Preprints PDF Link Abstract
V. Dumont, C. Garner, A. Trivedi, C. Jones, V. Ganapati, J. Mueller, T. Perciano, M. Kiran, M. Day. "HYPPO: A Surrogate-Based Multi-Level Parallelism Tool for Hyperparameter Optimization." Oct. 2021.
Abstract:

We present a new software, HYPPO, that enables the automatic tuning of hyperparameters of various deep learning (DL) models. Unlike other hyperparameter optimization (HPO) methods, HYPPO uses adaptive surrogate models and directly accounts for uncertainty in model predictions to find accurate and reliable models that make robust predictions. Using asynchronous nested parallelism, we are able to significantly alleviate the computational burden of training complex architectures and quantifying the uncertainty. HYPPO is implemented in Python and can be used with both TensorFlow and PyTorch libraries. We demonstrate various software features on time-series prediction and image classification problems as well as a scientific application in computed tomography image reconstruction. Finally, we show that (1) we can reduce by an order of magnitude the number of evaluations necessary to find the most optimal region in the hyperparameter space and (2) we can reduce by two orders of magnitude the throughput for such HPO process to complete.

Y. F. Cheng, Z. Sabry, M. Strachan, S. Cornell, J. Chanenson, E.-M. S. Collins, V. Ganapati. "Deep Learned Optical Multiplexing for Multi-Focal Plane Microscopy." Jul. 2019.
Abstract:

To obtain microscope images at multiple focal planes, the distance between the objective and sample can be mechanically adjusted. Images are acquired sequentially at each axial distance. Digital refocusing with a light-emitting diode (LED) array microscope allows elimination of this mechanical movement. In an LED array microscope, the light source of a conventional widefield microscope is replaced with a 2-dimensional LED matrix. A stack of images is acquired from the LED array microscope by sequentially illuminating each LED and capturing an image. Previous work has shown that we can achieve digital refocusing by post-processing this LED image stack. Though mechanical scanning is eliminated, digital refocusing with an LED array microscope has low temporal resolution due to the acquisition of multiple images.

In this work, we propose a new paradigm for multi-focal plane microscopy for live imaging, utilizing an LED array microscope and deep learning. In our deep learning approach, we look for a single LED illumination pattern that allows the information from multiple focal planes to be multiplexed into a single image. We jointly optimize this LED illumination pattern with the parameters of a post-processing deep neural network, using a training set of LED image stacks from fixed, not live, Dugesia japonica planarians. Once training is complete, we obtain multiple focal planes by inputting a single multiplexed LED image into the trained post-processing deep neural network. We demonstrate live imaging of a D. japonica planarian at 5 focal planes with our method.

V. Ganapati, T. P. Xiao, E. Yablonovitch. "Ultra-Efficient Thermophotovoltaics Exploiting Spectral Filtering by the Photovoltaic Band-Edge." Nov. 2016.
Abstract:

Thermophotovotaics convert thermal radiation from local heat sources to electricity. A new breakthrough in creating highly efficient thin-film solar cells can potentially enable thermophotovoltaic systems with unprecedented high efficiency. The current 28.8% single-junction solar efficiency record, by Alta Devices, was achieved by recognizing that a good solar cell needs to reflect infrared band-edge radiation at the back surface, to effectively recycle infrared luminescent photons. The effort to reflect band-edge luminescence in solar cells has serendipitously created the technology to reflect all infrared wavelengths, which can revolutionize thermophotovoltaics. We have never before had such high back reflectivity for sub-bandgap radiation, permitting step-function spectral control for the first time. Thus, contemporary efficiency advances in solar photovoltaic cells create the possibility of realizing a >50% efficient thermophotovoltaic system.


