Showing papers in "Optics Express in 2019"

Deep learning for accelerated all-dielectric metasurface design

Abstract: Deep learning has risen to the forefront of many fields in recent years, overcoming challenges previously considered intractable with conventional means. Materials discovery and optimization is one such field, but significant challenges remain, including the requirement of large labeled datasets and one-to-many mapping that arises in solving the inverse problem. Here we demonstrate modeling of complex all-dielectric metasurface systems with deep neural networks, using both the metasurface geometry and knowledge of the underlying physics as inputs. Our deep learning network is highly accurate, achieving an average mean square error of only 1.16 × 10−3 and is over five orders of magnitude faster than conventional electromagnetic simulation software. We further develop a novel method to solve the inverse modeling problem, termed fast forward dictionary search (FFDS), which offers tremendous controls to the designer and only requires an accurate forward neural network model. These techniques significantly increase the viability of more complex all-dielectric metasurface designs and provide opportunities for the future of tailored light matter interactions.

One-step robust deep learning phase unwrapping.

Abstract: Phase unwrapping is an important but challenging issue in phase measurement. Even with the research efforts of a few decades, unfortunately, the problem remains not well solved, especially when heavy noise and aliasing (undersampling) are present. We propose a database generation method for phase-type objects and a one-step deep learning phase unwrapping method. With a trained deep neural network, the unseen phase fields of living mouse osteoblasts and dynamic candle flame are successfully unwrapped, demonstrating that the complicated nonlinear phase unwrapping task can be directly fulfilled in one step by a single deep neural network. Excellent anti-noise and anti-aliasing performances outperforming classical methods are highlighted in this paper.

Learning from simulation: An end-to-end deep-learning approach for computational ghost imaging

Abstract: Artificial intelligence (AI) techniques such as deep learning (DL) for computational imaging usually require to experimentally collect a large set of labeled data to train a neural network. Here we demonstrate that a practically usable neural network for computational imaging can be trained by using simulation data. We take computational ghost imaging (CGI) as an example to demonstrate this method. We develop a one-step end-to-end neural network, trained with simulation data, to reconstruct two-dimensional images directly from experimentally acquired one-dimensional bucket signals, without the need of the sequence of illumination patterns. This is in particular useful for image transmission through quasi-static scattering media as little care is needed to take to simulate the scattering process when generating the training data. We believe that the concept of training using simulation data can be used in various DL-based solvers for general computational imaging.

100 m/500 Mbps underwater optical wireless communication using an NRZ-OOK modulated 520 nm laser diode

Abstract: In this paper, we proposed and experimentally demonstrated a long-distance high-speed underwater optical wireless communication (UOWC) system in a laboratory environment by using a low-cost green laser diode (LD) and power-efficient non-return-to-zero on-off keying (NRZ-OOK) modulation. The system successfully achieved a data rate of 500 Mbps through a 100 m tap-water channel by using a pigtailed single-mode fiber 520 nm green LD. The tap water was measured to have an attenuation coefficient comparable to pure seawater. The measured system bit error rate (BER) value of 2.5 × 10-3 was below the forward error correction (FEC) limit of 3.8 × 10-3 with 7% overhead. The distance can be extended if the received optical power is allowed to reduce to the minimum power to meet the data rate requirement. Based on the measured minimum required power and the power decay model in the water channel, the transmission performance was predicted to be 146 m/500 Mbps and 174 m/100 Mbps.

