Category Archives: Engineering

High-yield porphyrin production through metabolic engineering and biocatalysis – Nature.com

Hiroto, S., Miyake, Y. & Shinokubo, H. Synthesis and functionalization of porphyrins through organometallic methodologies. Chem. Rev. 117, 29103043 (2017).

Article CAS PubMed Google Scholar

Barr, H. et al. Eradication of high-grade dysplasia in columnar-lined (Barretts) oesophagus by photodynamic therapy with endogenously generated protoporphyrin IX. Lancet 348, 584585 (1996).

Article CAS PubMed Google Scholar

Surdel, M. C. et al. Antibacterial photosensitization through activation of coproporphyrinogen oxidase. Proc. Natl Acad. Sci. USA 114, E6652E6659 (2017).

Article CAS PubMed PubMed Central Google Scholar

Drury, S. L. et al. Simultaneous exposure to intracellular and extracellular photosensitizers for the treatment of Staphylococcus aureus infections. Antimicrob. Agents Chemother. 65, e0091921 (2021).

Poulos, T. L. Heme enzyme structure and function. Chem. Rev. 114, 39193962 (2014).

Article CAS PubMed PubMed Central Google Scholar

Zhang, J. et al. Recent advances in microbial production of high-value compounds in the tetrapyrrole biosynthesis pathway. Biotechnol. Adv. 55, 107904 (2022).

Article CAS PubMed Google Scholar

Choi, K. R., Yu, H. E., Lee, H. & Lee, S. Y. Improved production of heme using metabolically engineered Escherichia coli. Biotechnol. Bioeng. 119, 31783193 (2022).

Article CAS PubMed Google Scholar

Zhang, W., Lai, W. & Cao, R. Energy-related small molecule activation reactions: oxygen reduction and hydrogen and oxygen evolution reactions catalyzed by porphyrin- and corrole-based systems. Chem. Rev. 117, 37173797 (2017).

Article CAS PubMed Google Scholar

Singh, S. et al. Glycosylated porphyrins, phthalocyanines, and other porphyrinoids for diagnostics and therapeutics. Chem. Rev. 115, 1026110306 (2015).

Article CAS PubMed PubMed Central Google Scholar

Espinas, N. A., Kobayashi, K., Takahashi, S., Mochizuki, N. & Masuda, T. Evaluation of unbound free heme in plant cells by differential acetone extraction. Plant Cell Physiol. 53, 13441354 (2012).

Article CAS PubMed Google Scholar

In, M.-J., Kim, D. C., Chae, H. J. & Oh, N.-S. Effects of degree of hydrolysis and pH on the solubility of heme-iron enriched peptide in hemoglobin hydrolysate. Biosci. Biotechnol. Biochem. 67, 365367 (2003).

Article CAS PubMed Google Scholar

Lichtenthaler, H. K. & Buschmann, C. Extraction of phtosynthetic tissues: chlorophylls and carotenoids. Curr. Protoc. Food Anal. Chem. 1, F4.2.1F4.2.6 (2001).

Article Google Scholar

Kwon Seok, J., de Boer Arjo, L., Petri, R. & Schmidt-Dannert, C. High-level production of porphyrins in metabolically engineered Escherichia coli: systematic extension of a pathway assembled from overexpressed genes involved in heme biosynthesis. Appl. Environ. Microbiol. 69, 48754883 (2003).

Article PubMed PubMed Central Google Scholar

Bali, S. et al. Molecular hijacking of siroheme for the synthesis of heme and d1 heme. Proc. Natl Acad. Sci. USA 108, 1826018265 (2011).

Article CAS PubMed PubMed Central Google Scholar

Dailey Harry, A. et al. Prokaryotic heme biosynthesis: multiple pathways to a common essential product. Microbiol. Mol. Biol. Rev. 81, e00048-16 (2017).

PubMed PubMed Central Google Scholar

Dailey, H. A., Gerdes, S., Dailey, T. A., Burch, J. S. & Phillips, J. D. Noncanonical coproporphyrin-dependent bacterial heme biosynthesis pathway that does not use protoporphyrin. Proc. Natl Acad. Sci. USA 112, 22102215 (2015).

Article CAS PubMed PubMed Central Google Scholar

Fang, H. et al. Metabolic engineering of Escherichia coli for de novo biosynthesis of vitamin B12. Nat. Commun. 9, 4917 (2018).