Journal Publications PDF Link Abstract
Z. Omair, G. Scranton, L. M. Pazos-Outón, T. P. Xiao, M. A. Steiner, V. Ganapati, P. F. Peterson, J. Holzrichter, H. A. Atwater, E. Yablonovitch. "Ultraefficient thermophotovoltaic power conversion by band-edge spectral filtering." Proceedings of the National Academy of Sciences, Jul. 2019.
Abstract:

Thermophotovoltaic conversion utilizes thermal radiation to generate electricity in a photovoltaic cell. On a solar cell, the addition of a highly reflective rear mirror maximizes the extraction of luminescence, which in turn boosts the voltage. This has enabled the creation of record-breaking solar cells. The rear mirror also reflects low-energy photons back into the emitter, recovering the energy. This radically improves thermophotovoltaic efficiency. Therefore, the luminescence extraction rear mirror serves a dual function; boosting the voltage, and reusing the low-energy thermal photons. Owing to the dual functionality of the rear mirror, we achieve a thermophotovoltaic efficiency of 29.1% at 1,207 °C, a temperature compatible with furnaces, and a new world record at temperatures below 2,000 °C.

Y. F. Cheng, M. Strachan, Z. Weiss, M. Deb, D. Carone, V. Ganapati. "Illumination pattern design with deep learning for single-shot Fourier ptychographic microscopy." Optics Express, vol. 27, no. 2, pp. 644-656, Jan. 2019.
Abstract:

Fourier ptychographic microscopy allows for the collection of images with a high space-bandwidth product at the cost of temporal resolution. In Fourier ptychographic microscopy, the light source of a conventional widefield microscope is replaced with a light-emitting diode (LED) matrix, and multiple images are collected with different LED illumination patterns. From these images, a higher-resolution image can be computationally reconstructed without sacrificing field-of-view. We use deep learning to achieve single-shot imaging without sacrificing the space-bandwidth product, reducing the acquisition time in Fourier ptychographic microscopy by a factor of 69. In our deep learning approach, a training dataset of high-resolution images is used to jointly optimize a single LED illumination pattern with the parameters of a reconstruction algorithm. Our work paves the way for high-throughput imaging in biological studies.

A. Robey and V. Ganapati. "Optimal physical preprocessing for example-based super-resolution." Optics Express, vol. 26, no. 24, pp. 31333-31350, Nov. 2018.
Abstract:

Fourier ptychographic microscopy is a technique that achieves a high space-bandwidth product, i.e. high resolution and high field-of-view. In Fourier ptychographic microscopy, variable illumination patterns are used to collect multiple low-resolution images. These low-resolution images are then computationally combined to create an image with resolution exceeding that of any single image from the microscope. Due to the necessity of acquiring multiple low-resolution images, Fourier ptychographic microscopy has poor temporal resolution. Our aim is to improve temporal resolution in Fourier ptychographic microscopy, achieving single-shot imaging without sacrificing space-bandwidth product. We use example-based super-resolution to achieve this goal by trading off generality of the imaging approach. In example-based super-resolution, the function relating low-resolution images to their high-resolution counterparts is learned from a given dataset. We take the additional step of modifying the imaging hardware in order to collect more informative low-resolution images to enable better high-resolution image reconstruction. We show that this “physical preprocessing” allows for improved image reconstruction with deep learning in Fourier ptychographic microscopy. In this work, we use deep learning to jointly optimize a single illumination pattern and the parameters of a post-processing reconstruction algorithm for a given sample type. We show that our joint optimization yields improved image reconstruction as compared with sole optimization of the post-processing reconstruction algorithm, establishing the importance of physical preprocessing in example-based super-resolution.

V. Ganapati, M. A. Steiner, E. Yablonovitch. "The Voltage Boost Enabled by Luminescence Extraction in Solar Cells." IEEE Journal of Photovoltaics, vol. 6, no. 4, pp. 801-809, Apr. 2016.
Abstract:

