrum is like an "identity card" containing important information. Analyzing the spectral characteristics can allow us to obtain detailed information about the substance.

For example, we can determine the health of plants by analyzing the spectrum reflected by the plants; analyze the spectrum of starlight to understand the chemical composition and motion of stars; and in chemistry, analyzing the spectrum can determine the concentration of various chemical components in the solution.

The polarization state of light refers to the direction of light wave vibration. By analyzing the direction of light wave vibration, we can reveal information such as the surface morphology, molecular structure, and stress distribution of an object. The polarization of light is a vector property. When light passes through different media, its polarization state will change.

The human eye can directly or indirectly observe the intensity, color (wavelength), and phase information of light, but cannot perceive the polarization state of light. Therefore, the research and application of polarization optics started later and is more complicated.

Polarization detection can be applied in several areas: In the field of industrial inspection, polarization detection can detect tiny cracks or other defects on the metal surface, thereby ensuring the quality and reliability of the product; in the field of remote sensing, by analyzing the polarization state of light reflected from the surface, different types of vegetation and soil can be distinguished, and even environmental problems such as oil pollution can be detected, providing richer information than traditional remote sensing, which is helpful for environmental monitoring and natural resource management; in the field of biomedicine, polarization microscopy can be used to observe structural changes in cells and tissues, helping doctors to make more accurate diagnoses.

Problems and limitations of traditional detectors

Existing polarization and spectral detectors can usually only obtain polarization information at a fixed wavelength, or obtain spectral information at a fixed polarization state. However, in many scenes in nature, light fields may carry arbitrary polarization and intensity variations over a wide spectral range.

There are two main problems with existing polarization and spectral detectors: first, their detection capabilities are proportional to the time or space required; second, they cannot accurately detect high-dimensional light field information with arbitrarily varying polarization and intensity over a wide spectral range.

In order to solve the above problems, the team of Researcher Li Wei from the Changchun Institute of Optics, Fine Mechanics and Physics of the Chinese Academy of Sciences and the team of Professor Qiu Chengwei from the National University of Singapore cooperated to develop a high-dimensional light field detector based on dispersive surfaces.

This new detector can fully characterize high-dimensional light fields with arbitrarily varying polarization and intensity over a broadband spectral range using a single device through a single measurement. This groundbreaking research significantly improves the detection capability of light field information, greatly improves the flexibility and efficiency of detection, and opens up a new perspective for high-dimensional light field detection.

Schematic diagram of comparison with traditional light detection methods

(Image source: Dispersion-assisted high-dimensional photodetector)

Development of high-dimensional light field detectors

The core innovation of this research is to use simple thin film interfaces and spatial dispersion and frequency dispersion characteristics to project the high-dimensional information of light into the wave vector space. This design not only simplifies the complex structure of traditional spectral and polarization detectors, but also greatly improves the detection efficiency and accuracy.

First, we need to understand that the thin film interface is the transition between two materials with different refractive indices. It is like a "sorting" station for light . When light passes through these interfaces, the propagation direction and speed of light will change due to the difference in refractive index, resulting in refraction and reflection. Light of different wavelengths and polarization states is affected in different ways.

For example, some wavelengths of light may be strongly reflected at the interface, while other wavelengths may continue to propagate through the interface. By designing the thickness of the film and the material combination, the research team was able to achieve sensitivity to both the frequency and polarization of light in wavevector space.

When choosing thin film materials and structures, the magic of materials must be performed. The research team selected titanium dioxide (TiO₂) and silicon dioxide (SiO₂), two materials with "high transmittance" in the visible light band, as the main materials. High transmittance means that light can maintain a high energy transmission efficiency when passing through these materials, thereby improving the sensitivity and accuracy of the detector. In addition, the optical properties of these materials enable them to produce the desired scattering and refraction effects at different wavelengths and polarization states, thereby achieving efficient encoding of light field information.

