Progress 07/01/22 to 04/30/24

Outputs
Target Audience: Nothing Reported Changes/Problems:We encountered a number of problems during this project, all of which resulted in delays, and some of which resulted in changes to our research. In combination, they are why were were unable to achieve all of our goals, even with no-cost extensions. 1) Developing the hybrid numerical approximation/genetic algorithm software for designing our optics took far longer than expected. This had two causes. First, we were sidetracked trying to use a genetic algorithms (GA) development package that initially appeared that it would make our work much faster, but when we got into the details, we found that it had too many limitations for our work. As a result, we had to start over using a different GA package, costing us several weeks of work. Second, the numerical approximation used to solve the non-linear equations in designing our optics (Nelder-Mead) sometimes returns spurious results. We researched other approximations, but found that none of them could solve the equations, as the other possibilities required information about the optics we didn't have. We eventually developed work-arounds that recognized spurious results and attempted to prompt the Nelder-Mead method into producing correct results. This also slowed us by several weeks. 2) Companies that had indicated that they could make our prototype optics either went out of business or changed business between the time when we submitted our proposal and when we needed to have optical prototypes made. While we weren't able to get a definitive answer, this appears to have been caused by economic effects of the pandemic, which reduced scientific and engineering research of all kinds, including prototyping activities. Regardless of the cause, this necessitated a search for a manufacturer that could make our prototypes, which also took a few weeks. 3) Products and services we expected to use also weren't available. These included making the custom PV cells for Objective 2, using a tunable monochromator for Objectives 1 and 2, and bonding our optical prototypes. The first two of these were due to unexpected technological changes affecting products/services distant from the change. For example, tunable monochromators were previously used for testing optical fibers for telecommunications, but advances in laser diodes has lead to laser systems replacing the monochromators, causing the rental market to collapse. The inability to find professional bonding services for our prototype optics ultimately stopped our progress. We were unable to producing bonds clear enough for our testing. What we could make lowered the spectral resolution efficiency of our optics, and despite our attempts, we were unable to improve our technique enough to go further with our remaining time and funding. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Project Results Over the Full Project Duration This project addressed the economics of farming, and specificaly, it addressed the high cost of electricity paid by farmers. This project offered a solution to those high costs, which not only has a lower cost than than either grid-provided electricity or current silicon photovoltaic (PV) panels ("solar panels"), but takes up less space than current PV panels, allowing farmers to either use less space for PV electricity production, or produce more PV electricity for their own needs using the same amount of space as current silicon PV panels The high cost of electricity is particularly a problem for poultry farmers, who use a lot of electricity to keep their chickens at the proper temperature, but who have relatively limited space on their farms to produce electricity. Poultry farmers thus stand to benefit the most from our project. This project proposed using a new, inexpensive optical systems to break up (resolve) light into its component colors, producing an image like a visible spectrum of light made by a simple prism. This allows the use of PV cells matched for highest efficiency with each color, producing more electricity than current PV cells alone, which respond best only to infrared light. However, in addition to resolving light, the produced resolved-light image is five-fold smaller than the size of optics, reducing the amount of PV cells needed, lowering their cost. The production of color-resolved light concentrated to a small area had never been successfully demonstrated except by the use of complex, expensive, laboratory equipment which wasn't suitable practical use. We have made a technological advance that suggests it is possible to do that with inexpensive, commodity optical materials, which can be mass-produced for large-scale use. With respect to each objective, we accomplished: 1) We designed the optical system we proposed, a compound Fresnel prism array (CFPA) using a combination of traditional mathematical methods and genetic algorithms. The CFPA design uses inexpensive commodity materials. We then simulated its output using industry-standard optics simulation software. This showed that the CFPA was capable of resolving light with an efficiency greater than 99%. That is, over 99% of of the light ended up sorted in the usual red to violet order seen in visible spectra. The simulation also showed the the light was concentrated in an area 5x smaller than the size of the CFPA. Next, we had the parts of the CFPA made, and we attempted to bond them with optical adhesives. The bonding process didn't work well, so that when thin laser beams, only 2 mm in diameter, were scanned across the CFPA, the output was widened to approximately 12 mm. However, the roughly 5x overall size reduction remained: a roughly 200 mm wide CFPA put the light in an area 40 mm wide. This meant that the efficiency of resolving light fell dramatically, to 70%. This is, however, a better result than had previously been demonstrated, and strongly suggests that with better bonding, the CFPA will achieve its simulated potential. 2) We designed a set of PV cells that can work with the CFPA, but due to our original semiconductor manufacturer leaving the business, we constructed a proxy made of different materials. It's suitable for lab testing, but not for electricity generation. As a result of these accomplishments, we feel that even though we did not complete this project's goals entirely, we will be able to not only complete those goals, but achieve further goals including demonstrating that our system can mass produced, in future work. Project Results for Final Reporting Period Only We had four sets of our prototype optical components 3D printed from non-optical-quality materials. We tested different combinations of bonding methods using these components, including vacuum-bagging, uniform weighting/biased weighting, and amount of adhesive used. We selected the best bonding method available, and used it to bond our remaining set of prototype components. Microscope analysis showed some problems with it - voids and incomplete adhesive coverage - however, many fewer flaws were seen this time relative to our first attempt. This was confirmed by testing with three colors of laser light: 405 nm blue-violet, 650 nm red, and 780 nm near infrared. Each yielded scattered light with an average beam-width of 12 mm. The colors were in the correct spectral order, with blue-violet light at one end of the target, near infrared at the other, and red slightly shifted toward the blue from the infrared. Thus, we have excellent qualititative spectral resolution, and roughly 70% quantitative resolution.