Over the past few years, the application of the physical principle, i.e., “luminescence extraction,” has produced record voltages and efficiencies in photovoltaic cells. Luminescence extraction is the use of optical design, such as a back mirror or textured surfaces, to help internal photons escape out of the front surface of a solar cell. The principle of luminescence extraction is exemplified by the mantra “a good solar cell should also be a good LED.” Basic thermodynamics says that the voltage boost should be related to concentration ratio \(C\) of a resource by \(\Delta V = (kT/q) \ln\{C\}\). In light trapping (i.e., when the solar cell is textured and has a perfect back mirror), the concentration ratio of photons \(C = {4n^2}\); therefore, one would expect a voltage boost of \(\Delta V = (kT/q) \ln\{4n^2\}\) over a solar cell with no texture and zero back reflectivity, where n is the refractive index. Nevertheless, there has been ambiguity over the voltage benefit to be expected from perfect luminescence extraction. Do we gain an open-circuit voltage boost of \(\Delta V = (kT/q) \ln\{n^2\}\), \(\Delta V = (kT/q) \ln\{2n^2\}\), or \(\Delta V = (kT/q) \ln\{4n^2\}\)? What is responsible for this voltage ambiguity \(\Delta V = (kT/q) \ln\{4\}\approx 36\) \(mV\)? We show that different results come about, depending on whether the photovoltaic cell is optically thin or thick to its internal luminescence. In realistic intermediate cases of optical thickness, the voltage boost falls in between: \(\ln\{n^2\} < (q\Delta V/kT) < \ln\{4n^2\}\).

V. Ganapati, C.-S. Ho, E. Yablonovitch. "Air Gaps as Intermediate Selective Reflectors to Reach Theoretical Efficiency Limits of Multibandgap Solar Cells." IEEE Journal of Photovoltaics, vol. 5, no. 1, pp. 410-417, Jan. 2015.
Abstract:

Efficient external luminescence is a prerequisite for high-voltage solar cells. To approach the Shockley–Queisser limit, a highly reflective rear mirror is required. This mirror enhances the voltage of the solar cell by providing internally luminescent photons with multiple opportunities for escaping out the front surface. Likewise, intermediate reflectors in a multibandgap solar cell can assist external luminescence to enhance the voltage for each cell in a stack. These intermediate reflectors must also transmit the subbandgap photons to the next cell in the stack. A practical implementation of an intermediate selective reflector is an air gap sandwiched by antireflection coatings. The air gap provides perfect reflection for angles outside the escape cone, and the antireflection coating transmits angles inside the escape cone. As the incoming sunlight is within the escape cone, it is transmitted on to the next cell, while most of the internally trapped luminescence is reflected. We calculate that air gap intermediate reflectors, along with a rear mirror, can provide an absolute efficiency increase of ~5% in multibandgap cells.

G. Scranton, S. Bhargava, V. Ganapati, E. Yablonovitch. "Single spherical mirror optic for extreme ultraviolet lithography enabled by inverse lithography technology." Optics Express (OSA), vol. 22, no. 21, Oct. 2014.
Abstract:

Traditionally, aberration correction in extreme ultraviolet (EUV) projection optics requires the use of multiple lossy mirrors, which results in prohibitively high source power requirements. We analyze a single spherical mirror projection optical system where aberration correction is built into the mask itself, through Inverse Lithography Technology (ILT). By having fewer mirrors, this would reduce the power requirements for EUV lithography. We model a single spherical mirror system with orders of magnitude more spherical aberration than would ever be tolerated in a traditional multiple mirror system. By using ILT, (implemented by an adjoint-based gradient descent optimization algorithm), we design photomasks that successfully print test patterns, in spite of these enormous aberrations. This mathematical method was tested with a 6 plane wave illumination source. Nonetheless, it would have poor power throughput from a totally incoherent source.

V. Ganapati, O. D. Miller, E. Yablonovitch. "Light Trapping Textures Designed by Electromagnetic Optimization for Subwavelength Thick Solar Cells." IEEE Journal of Photovoltaics, vol. 4, no. 1, pp. 175-182, Jan. 2014.
Abstract:

Light trapping in solar cells allows for increased current and voltage, as well as reduced materials cost. It is known that in geometrical optics, a maximum \(4n^2\) absorption enhancement factor can be achieved by randomly texturing the surface of the solar cell, where n is the material refractive index. This ray-optics absorption enhancement (AE) limit only holds when the thickness of the solar cell is much greater than the optical wavelength. In subwavelength thin films, the fundamental questions remain unanswered: 1) what is the subwavelength AE limit and 2) what surface texture realizes this optimal AE? We turn to computational electromagnetic optimization in order to design nanoscale textures for light trapping in subwavelength thin films. For high-index thin films, in the weakly absorbing limit, our optimized surface textures yield an angle- and frequency-averaged enhancement factor ~39. They perform roughly 30% better than randomly textured structures, but they fall short of the ray optics enhancement limit of \(4n^2 \sim 50\).