Article PubMed PubMed Central Google Scholar

Chen, G. E. et al. Complete enzyme set for chlorophyll biosynthesis in Escherichia coli. Sci. Adv. 4, eaaq1407 (2018).

Article PubMed PubMed Central Google Scholar

Nielsen, M. T. et al. Assembly of highly standardized gene fragments for high-level production of porphyrins in E. coli. ACS Synth. Biol. 4, 274282 (2015).

Article CAS PubMed Google Scholar

Zhang, J. et al. Heme biosensor-guided in vivo pathway optimization and directed evolution for efficient biosynthesis of heme. Biotechnol. Biofuels Bioprod. 16, 33 (2023).

Article PubMed PubMed Central Google Scholar

Dai, J. et al. Differential gene content and gene expression for bacterial evolution and speciation of Shewanella in terms of biosynthesis of heme and heme-requiring proteins. BMC Microbiol. 19, 173 (2019).

Article PubMed PubMed Central Google Scholar

Ouchane, S., Picaud, M., Therizols, P., Reiss-Husson, F. & Astier, C. Global regulation of photosynthesis and respiration by FnrL: the first two targets in the tetrapyrrole pathway. J. Biol. Chem. 282, 76907699 (2007).

Article CAS PubMed Google Scholar

Toriya, M. et al. Zincphyrin, a novel coproporphyrin III with zinc from Streptomyces sp. J. Antibiot. (Tokyo) 46, 196200 (1993).

Article CAS PubMed Google Scholar

Nguyen, H. T. et al. Exploration of cryptic organic photosensitive compound as Zincphyrin IV in Streptomyces venezuelae ATCC 15439. Appl. Microbiol. Biotechnol. 104, 713724 (2020).

Article CAS PubMed Google Scholar

Cleary, J. L., Kolachina, S., Wolfe, B. E. & Sanchez, L. M. Coproporphyrin III produced by the bacterium Glutamicibacter arilaitensis binds zinc and is upregulated by fungi in cheese rinds. mSystems. 3, e00036-18 (2018).

Zhao, X. R., Choi, K. R. & Lee, S. Y. Metabolic engineering of Escherichia coli for secretory production of free haem. Nat. Catal. 1, 720728 (2018).

Article CAS Google Scholar

Ko, Y. J. et al. Animal-free heme production for artificial meat in Corynebacterium glutamicum via systems metabolic and membrane engineering. Metab. Eng. 66, 217228 (2021).

Article CAS PubMed Google Scholar

Ishchuk, O. P. et al. Genome-scale modeling drives 70-fold improvement of intracellular heme production in Saccharomyces cerevisiae. Proc. Natl Acad. Sci. USA 119, e2108245119 (2022).

Article CAS PubMed PubMed Central Google Scholar

Choi, K. R., Yu, H. E. & Lee, S. Y. Production of zinc protoporphyrin IX by metabolically engineered Escherichia coli. Biotechnol. Bioeng. 119, 33193325 (2022).

Article CAS PubMed Google Scholar

Nelson, N. Metal ion transporters and homeostasis. EMBO J. 18, 43614371 (1999).

Article CAS PubMed PubMed Central Google Scholar

Frunzke, J., Gtgens, C., Brocker, M. & Bott, M. Control of heme homeostasis in Corynebacterium glutamicum by the two-component system HrrSA. J. Bacteriol. 193, 12121221 (2011).

Article CAS PubMed PubMed Central Google Scholar

Koripella, R. K. et al. Mechanism of elongation factor-G-mediated fusidic acid resistance and fitness compensation in Staphylococcus aureus. J. Biol. Chem. 287, 3025730267 (2012).

Article CAS PubMed PubMed Central Google Scholar

Kojima, I., Maruhashi, K., Sato, H. & Fujiwara, Y. A highly active producer of coproporphyrin III and uroporphyrin III. J. Ferment. Bioeng. 76, 527529 (1993).

Article CAS Google Scholar

Zhang, L. et al. Phosphate limitation increases coenzyme Q10 production in industrial Rhodobacter sphaeroides HY01. Synth. Syst. Biotechnol. 4, 212219 (2019).

Article PubMed PubMed Central Google Scholar

Shi, T. et al. Screening and engineering of high-activity promoter elements through transcriptomics and red fluorescent protein visualization in Rhodobacter sphaeroides. Synth. Syst. Biotechnol. 6, 335342 (2021).