Highly efficient GaN-based high-power flip-chip light-emitting diodes

Abstract: High-power flip-chip light-emitting diodes (FCLEDs) suffer from low efficiencies because of poor p-type reflective ohmic contact and severe current crowding. Here, we show that it is possible to improve both the light extraction efficiency (LEE) and current spreading of an FCLED by incorporating a highly reflective metallic reflector made from silver (Ag). The reflector, which consists of an Ag film covered by three pairs of TiW/Pt multilayers, demonstrates high reflectance of 95.0% at 460 nm at arbitrary angles of incidence. Our numerical simulation and experimental results reveal that the FCLED with Ag-based reflector exhibits higher LEE and better current spreading than the FCLED with indium-tin oxide (ITO)/distributed Bragg reflector (DBR). As a result, the external quantum efficiency (EQE) of FCLED with Ag-based reflector was 6.0% higher than that of FCLED with ITO/DBR at 750 mA injection current. Our work also suggests that the EQE of FCLED with the Ag-based reflector could be further enhanced 5.2% by replacing the finger-like n-electrodes with three-dimensional (3D) vias n-electrodes, which spread the injection current uniformly over the entire light-emitting active region. This study paves the way towards higher-performance LED technology.

Deep learning wavefront sensing

Abstract: We present a new class of wavefront sensors by extending their design space based on machine learning. This approach simplifies both the optical hardware and image processing in wavefront sensing. We experimentally demonstrated a variety of image-based wavefront sensing architectures that can directly estimate Zernike coefficients of aberrated wavefronts from a single intensity image by using a convolutional neural network. We also demonstrated that the proposed deep learning wavefront sensor can be trained to estimate wavefront aberrations stimulated by a point source and even extended sources.

Optimization of thermo-optic phase-shifter design and mitigation of thermal crosstalk on the SOI platform.

Abstract: We first optimize the design and compare the performance of thermo-optic phase-shifters based on TiN metal and N++ doped silicon, in the same SOI process. The designs don’t require special material processing, show negligible loss, and have very stable power consumption. The optimum TiN design has a switching powerPπ=21.4 mW and a time constantτ=5.6 µs, whereasPπ=22.8 mW andτ=2.2 µs for the best N++ Si design, enabling 2.5x faster switching compared to the metal heater. Doped-Si-based heaters are therefore the most practical and efficient on standard SOI. In addition, to optimize the layout density of highly integrated dies, we experimentally characterize internal and external thermal crosstalk for tunable Mach-Zehnder interferometers (MZIs) based on both heater designs for various power, distances, and etching patterns. Deep trenches are the best structures not involving special fabrication techniques to mitigate heat leakage affecting phase-sensitive devices close to heaters. Given the numerous applications of thermal tuners, this work is relevant to almost all silicon photonics designers.

Topology optimization of freeform large-area metasurfaces

Abstract: We demonstrate optimization of optical metasurfaces over 105–106 degrees of freedom in two and three dimensions, 100–1000+ wavelengths (λ) in diameter, with 100+ parameters per λ2. In particular, we show how topology optimization, with one degree of freedom per high-resolution “pixel,” can be extended to large areas with the help of a locally periodic approximation that was previously only used for a few parameters per λ2. In this way, we can computationally discover completely unexpected metasurface designs for challenging multi-frequency, multi-angle problems, including designs for fully coupled multi-layer structures with arbitrary per-layer patterns. Unlike typical metasurface designs based on subwavelength unit cells, our approach can discover both sub- and supra-wavelength patterns and can obtain both the near and far fields.

An all-optical neuron with sigmoid activation function.

Abstract: We present an all-optical neuron that utilizes a logistic sigmoid activation function, using a Wavelength-Division Multiplexing (WDM) input & weighting scheme. The activation function is realized by means of a deeply-saturated differentially-biased Semiconductor Optical Amplifier-Mach-Zehnder Interferometer (SOA-MZI) followed by a SOA-Cross-Gain-Modulation (XGM) gate. Its transfer function is both experimentally and theoretically analyzed, showing excellent agreement between theory and experiment and an almost perfect fitting with a logistic sigmoid function. The optical sigmoid transfer function is then exploited in the experimental demonstration of a photonic neuron, demonstrating successful thresholding over a 100psec-long pulse sequence with 4 different weighted-and-summed power levels.

Kilopixel array of superconducting nanowire single-photon detectors.