V. Ganapati, D.P. Fenning, M.I. Bertoni, C.E. Kendrick, A.E. Fecych, J.M. Redwing, T. Buonassisi. "Seeding of silicon wire growth by out-diffused metal precipitates." Small, vol. 7, no. 5, pp. 563-567, Mar. 2011.
Abstract:

We propose the out-diffused metal precipitates (OMP) method to seed metal catalysts for bottom-up silicon wire growth. We first in-diffuse the silicon substrate with a fast-diffusing metal (e.g., copper), and then anneal with a temperature profile tuned to out-diffuse the metal to favorable nucleation sites on the surface. Vapor–liquid–solid (VLS) silicon wire growth on seeds from the OMP method is demonstrated. The OMP method has the potential to seed wires of any size at any position on a three-dimensional surface, in a high-throughput manner.

V. Ganapati, S. Schoenfelder, S. Castellanos, S. Oener, R. Koepge, A. Sampson, M.A. Marcus, B. Lai, H. Morhenn, G. Hahn, J. Bagdahn, T. Buonassisi. "Infrared birefringence imaging of residual stress and bulk defects in multicrystalline silicon." Journal of Applied Physics, vol. 108, Jul. 2010.
Abstract:

This manuscript concerns the application of infrared birefringence imaging (IBI) to quantify macroscopic and microscopic internal stresses in multicrystalline silicon mc-Si solar cell materials. We review progress to date, and advance four closely related topics. (1) We present a method to decouple macroscopic thermally-induced residual stresses and microscopic bulk defect related stresses. In contrast to previous reports, thermally-induced residual stresses in wafer-sized samples are generally found to be less than 5 MPa, while defect-related stresses can be several times larger. (2) We describe the unique IR birefringence signatures, including stress magnitudes and directions, of common microdefects in mc-Si solar cell materials including: \(\beta-SiC\) and \(\beta-Si_3N_4\) microdefects, twin bands, nontwin grain boundaries, and dislocation bands. In certain defects, local stresses up to 40 MPa can be present. (3) We relate observed stresses to other topics of interest in solar cell manufacturing, including transition metal precipitation, wafer mechanical strength, and minority carrier lifetime. (4) We discuss the potential of (IBI) as a quality-control technique in industrial solar cell manufacturing.


Conference Papers PDF Link Abstract
T. P. Xiao, G. Scranton, V. Ganapati, J. Holzrichter, P. F. Peterson, E. Yablonovitch. "Enhancing the Efficiency of Thermophotovoltaics with Photon Recycling." Conference on Lasers and Electro-Optics (CLEO), Jun. 2016.
Abstract:

We show that photon recycling utilizing the spectral selectivity of the photovoltaic band edge enables 48% thermophotovoltaic heat to electricity conversion efficiency at \(1200^{\circ} C\) with \(ln_{0.47}Ga_{0.53}As\) cells, and present experimental methods to demonstrate this concept.

G. Scranton, T. P. Xiao, V. Ganapati, J. Holzrichter, P. F. Peterson, E. Yablonovitch. "Highly Efficient Thermophotovoltaics Enabled by Photon Re-Use." IEEE Photovoltaic Specialists Conference (PVSC 43), Jun. 2016.
Abstract:

Thermophotovoltaic efficiencies above 50% may soon be realizable due to recent advances in thin film photovoltaics. Highly efficient thin film photovoltaics cells are highly reflective in the below bandgap spectral region. In a thermophotovoltaic system, this allows the below bandgap radiation to be reflected back to the emitter, so that their energy can be used to reheat the source, rather than being lost. In this work, we present a substantial improvement in the thermophotovoltaic conversion efficiency in the presence of photon recycling. We also predict the achievable conversion efficiency for a system that uses \(ln_{0.53}Ga_{0.47}As\) photovoltaic cells, and present an experimental cavity to be used for future efficiency measurements.