Publications

Progress 07/01/23 to 04/30/24

Outputs
Target Audience: Nothing Reported Changes/Problems:We encounted four major problems during this reporting period's work, three of which resulted in major changes to our now, and the last of which will causes changes to future work. 1) The four month long delay in receiving the components of our optical prototype resulted us needing to request a no-cost extension, which was granted. 2) The rise of so-called "AI" generative systems and 5G telephone networks has led manufacturers of non-silicon semiconductors to emphasis production of gallium arsenide (GaAs) components, sometimes to the exclusion of all other products. Specifically, the manufacturer we had selected and received a quote from to make a gallium indium phosphide (GaInP) PV cell had left that business. When we contacted other companies, we found that they had so much work for GaAs semiconductors that they weren't willing to make a GaInP component until far after this project's completion date. 3) We were very surprised that none of the companies that routinely offer optical services, including bonding optical components with standard optical adhesives, were willing to work on our prototype. This led to us having to unexpectedly do the work ourselves. We're not experts and we didn't do a good job bonding the components. This has prevented us from carrying out expected experiments during this reporting period, and will lead to further delays as we need to develop a better bonding process. 4) We were also surprised to discover that an instrument we intended to use during testing of the combined optical system (CFPA) and test PV module, a tunable monochromator, can no longer be rented, only purchased. A tunable monochromator produces single colors of light that can be changed (tuned) from the near infrared to the near ultraviolet. This is very useful for testing the color response of PV cells and their associated optical systems, which is our objective 3 for this project. We intended to rent a tunable monochromator for this project, at a cost of a few thousand dollars. As they are no longer available to rent, we would have to purchase one for near $50,000, which we do not have the funding or approval to do. We also searched for suitable substitute equipment at academic and federal government laboratories, but were unable to locate any. As a result, we will have to use discrete (single-color) lasers for testing the color response of our system. We received permission do make this change from our grant administrator. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?During the next period of work, we will test multiple methods of bonding the second optical prototype, using inexpensive 3-D printed parts with the same shapes as the prototype components, but using materials that aren't optical quality. Once we have a working bonding process, we will use it to assemble the remaining prototype and complete this project's objectives.

Impacts
What was accomplished under these goals? As of the end of the previous progress report, we expected to have a manufactured optical prototype of our proposed light concentrator in hand by October. However, late in the design process, it became apparent that we had a serious miscommunication with the manufacturer about the design, and between correcting that, the end of year holidays, and additional work the manufacturer had taken on in the meantime, we did not receive the prototype until the first of February. However, to make up for the delay, the manufacturer provided us with parts for two prototypes. While waiting for the optical prototype to arrive, we designed a P-MJPV module to go with the optical prototype, accomplishing the first part of our second objective. Upon discovering that the expected manufacturer for the P-MJPV module had left the business of making the components, we searched for other companies to make it. However, they were booked up with other projects. As a result, we received permission from our grant administrator to substitute a module made from a pair of side-by-side silicon PV cells, and the efficiency of the high band-gap cell would be calculated from the data gathered rather than being measured directly. We then constructed this PV module, completing the second part of our second objective. By that time, the optical prototype components had arrived, and we looked for a service provider who could bond them with an optical adhesive. Because of the prototype components' unusual shapes, nobody was willing to do this work for us. Instead, we developed what we believed was an adequate process for bonding them ourselves. While the bonding appeared acceptable to the naked eye, laboratory testing by shining a 650 nm red laser through the light concentrator revealed that the light was greatly scattered at the adhesive layer, turning an approximately 2 mm diameter input beam into a 55 mm-long output streak. As the PV module we intended to use with this the optical prototype was only 40 mm across, this was clearly not functional. Microscope examination of the bonded prototype revealed numerous tiny voids, areas of poor adhesive coverage, and spots of smeared adhesive. All of these doubtless contributed to scattering the laser light. Further, direct measurements of the unassembled components of the second prototype with the laser found negligible scattering from the optics themselves. This confirms that the bonding process is the problem.