Abstract: We present a 1024-element near-infrared imaging array of superconducting nanowire single photon detectors (SNSPDs) using a 32×32 row-column multiplexing architecture. The array has an active area of 0.96 × 0.96 mm, making it the largest SNSPD array reported to date in terms of both active area and pixel count. Using a 64-channel time-tagging readout, we have characterized the array’s yield, efficiency, and timing resolution. Large arrays of SNSPDs are desirable for applications such as imaging, spectroscopy, or particle detection.

Heterogeneous silicon photonics sensing for autonomous cars.

Abstract: Heterogeneous silicon photonics is uniquely positioned to address the photonic sensing needs of upcoming autonomous cars and provide the necessary cost reduction for widespread deployment. This is because it allows for wafer-scale active/passive integration, including optical sources. We present our recent research and the development of interferometric optical gyroscopes and LiDAR sensors. More specifically, we show a fully integrated gyroscope front-end occupying an area of only 4.5 mm2. We also show the first dense pitch optical phased array using heterogeneous phase shifters. The 4 µm pitch heterogeneous phase shifters provide very low V2π of only 0.35-1.4 V across 200 nm, low residual amplitude modulation of only 0.1-0.15 dB for 2π phase shift, extremely low static power consumption ( 1 GHz). All of these factors make them ideal for next-generation LiDAR systems that employ optical phased arrays.

Absorption and slow-light analysis based on tunable plasmon-induced transparency in patterned graphene metamaterial.

Abstract: We propose a novel simple patterned monolayer graphene metamaterial structure based on tunable terahertz plasmon-induced transparency (PIT). Destructive interference in this structure causes pronounced PIT phenomenon, and the PIT response can be dynamically controlled by voltage since the existence of continuous graphene bands in the structural design. The theoretical transmission of this structure is calculated by coupled mode theory (CMT), and the results are highly consistent with the simulation curve. After that, the influence of the graphene mobility on the PIT response and absorption characteristics is researched. It is found that the absorption efficiency of our designed structure can reach up to 50%, meaning the structure is competent to prominent terahertz absorber. Moreover, the slow-light performance of this structure is discussed via analyzing the group refractive index and phase shift. It shows that the structure possesses a broad group refractive index band with ultra-high value, and the value is up to 382. This work will diversify the designs for versatile tunable terahertz devices and micro-nano slow-light devices.

Dynamically tunable dual plasmon-induced transparency and absorption based on a single-layer patterned graphene metamaterial

Abstract: Dual plasmon-induced transparency (PIT) and plasmon-induced absorption (PIA) are simultaneously achieved in an integrated metamaterial composed of single layer of graphene. Electric field distribution and coupled mode theory (CMT) are used to demonstrate the physical mechanism of dual PIT and PIA, and the theoretical result of CMT is highly consistent with the finite-difference time-domain (FDTD) method simulation result. Further research shows that both the dual PIT and PIA phenomenon can be effectively modulated by the Fermi level, the carrier mobility of the graphene and the refractive index of the surrounding environment. It is meaningful that the absorption of the dual PIA spectrum can be abruptly increased to 93.5% when the carrier mobility of graphene is 0.8m2/Vs. In addition, the group index can be as high as 328. Thus, our work can pave new way for developing excellent slow-light and light absorption functional devices.

Optimisation of colour generation from dielectric nanostructures using reinforcement learning.

Abstract: Recently, a novel machine learning model has emerged in the field of reinforcement learning known as deep Q-learning. This model is capable of finding the best possible solution in systems consisting of millions of choices, without ever experiencing it before, and has been used to beat the best human minds at complex games such as, Go and chess, which both have a huge number of possible decisions and outcomes for each move. With a human-level intelligence, it has solved the problems that no other machine learning model has done before. Here, we show the steps needed for implementing this model to an optical problem. We investigate the colour generation by dielectric nanostructures and show that this model can find geometrical properties that can generate much purer red, green and blue colours compared to previously reported results. The model found these results in 9000 steps from a possible 34.5 million solutions. This technique can easily be extended to predict and optimise the design parameters for other optical structures.