V. Ganapati, L. Waller, E. Yablonovitch. "Adjoint Method for Phase Retrieval." Computational Optical Sensing and Imaging (COSI), Jun. 2014.
Abstract:

We describe an adjoint method for phase retrieval of a wavefront from measurements of intensity along the axial direction, assuming Fresnel propagation. This method allows efficient computation of gradients for iterative phase retrieval.

C. Lalau-Keraly, S. Bhargava, V. Ganapati, E. Yablonovitch. "Shape Optimization of Nanophotonic Devices Using the Adjoint Method." Conference on Lasers and Electro-Optics (CLEO), Jun. 2014.
Abstract:

We use an adjoint method integrated with a classical Maxwell solver to optimize several nanophotonic devices, providing an adaptable and efficient tool for photonics design.

S. Bhargava, O. D. Miller, V. Ganapati, E. Yablonovitch. "Inverse Design of Optical Antennas for Sub-Wavelength Energy Delivery." Conference on Lasers and Electro-Optics (CLEO), Jun. 2013.
Abstract:

We report using Inverse Electromagnetic Design to computationally optimize optical antenna shapes. Optimized antennas deliver 10% of incident power to a \(50\times40\times10\) nm\(^3\) spot in a practical magnetic recording medium for Heat Assisted Magnetic Recording.

V. Ganapati, O. D. Miller, E. Yablonovitch. "Inverse electromagnetic design for subwavelength light trapping in solar cells." IEEE Photonics Conference (IPC), pp. 191-192, Sep. 2012.
Abstract:

In the subwavelength regime, the optimal surface texture for light trapping in solar cells remains to be found. We use computational inverse electromagnetic design to find the optimal nanoscale surface texture.

V. Ganapati, O. D. Miller, E. Yablonovitch. "Spontaneous symmetry-breaking in the optimization of subwavelength solar cell textures for light trapping." IEEE Photovoltaic Specialists Conference (PVSC 38), Jun. 2012.
Abstract:

Light trapping in solar cells allows for increased efficiency and reduced materials cost. It is well known that a \(4n^2\) factor of enhancement in absorption can be achieved by randomly texturing the surface of the solar cell, where n is the refractive index of the material. However, this limit only holds when the thickness of the solar cell is much greater than the wavelength of light. In the subwavelength regime, the fundamental question remains unanswered: what surface texture realizes the optimal absorption enhancement? We turn to computational inverse electromagnetic design in order to find this optimal nanoscale texture for light trapping, and observe spontaneous symmetry breaking in the final design. We achieve a factor of \(40\) in enhancement at normal incidence and above \(20\) for angle averaged incidence (averaged over an energy bandwidth of \(1/8\)) for \(n = 3.5\).

O. D. Miller, V. Ganapati, E. Yablonovitch. "Inverse Design of a Nano-Scale Surface Texture for Light Trapping." CLEO: Science and Innovations, May. 2012.
Abstract:

We introduce computational inverse design to optimize nano-scale surface textures for light trapping. The approach yields a structure with a 40.8 absorption enhancement factor, the highest reported for a high-index material in the full-wave domain

V. Ganapati, S. Schoenfelder, S. Castellanos, S. Oener, T. Buonassisi. "Infrared birefringence imaging of residual stress and bulk defects in multicrystalline silicon." IEEE Photovoltaic Specialists Conference (PVSC 35), Jun. 2010.
Abstract:

We explore the potential of infrared birefringence imaging (IBI) to reveal a complete picture of macro- and microscopic internal stresses and their origins in multicrystalline silicon (mc-Si). We present a method to decouple macroscopic thermally induced residual stresses and microscopic bulk defect-related stresses, and validate this method in mc-Si wafers via microstructural analysis. We then describe the unique IR birefringence signatures, including stress magnitudes and directions, of common microdefects in mc-Si solar cell materials: \(\beta-SiC\) and \(\beta-Si_3 N_4\) microdefects, twin bands, non-twin grain boundaries, and dislocation bands. We relate observed stresses to other topics of interest in solar cell manufacturing, including wafer mechanical strength and minority carrier lifetime.

