Image transmission through a multi-mode fiber is a difficult task given the complex interference of light through the fiber that leads to random speckle patterns at the distal end of the fiber. With traditional methods and techniques, it is impractical to reconstruct a high-resolution input image by using the information obtained from the intensity of the corresponding output speckle alone. In this work, we train three Convolutional Neural Networks (CNNs) with input-output couples of a multi-mode fiber and test the learning with images outside the learning set. The three implemented deep learning models have modern UNet, ResNet and VGGNet architectures and are trained with 31,200 grey-scale handwritten letters of the Latin alphabet. After the training, 5,200 images outside the learning set are used for testing and it was shown that the models successfully reconstruct the input images from the output random speckle patterns with average fidelities ranging from 81% to 90%. Our results show the superiority of the ResNet based architecture over UNet and VGGNet in reconstruction accuracy, achieving up to 97% fidelity in a short amount of time. This can be attributed to the success of the ResNet architecture in learning non-linear systems compared to its counterparts. We believe that the implementation of machine learning techniques to imaging, along with its contributions to biophysics, can reshape the telecommunication industry and thus will be a cornerstone in future optics and photonics studies.
Conversion of solar energy into electricity is crucial to meet our ever-growing energy needs. The broadband spectrum of the sunlight limits the conversion efficiency of the single- and multi-junction based solar cells. Moreover, the angle of incident radiation dramatically decreases the amount of converted energy. In fact, diffractive optical elements (DOE) designed for spectrally splitting solar light are optimized for normal incidence, and their performance drastically decreases under angled-illumination. Unfortunately, once the number of design parameters two of whose are the number of wavelengths and number of incident angles increases, computational expense for DOEs design rises. Here, we design DOEs which concentrate and split the broadband radiation under angled-illumination. In our design, we take thin, transparent and cost-effective materials into account, and we manage to disperse broadband radiation 400 nm - 1100 nm into two separate bands which are the visible band 400 nm - 700 nm and the short-IR band 701 nm - 1100 nm. Here we optimize the DOEs for angled-illumination using computationally cost-effective approaches. We observe that spectral splitting of the broadband light is less sensitive to variation of incident angle of solar radiation once DOE optimization performed for the area which is half of the output plane. As a result, 8% and 18% excess solar energy conversion can be achieved within the visible band and the short-IR band, respectively. What's interesting is that less than 0.6% deviation in output intensity can be observed when a single DOE is illuminated at angle spans from 0 to 80 degrees.
Elastic scattering intensity calculations at 90° and 0° for the transverse electric and transverse magnetic polarized light were performed at 1200nm for a 50 μm radius and 3.5 refractive index silicon microsphere. The mode spacing between morphology dependent resonances was found to be 1.76 nm. The linewidth of the morphology dependent resonances was observed to be 0.02 nm, which leads to a quality factor on the order of 104.
A silicon microsphere coupled to a silica optical fiber half coupler is excited using a diode laser operating at 1.55 μm. The transmitted and the 90o elastically scattered light signals are modulated with an electrical square wave applied to the silicon microsphere.