Achieving the laser intensity of 5.5×10 22 W/cm 2 with a wavefront-corrected multi-PW laser

Abstract: The generation of ultrahigh intensity laser pulses was investigated by tightly focusing a wavefront-corrected multi-petawatt Ti:sapphire laser. For the wavefront correction of the PW laser, two stages of deformable mirrors were employed. The multi-PW laser beam was tightly focused by an f/1.6 off-axis parabolic mirror and the focal spot profile was measured. After the wavefront correction, the Strehl ratio was about 0.4, and the spot size in full width at half maximum was 1.5×1.8 μm2, close to the diffraction-limited value. The measured peak intensity was 5.5×1022 W/cm2, achieving the highest laser intensity ever reached.

Machine learning approach for computing optical properties of a photonic crystal fiber

Abstract: Photonic crystal fibers (PCFs) are the specialized optical waveguides that led to many interesting applications ranging from nonlinear optical signal processing to high-power fiber amplifiers. In this paper, machine learning techniques are used to compute various optical properties including effective index, effective mode area, dispersion and confinement loss for a solid-core PCF. These machine learning algorithms based on artificial neural networks are able to make accurate predictions of above-mentioned optical properties for usual parameter space of wavelength ranging from 0.5-1.8 µm, pitch from 0.8-2.0 µm, diameter by pitch from 0.6-0.9 and number of rings as 4 or 5 in a silica solid-core PCF. We demonstrate the use of simple and fast-training feed-forward artificial neural networks that predicts the output for unknown device parameters faster than conventional numerical simulation techniques. Computation runtimes required with neural networks (for training and testing) and Lumerical MODE solutions are also compared.

Ultra-broadband perfect absorber utilizing refractory materials in metal-insulator composite multilayer stacks

Abstract: We present an ultra-broadband perfect absorber composed of metal-insulator composite multilayer (MICM) stacks by placing the insulator-metal-insulator (IMI) grating on the metal-insulator-metal (MIM) film stacks. The absorber shows over 90% absorption spanning between 570 nm and 3539 nm, with an average absorption of 97% under normal incidence. The ultra-broadband perfect absorption characteristics are achieved by the synergy of guided mode resonances (GMRs), localized surface plasmons (LSPs), propagating surface plasmons (PSPs), and cavity modes. The polarization insensitivity is demonstrated by analyzing the absorption performance over arbitrary polarization angles. The ultra-broadband absorption remains more than 80% over a wide incident angle up to 50°, for both transverse electric (TE) and transverse magnetic (TM) modes. The ultra-broadband perfect absorber has tremendous potential for various applications, such as solar thermal energy harvesting, thermoelectrics, and imaging.

Single-mode, low loss hollow-core anti-resonant fiber designs.

Abstract: In this paper, we numerically investigate various hollow-core anti-resonant (HC-AR) fibers towards low propagation and bend loss with effectively single-mode operation in the telecommunications window. We demonstrate how the propagation loss and higher-order mode modal contents are strongly influenced by the geometrical structure and the number of the anti-resonant cladding tubes. We found that 5-tube nested HC-AR fiber has a wider anti-resonant band, lower loss, and larger higher-order mode extinction ratio than designs with 6 or more anti-resonant tubes. A loss ratio between the higher-order modes and fundamental mode, as high as 12,000, is obtained in a 5-tube nested HC-AR fiber. To the best of our knowledge, this is the largest higher-order mode extinction ratio demonstrated in a hollow-core fiber at 1.55 μm. In addition, we propose a modified 5-tube nested HC-AR fiber, with propagation loss below 1 dB/km from 1330 to 1660 nm. This fiber also has a small bend loss of ~15 dB/km for a bend radius of 1 cm.

Rydberg-atom-based digital communication using a continuously tunable radio-frequency carrier.