Publications

Progress 07/01/22 to 06/30/23

Outputs
Target Audience: Nothing Reported Changes/Problems:We encounter four major problems in our research to date. 1) Unexpected Complexity of the Underlying Software Our software for designing our optics makes use of software developed by two academic researchers, Nathan Hagen and Tomasz S. Tkaczyk, for designing individual compound prisms. In our preliminary research, which led to this project, their software seemed adequate for the modifications we expected to make to it for this project. However, it turns out that the overall organization of the software, both in terms of how data is stored in it, and how different parts of the software are executed and which other parts of the software they make use of, make it difficult to modify at all, and especially to adapt to new purposes and connect to other pieces of software. Rewriting Hagen and Tkacyzk's software was a continual drain on researcher time for several months. However, we have finished the stage of the project during which significant changes to their software are necessary. 2) Supplier Cost Information Delays Our initial filtering of available optical materials yielded about twenty potentially suitable materials and nearly 100 families of starting points for optical designs. We then needed to get prices for these materials to move to the next step of design, which was driven by cost-effectiveness. The materials suppliers were slow to respond, and we didn't have any prices until roughly six weeks after our initial requests, and the final information we needed - which turned out to be that several promising materials were no longer made - took nearly three months. During this time, we could make only limited progress, since the results of the cost effectiveness step drove the rest of the research. 3) PI Illness The project PI suffered a medical issue that set the project back for several more weeks immediately following the previous delay. 4) Unexpected Sensitivity of the Core Software Algorithm Hagen and Tkacyzk's software relies on an algorithm to find optimal solutions to difficult mathematical problems that lack key pieces of information about the problem. Many sorts of optical design problems, including the ones involved in the project, fall into this category. The algorithm, called Nelder-Mead after its inventors, is one of the few methods for trying to find solutions to these optical problems. It doesn't always find solutions. For designing optics, this is expected: not every arrangement of materials and shapes of those materials will solve any given optical problem. During our preliminary research and the early stages of the project, the Nelder-Mead algorithm appeared to be working correctly. It often found viable optical designs for choices of optical materials, but not always. However, as we started research on specific designs, it became apparent that the Nelder-Mead approach had some problems. When designing individual sub-parts of an optical system, working from the system's top to its bottom, it would often find results for the top, but then not the bottom... unless the design problem was presented in reverse order (bottom to top), in which case it worked fine. Similarly, it would sometimes be unable to find solutions for small parts of the system, even if adjacent parts had good solutions. Investigating these errors led us to realize that Nelder-Mead is extremely sensitive to the initial conditions it's presented with for solving an optical problem. Knowing this, we were able to fix many of its problems, for example, when it encountered a spot it couldn't solve on its own, we would start it with the solution to a nearby spot it could. These sorts of fixes have been merged into our software so that they now happen automatically. This did take time, however, as researching and compensating for the limits of Nelder-Mead wasn't expected to be part of this project. There are also a few lingering problems with the algorithm that require work "by hand". When they occur, we need to take time to edit the software to make it work correctly. Unfortunately, attempting the replace Nelder-Mead with another similar algorithm only confirmed that, despite its limitations, it is the best solution available for this class of optical problem. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?By the middle of October, we expect to have the manufactured optical prototype in hand. The prototype-maker believes they can make what we need, and that it won't be a long or complicated process. Immediately thereafter, we will verify its optical properties the laboratory, and that will complete our first Objective. We expect to do this by the end of October. Despite Objective 1 being just the first of our four Objectives, it contained the bulk of the new scientific research. Objective 2 is already being planned. We believe we've identified at least two sources that can provide the photovoltaic materials we need for that step, and the remainder of the materials and equipment is readily available. The approximate sizes and shapes of the photovoltaic materials are known from the work we've already completed, and will need only small changes to produce the exact dimensions. As this Objective is more engineering than science, we expect rapid progress on it. Objective 3 simply combines the products of the first two Objectives and tests them together. This is a short stage. Objective 4 is roughly half science and half engineering, but the contractor is already at work on both the long-wave thin-film P-MJPV cell and the low-wind resistance shape needed for it. While there is a significant amount of research in this stage, it's being pursued in parallel with our current project. We believe that both our work and our contractor's work will be ready in time for completion of this Objective during our remaining time in our project extension.