Article CAS PubMed PubMed Central Google Scholar

Lee, S. Q. E., Tan, T. S., Kawamukai, M. & Chen, E. S. Cellular factories for coenzyme Q10 production. Microb. Cell. Fact. 16, 39 (2017).

Article PubMed PubMed Central Google Scholar

Lu, W. et al. Identification and elimination of metabolic bottlenecks in the quinone modification pathway for enhanced coenzyme Q10 production in Rhodobacter sphaeroides. Metab. Eng. 29, 208216 (2015).

Article CAS PubMed Google Scholar

Wang, Z.-J. et al. Oxygen uptake rate controlling strategy balanced with oxygen supply for improving coenzyme Q10 production by Rhodobacter sphaeroides. Biotechnol. Bioprocess Eng. 25, 459469 (2020).

Article CAS Google Scholar

Klaus, O. et al. Engineering phototrophic bacteria for the production of terpenoids. Curr. Opin. Biotechnol. 77, 102764 (2022).

Article CAS PubMed Google Scholar

Qiang, S. et al. Elevated -carotene synthesis by the engineered rhodobacter sphaeroides with enhanced CrtY expression. J. Agric. Food Chem. 67, 95609568 (2019).

Article CAS PubMed Google Scholar

Orsi, E. et al. Growth-uncoupled isoprenoid synthesis in Rhodobacter sphaeroides. Biotechnol. Biofuels 13, 123 (2020).

Article CAS PubMed PubMed Central Google Scholar

Hu, J., Yang, H., Wang, X., Cao, W. & Guo, L. Strong pH dependence of hydrogen production from glucose by Rhodobacter sphaeroides.Int. J. Hydrog. Energy 45, 94519458 (2020).

Article CAS Google Scholar

Li, S. et al. Photoautotrophic hydrogen production of Rhodobacter sphaeroides in a microbial electrosynthesis cell. Bioresour. Technol. 320, 124333 (2021).

Article CAS PubMed Google Scholar

Orsi, E., Beekwilder, J., Eggink, G., Kengen, S. W. M. & Weusthuis, R. A. The transition of Rhodobacter sphaeroides into a microbial cell factory. Biotechnol. Bioeng. 118, 531541 (2021).

Article CAS PubMed Google Scholar

Oh, J.-I. & Kaplan, S. Generalized approach to the regulation and integration of gene expression. Mol. Microbiol. 39, 11161123 (2001).

Article CAS PubMed Google Scholar

Imam, S., Noguera, D. R. & Donohue, T. J. Global analysis of photosynthesis transcriptional regulatory networks. PLoS Genet. 10, e1004837 (2014).

Article PubMed PubMed Central Google Scholar

Kang, Z. et al. Recent advances in microbial production of -aminolevulinic acid and vitamin B12. Biotechnol. Adv. 30, 15331542 (2012).

Article CAS PubMed Google Scholar

Nishikawa, S. et al. Rhodobacter sphaeroides mutants which accumulate 5-aminolevulinic acid under aerobic and dark conditions. J. Biosci. Bioeng. 87, 798804 (1999).

Article CAS PubMed Google Scholar

Urakami, T. & Yoshida, T. Production of ubiquinone and bacteriochlorophyll a by Rhodobacter sphaeroides and Rhodobacter sulfidophilus. J. Ferment. Bioeng. 76, 191194 (1993).

Article CAS Google Scholar

Zeilstra-Ryalls, J. H. & Kaplan, S. Aerobic and anaerobic regulation in Rhodobacter sphaeroides 2.4.1: the role of the fnrL gene. J. Bacteriol. 177, 64226431 (1995).

Article CAS PubMed PubMed Central Google Scholar

Wei, W. et al. Lysine acetylation regulates the function of the global anaerobic transcription factor FnrL in Rhodobacter sphaeroides. Mol. Microbiol. 104, 278293 (2017).

Article CAS PubMed Google Scholar

Read the original:

High-yield porphyrin production through metabolic engineering and biocatalysis - Nature.com

Howard University engineering school gets a record $5 million donation – WTOP

Howard University's College of Engineering and Architecture has received a $5 million gift from design software giant Autodesk, the school's largest-single unrestricted donation in its 90-year history.

Howard Universitys College of Engineering and Architecture has received a $5 million gift from design software giant Autodesk, the schools largest-single unrestricted donation in its 90-year history.

Last year, Autodesk donated $1 million to Howard Universitys Department of Mechanical Engineering.