Abstract: Up to now, the measurement of radio-frequency (RF) electric field achieved using the electromagnetically-induced transparency (EIT) of Rydberg atoms has proved to be of high-sensitivity and shows a potential to produce a promising atomic RF receiver at resonance between two chosen Rydberg states. In this paper, we study the extension of the feasibility of digital communication via this quantum-based antenna over a continuously tunable RF-carrier at off-resonance. Our experiment shows that the digital communication at a rate of 500 kbps can be performed reliably within a tunable bandwidth of 200 MHz near a 10.22 GHz carrier. Outside of this range, the bit error rate (BER) increases, rising to, for example, 15% at an RF-detuning of ±150 MHz. In the measurement, the time-varying RF field is retrieved by detecting the optical power of the probe laser at the center frequency of RF-induced symmetric or asymmetric Autler-Townes splitting in EIT. Prior to the digital test, we studied the RF-reception quality as a function of various parameters including the RF detuning and found that a choice of linear gain response to the RF-amplitude can suppress the signal distortion. The modulating signal can be decoded at speeds up to 500 kHz in the tunable bandwidth. Our test consolidates the physical basis for reliable communication and spectral sensing over a wider broadband RF-carrier, which paves a way for the concurrent multi-channel communications founded on the same pair of Rydberg states.

Three-dimensional single-photon imaging through obscurants

Abstract: We investigate the depth imaging of objects through various densities of different obscurants (water fog, glycol-based vapor, and incendiary smoke) using a time-correlated single-photon detection system which had an operating wavelength of 1550 nm and an average optical output power of approximately 1.5 mW. It consisted of a monostatic scanning transceiver unit used in conjunction with a picosecond laser source and an individual Peltier-cooled InGaAs/InP single-photon avalanche diode (SPAD) detector. We acquired depth and intensity data of targets imaged through distances of up to 24 meters for the different obscurants. We compare several statistical algorithms which reconstruct both the depth and intensity images for short data acquisition times, including very low signal returns in the photon-starved regime.

MXene-based saturable absorber for femtosecond mode-locked fiber lasers.

Abstract: We report simple and compact all-fiber erbium-doped soliton and dispersion-managed soliton femtosecond lasers mode-locked by the MXene Ti3C2Tx. A saturable absorber device fabricated by optical deposition of Ti3C2Tx onto a microfiber exhibits strong saturable absorption properties, with a modulation depth of 11.3%. The oscillator operating in the soliton regime produces 597.8 fs-pulses with 5.21 nm of bandwidth, while the cavity with weak normal dispersion (~0.008 ps2) delivers 104 fs pulses with 42.5 nm of bandwidth. Our results contribute to the growing body of work studying the nonlinear optical properties of MXene that underpin new opportunities for ultrafast photonic technology.

Mesoscale laser 3D printing

Abstract: 3D meso scale structures that can reach up to centimeters in overall size but retain micro- or nano-features, proved to be promising in various science fields ranging from micro-mechanical metamaterials to photonics and bio-medical scaffolds. In this work, we present synchronization of the linear and galvanometric scanners for efficient femtosecond 3D optical printing of objects at the meso-scale (from sub-μm to sub-cm spanning five orders of magnitude). In such configuration, the linear stages provide stitch-free structuring at nearly limitless (up to tens-of-cm) working area, while galvo-scanners allow to achieve translation velocities in the range of mm/s-cm/s without sacrificing nano-scale positioning accuracy and preserving the undistorted shape of the final print. The principle behind this approach is demonstrated, proving its inherent advantages in comparison to separate use of only linear stages or scanners. The printing rate is calculated in terms of voxels/s, showcasing the capability to maintain an optimal feature size while increasing throughput. Full capabilities of this approach are demonstrated by fabricating structures that reach millimeters in size but still retain sub-μm features: scaffolds for cell growth, microlenses, and photonic crystals. All this is combined into a benchmark structure: a meso-butterfly. Provided results show that synchronization of two scan modes is crucial for the end goal of industrial-scale implementation of this technology and makes the laser printing well aligned with similar approaches in nanofabrication by electron and ion beams.