This sensitivity to the frequency and polarization of light can be easily resonantly designed and amplified. To demonstrate this, the research team constructed a Fabry–Pérot cavity. The Fabry–Pérot cavity consists of two highly reflective mirrors that are parallel to each other. The light is reflected multiple times in the cavity, so that light of different wavelengths and polarization states will experience different interference effects, thereby producing specific transmission characteristics in the cavity. In this way, the sensitivity to the frequency and polarization of light is further enhanced.

Schematic diagram of a single high-dimensional detection method

(Image source: Dispersion-assisted high-dimensional photodetector)

In addition, by stacking multilayer thin film structures on both sides, the research team not only achieved broadband spectral detection, but also full Stokes polarization detection. Full Stokes polarization detection is an advanced optical characterization technology that can simultaneously measure all four Stokes parameters (S₀, S₁, S₂, S₃) that describe the polarization state of light , providing complete information on the polarization properties of light.

These four parameters represent the total intensity of light, the difference between the linear polarization intensity in the horizontal/vertical direction and the +45°/-45° direction, and the difference between the left-handed/right-handed circular polarization intensity. Traditional polarization detection methods can usually only measure one or two of these parameters, making it difficult to fully characterize the polarization state of light.

In order to extract useful information from complex light field data, the research team used a deep residual network (ResNet) for decoding. ResNet is a common neural network architecture, which can be imagined as a shortcut in a complex maze. These shortcuts help the model find the exit more quickly, that is, the useful information in the data, so that the network can be trained more deeply and efficiently.

Compared with other neural network structures, ResNet performs better in learning effect and training time, and can efficiently process high-dimensional matrix data.

When light passes through the designed film interface and cavity structure, the polarization and spectral information of the light is encoded into complex two-dimensional transmission distribution information. This information is recorded by a high-resolution camera and input into a pre-trained ResNet model.

Since complex polarization and spectral information are encoded at the same time, and additional system noise is introduced in the experiment, other information extraction methods used in the early stage can only extract and analyze the linear polarization state. Through testing, the research team finally determined that the ResNet model can successfully extract high-dimensional information.

Achieve more efficient and accurate light field detection

In order to verify the actual detection effect of the high-dimensional light field detector, the research team selected two typical high-dimensional light fields for experimental testing: one is a two-color dual-polarization laser field, and the other is a reflected light field generated by broadband light irradiating a gold surface. These two light fields have complex polarization and spectral information, which are ideal for testing the performance of the detector.

In these experiments, the research team demonstrated that the designed high-dimensional light field detector can accurately detect the information of these complex light fields using a single measurement.

For example, in the test of dual-color dual-polarization laser field, the detector can simultaneously obtain light field information of two different wavelengths and polarization states, and accurately restore its spectrum and polarization characteristics. In the broadband light reflection experiment, the detector also successfully captured the intensity of each wavelength in the reflected light field and the changes in polarization state, thereby more comprehensively describing the light field information.

To further demonstrate the integration and convenience of the detector, the research team combined the film with a microlens array and a large-area imaging sensor array in a "sandwich" manner to develop an ultra-integrated high-dimensional light field imager that does not require alignment and can perform single-shot measurements.

The imager can simultaneously obtain the target's polarization and spectral information in one imaging, achieving high-precision and high-dimensional imaging. This design not only reduces the complexity of the experimental device, but also significantly improves the detection efficiency and accuracy.

High-dimensional photodetector and imager experiments

(Image source: Dispersion-assisted high-dimensional photodetector)

In the future, the research team plans to achieve similar high-dimensional detection effects in other wide-band ranges and explore high-dimensional detection systems in practical applications, such as high-dimensional detection in outdoor scenes. At the same time, the team will use material systems such as metasurfaces and two-dimensional materials to further miniaturize detectors and reduce the amount of prior data required to achieve more efficient and accurate light field detection.

References:

1.Fan, Y., Huang, W., Zhu, F. et al. Dispersion-assisted high-dimensional photodetector. Nature 630, 77–83 (2024).