Impacts
What was accomplished under these goals? We have nearly completed our first objective, the design, production, and testing of a CFPA with the stated properties, separating sunlight into two areas on a solar cell with 95% accuracy, allowing long-wave light to be converted to electricity by a silicon solar cell, which works well with long-wave light, and allowing shorter-wave light to be converted by solar cell chemistry more efficient for it as well. We completed the first step of that goal by adapting our previously-used optical modeling software to work with genetic algorithms, a class of rapid problem solvers that mimic natural evolution by choosing a set of possible solutions, selecting the best ones, making random changes to them and allowing them to "breed" with other good solutions, and repeating the process until no better solutions are found. Multiple runs of this method are used with different starting sets to probe the possible solutions. Using genetic algorithms, we were able to systematically probe the roughly six billion combinations of the 1,800 optical materials - plastics, glasses, and related materials - available on the market today. We were left with roughly 100 families of candidates to pursue further, made from approximately twenty different optical materials. These roughly 100 families fell into four groups: 1) Three kinds of any optical plastic, including both generic and proprietary plastics 2) Three kinds of all-generic plastics, of which there was only a single viable combination 3) Two kinds of any optical plastic and one layer of water 4) Two kinds of all-generic plastics and one layer of water, which also had a single viable combination The last two need some explanation. The water layer is contained in miniature hollow prisms formed by one of the adjacent plastic layers. The shape of the hollow prism allows water to act as though it was solid material in terms of how it interacts with light. No combinations of optical glasses, with or without optical plastics, yielded good solutions. Once we had information on the costs of the optical plastics, we were left with only groups 2 and 4, those involving generic plastics and possibly water. All of the proprietary materials were so expensive that even just a single millimeter-thick layer of them exceeded the cost of standard solar panels, which priced them out of any potential competition with existing solar energy products. For example, AngstromLink produces a series of proprietary materials with good optical properties for our use. The least expensive of them sells for roughly $143/kg. Given its density, a one square meter sheet a millimeter thick would cost almost $179. The solar modules we're researching will ideally produce 380 W with a square-meter aperture, so, the AngstromLink plastics would cost $0.47/W. By comparison, for utility-scale solar power plants, solar modules cost roughly $0.40/W. This left us two combinations of materials to examine in detail: the combination in group 2, SAN (styrene acrylonitrile), polycarbonate, and polystyrene; and water, acrylic, and polycarbonate. Detailed analysis of those combinations, along with some other potential combinations to serve as controls, revealed that the water, acrylic, and polycarbonate combination was the least expensive, and had excellent optical properties. In fact, we were able to show that the water layer was optional. Including a layer of water-filled prisms reduced the materials cost and improved the optical properties vs a two-layer optical system of just acrylic and polycarbonate prisms. However, we weren't able to determine the manufacturing costs of the water-filled prisms when scaled up for high-volume production. Since the project was behind schedule, and since the two-layer optical system is adequate, we decided to pursue that with the option to return to the three-layer system that includes water. In contrast to the AngstromLink materials, the two-layer acrylic-polycarbonate optics would cost approximately $0.015/W in bulk. This price is more than acceptable. We have simulated the acrylic-polycarbonate optical system in detail using OpticStudio (a commercial optics simulation) and verified that it meets the 95% light-sorting goal we stated initially. 99+% of the long wave (deep red to infrared) light is directed to one area on the simulated solar cell, and 95% of the shorter wave (red to violet) light is directed to the other area. We may be able to reach 99+% sorting overall by adjusting the goal areas slightly. When talking to a company that makes prototype optical systems, it became apparent that small changes would be needed to the design to allow it to be prototyped. Not all of the changes will be necessary for a production optical system. However, we will re-run our simulations to verify we can achieve 95% sorting with the prototype design, as that will guarantee that a production design will achieve our design goal as well.

Publications