Fourier light-field microscopy

Abstract: Observing the various anatomical and functional information that spans many spatiotemporal scales with high resolution provides deep understandings of the fundamentals of biological systems. Light-field microscopy (LFM) has recently emerged as a scanning-free, scalable method that allows for high-speed, volumetric imaging ranging from single-cell specimens to the mammalian brain. However, the prohibitive reconstruction artifacts and severe computational cost have thus far limited broader applications of LFM. To address the challenge, in this work, we report Fourier LFM (FLFM), a system that processes the light-field information through the Fourier domain. We established a complete theoretical and algorithmic framework that describes light propagation, image formation and system characterization of FLFM. Compared with conventional LFM, FLFM fundamentally mitigates the artifacts, allowing high-resolution imaging across a two- to three-fold extended depth. In addition, the system substantially reduces the reconstruction time by roughly two orders of magnitude. FLFM was validated by high-resolution, artifact-free imaging of various caliber and biological samples. Furthermore, we proposed a generic design principle for FLFM, as a highly scalable method to meet broader imaging needs across various spatial levels. We anticipate FLFM to be a particularly powerful tool for imaging diverse phenotypic and functional information, spanning broad molecular, cellular and tissue systems.

Learning to decompose the modes in few-mode fibers with deep convolutional neural network

Abstract: We introduce a deep-learning technique to perform complete mode decomposition for few-mode optical fibers for the first time. Our goal is to learn a fast and accurate mapping from near-field beam patterns to the complete mode coefficients, including both modal amplitudes and phases. We train the convolutional neural network with simulated beam patterns and evaluate the network on both the simulated beam data and the real beam data. In simulated beam data testing, the correlation between the reconstructed and the ideal beam patterns can achieve 0.9993 and 0.995 for 3-mode case and 5-mode case, respectively. While in the real 3-mode beam data testing, the average correlation is 0.9912 and the mode decomposition can be potentially performed at 33 Hz frequency on a graphic processing unit, indicating real-time processing ability. The quantitative evaluations demonstrate the superiority of our deep learning–based approach.

Laser frequency sweep linearization by iterative learning pre-distortion for FMCW LiDAR.

Abstract: We report on a laser frequency sweep linearization method by iterative learning pre-distortion for frequency-modulated continuous-wave (FMCW) light detection and ranging (LiDAR) systems. A pre-distorted laser drive voltage waveform that results in a linear frequency sweep is obtained by an iterative learning controller, and then applied to the FMCW LiDAR system. We have also derived a fundamental figure of merit for the maximum residual nonlinearity needed to achieve the transform-limited range resolution. This method is experimentally tested using a commercial vertical cavity surface-emitting laser (VCSEL) and a distributed feedback (DFB) laser, achieving less than 0.005% relative residual nonlinearity of frequency sweep. With the proposed method, high-performance FMCW LiDAR systems can be realized without expensive linear lasers, complex linearization setups, or heavy post-processing.

Ultrasensitive strain sensor based on Vernier- effect improved parallel structured fiber-optic Fabry-Perot interferometer

Abstract: A novel parallel structured fiber-optic Fabry-Perot interferometer (FPI) based on Vernier-effect is theoretically proposed and experimentally demonstrated for ultrasensitive strain measurement. This proposed sensor consists of open-cavity and closed-cavity fiber-optic FPI, both of which are connected in parallel via a 3 dB coupler. The open-cavity is implemented for sensing, while the closed-cavity for reference. Experimental results show that the proposed parallel structured fiber-optic FPI can provide an ultra-high strain sensitivity of −43.2 pm/μe, which is 4.6 times higher than that of a single open-cavity FPI. Furthermore, the sensor is simple in fabrication, robust in structure, and stable in measurement. Finally, the parallel structured fiber-optic FPI scheme proposed in this paper can also be applied to other sensing field, and provide a new perspective idea for high sensitivity sensing.