S. Schoenfelder, A. Sampson, V. Ganapati, R. Koepge, S. Castellanos, S. Oener, T. Buonassisi. "Quantitative stress measurements of bulk microdefects in multicrystalline silicon." 24th European Photovoltaic Solar Energy Conference, pp. 977-980, Sep. 2009.
Abstract:

In this work, inclusions of silicon carbide and silicon nitride in multicrystalline silicon are investigated experimentally by photoelasticity and theoretically by numerical simulation in a finite-element model. Large tensile stresses were observed in experiment at the interface from silicon carbide to silicon while silicon nitride induced much lower stresses. The results of the finite-element model indicate that the stresses are induced while cooling during crystallization, by the mismatch of coefficients of thermal expansion.


Patents PDF Link Abstract
E. Rephaeli, V. Ganapati, D. Piponi, T. Teisseyre. "Optical implementation of machine learning for real time increased contrast via multiple wavelength illumination with tunable power." United States Patent 10,445,607 (2019).
Abstract:

An imaging system (e.g., hyperspectral imaging system) receives an indication to compare a first object and a second object (e.g., two anatomical structures or organs in a medical environment). The imaging system accesses a classification vector for the first object and the second object, the classification vector having been extracted by separating a plurality of collected reflectance values for the first object from a plurality of collected reflectance values for the second object. A set of optimal illumination intensities for one or more spectral illumination sources of the imaging system is determined based on the extracted classification vector. The first and second objects are illuminated with the determined illumination intensities. A high-contrast image of the first and second objects is provided for display, such that the two objects can be readily distinguished in the image. The intensity of pixels in the image is determined by the illumination intensities.

V. Ganapati, E. Rephaeli. "System and method for 3D scene reconstruction with dual complementary pattern illumination." United States Patent None (2019).
Abstract:

An apparatus, system and process for utilizing dual complementary pattern illumination of a scene when performing depth reconstruction of the scene are described. The method may include projecting a first reference image and a complementary second reference image on a scene, and capturing first image data and second image data including the first reference image and the complementary second reference image on the scene. The method may also include identifying features of the first reference image from features of the complementary second reference image. Furthermore, the method may include performing three dimensional (3D) scene reconstruction for image data captured by the imaging device based on the identified features in the first reference image.

V. Ganapati, E. Rephaeli, D. Piponi. "Simultaneous visible and fluorescence endoscopic imaging." United States Patent 10,386,627 (2019).
Abstract:

An endoscope apparatus includes a fiber optic cable with a proximal end and a distal end opposite the proximal end. The endoscope apparatus also includes a light source optically coupled to the proximal end of the fiber optic cable to emit visible light and excitation light into the fiber optic cable for output from the distal end. The light source is configured to emit both the visible light and the excitation light simultaneously, and a wavelength of the excitation light is outside a wavelength spectrum of the visible light. An image sensor coupled to the distal end of the fiber optic cable and positioned to receive a reflection of the visible light as reflected visible light.

V. M. Acosta and V. Ganapati. "Parallel programmable array microscope." United States Patent 9,535,242 (2017).
Abstract:

A microscope system includes a sample tray to hold a plurality of samples to be imaged in parallel. An illumination source generates illumination light and a plurality of spatial light modulators are each positioned to spatially modulate the illumination light onto a corresponding one of the samples. Relay optics are positioned in an optical path between the sample tray and the plurality of spatial light modulators to image the samples onto the plurality of spatial light modulators. A plurality of first cameras is positioned to capture images in parallel of the samples in the sample tray by imaging the plurality of spatial light modulators.