Sub-wavelength-pitch silicon-photonic optical phased array for large field-of-regard coherent optical beam steering

Abstract: This paper reports on large field-of-regard, high-efficiency, and large aperture active optical phased arrays (OPAs) for optical beam steering in LIDAR systems. The fabricated 5 mm-long silicon photonic OPA with a 1.3 μm waveguide pitch achieved adjacent waveguide crosstalk below −12dB. A relatively large and uniform emission aperture has been achieved with a low-contrast silicon nitride assisted grating (~20 dB/cm) whose emission profile can be further optimized using an apodized design. The fabricated silicon-photonic OPA demonstrated > 40° lateral beam steering with no sidelobes in a ± 33° field-of-regard and 3.3° longitudinal beam steering via wavelength tuning by 20 nm centered at 1550 nm. We have fully integrated the silicon photonic OPA device with electronic controls and successfully demonstrated 2-dimensional coherent optical beam steering of pre-planned far-field patterns. Future improvements include placement of a distributed Bragg reflector (DBR) underneath the grating emitter in order to achieve nearly a factor of two improvement in emission efficiency.

Nonlinear optical properties of MoS2-WS2 heterostructure in fiber lasers.

Abstract: As a saturable absorption material, the heterostructure with the van der Waals structure has been paid much attention in material science. In general, the heterogeneous combination is able to neutralize, or even exceed, the individual material's advantages in some aspects. In this paper, which describes the magnetron sputtering deposition method, the tapered fiber is coated by the MoS2-WS2 heterostructure, and the MoS2-WS2 heterostructure saturable absorber (SA) is fabricated. The modulation depth of the prepared MoS2-WS2 heterostructure SA is measured to be 19.12%. Besides, the theoretical calculations for the band gap and carrier mobility of the MoS2-WS2 heterostructure are provided. By employing the prepared SA, a stable and passively erbium-doped fiber laser is implemented. The generated pulse duration of 154 fs is certified to be the shortest among all fiber lasers based on transition mental dichalcogenides. Results in this paper provide the new direction for the fabrication of ultrafast photon modulation devices.

Multicolor localization microscopy and point-spread-function engineering by deep learning.

Abstract: Deep learning has become an extremely effective tool for image classification and image restoration problems. Here, we apply deep learning to microscopy and demonstrate how neural networks can exploit the chromatic dependence of the point-spread function to classify the colors of single emitters imaged on a grayscale camera. While existing localization microscopy methods for spectral classification require additional optical elements in the emission path, e.g., spectral filters, prisms, or phase masks, our neural net correctly identifies static and mobile emitters with high efficiency using a standard, unmodified single-channel configuration. Furthermore, we show how deep learning can be used to design new phase-modulating elements that, when implemented into the imaging path, result in further improved color differentiation between species, including simultaneously differentiating four species in a single image.

Rapid and robust two-dimensional phase unwrapping via deep learning.

Abstract: Two-dimensional phase unwrapping algorithms are widely used in optical metrology and measurements. The high noise from interference measurements, however, often leads to the failure of conventional phase unwrapping algorithms. In this paper, we propose a deep convolutional neural network (DCNN) based method to perform rapid and robust two-dimensional phase unwrapping. In our approach, we employ a DCNN architecture, DeepLabV3+, with noise suppression and strong feature representation capabilities. The employed DCNN is first used to perform semantic segmentation to obtain the segmentation result of the wrapped phase map. We then combine the wrapped phase map with the segmentation result to generate the unwrapped phase. We benchmarked our results by comparing them with well-established methods. The reported approach out-performed the conventional path-dependent and path-independent algorithms. We also tested the robustness of the reported approach using interference measurements from optical metrology setups. Our results, again, clearly out-performed the conventional phase unwrap algorithms. The reported approach may find applications in optical metrology and microscopy imaging.