The school will use the latest gift to fund its new state-of-the-art design and make lab, which it plans to unveil this fall. In addition to interdisciplinary collaboration for Howard University students, the labor will be used for K-12 programs for young students

Autodesk offers its design software suites to students and educators for free worldwide. Howard Universitys College of Engineering and Architecture has been using Autodesk software in classrooms for the past six years.

Howard is a leader in graduating Black engineers, and this gift will expand our ability to provide our students with the best and latest resources throughout the course of their education here, said Howard University president Ben Vinson III.

Howard quotes research from career and jobs website Zippia that found only 3% of mechanical engineers in the U.S. are Black. Howard said it is committed to increasing opportunities.

Get breaking news and daily headlines delivered to your email inbox by signing up here.

2024 WTOP. All Rights Reserved. This website is not intended for users located within the European Economic Area.

See the rest here:

Howard University engineering school gets a record $5 million donation - WTOP

Introducing USC’s Molinaroli College of Engineering and Computing – University of South Carolina

Posted on: June 6, 2024; Updated on: June 6, 2024 By Gregory Hardy, ghardy@sc.edu

The University of South Carolina College of Engineering and Computing has a new name that reflects the vision and generosity of generations of the Molinaroli family.

The official name, the University of South Carolina Molinaroli College of Engineering and Computing, was announced on Thursday (June 6) during a signage unveiling ceremony at the colleges Swearingen Engineering Center on Main Street.

The Molinaroli family boasts strong ties to South Carolina and a deep commitment to the future of the university and the Palmetto State. Seven family members have graduated from USC since the 1930s and another will begin this fall. Their engineering impact spans from the Charleston Navy Yard to the South Carolina Department of Transportation and Johnson Controls.

The $30 million investment initiated by 1983 alumnus Alex Molinaroli and his wife, Kristin Ihle Molinaroli, will be used to establish the college as a leading educational and research institution throughout the Mid-Atlantic and Southeast and a driving force behind South Carolinas economic growth.

Among those expressing gratitude Thursday was USC President Michael Amiridis, who served as the dean of the College of Engineering and Computing from 2006 to 2009.

With this extraordinary gift from Alex and Kristin Molinaroli, we are embarking on a new era for the College of Engineering and Computing, Amiridis said. Their generous commitment will empower the college to enhance its academic stature and explore its fullest potential for exceptional education, research, workforce development and economic impact in South Carolina and beyond.

This marks the universitys fourth academic unit to be named for a donor. The Darla Moore School of Business was named for financial investor and alumna Darla Moore in 1998, and the Arnold School of Public Health was named in 2000 for business leader Norman J. Arnold. Most recently, the Joseph F. Rice School of Law was named in November 2023 for plaintiffs trial lawyer and alumnus Joe Rice.

"A degree from the University of South Carolina can change your life and be an enabler to provide generational change for your family," says Alex Molinaroli. "The University of South Carolina always was and still is a family affair. It is a privilege for my familys name to be associated with the College of Engineering and Computing as it becomes an incubator and foundation developer for the students who will be the next Fortune 100 CEOs and successful entrepreneurs."

The former CEO of Johnson Controls, Alex Molinaroli has been a strong advocate of his alma mater for many years. He has played a pivotal role, offering valuable insights and guidance to college leadership. His previous contributions helped spur innovation and provide opportunities for student advancement.

"The University of South Carolina always was and still is a family affair. It is a privilege for my familys name to be associated with the College of Engineering and Computing as it becomes an incubator and foundation developer for the students who will be the next Fortune 100 CEOs and successful entrepreneurs."

The current gift will support the creation of new, high-demand programs; student and faculty recruitment; facility and equipment enhancements; and research start-ups.

The generous gift honors multiple generations of the Molinaroli family, including Alexs father, Adrian (51), and uncle, Remo (34), who were both first-generation engineering students at USC; his aunt, Elenora (33); his brother, Raymond (91), who is a Lowcountry civil engineer; and two cousins, Charles (65) and Marion (69).

We are grateful for the Molinarolis extraordinary generosity and shared vision for our colleges role in shaping a better future for our state, region and nation, says Hossein Haj-Hariri, dean of the College of Engineering and Computing. Their gift will enable us to enhance our educational, research and partnership efforts to transform our state.

Founded in 1909, the Molinaroli College of Engineering and Computing offers high-quality education through 40 degree programs, including the states only aerospace, biomedical and graduate nuclear engineering programs, as well as numerous computing and AI options.

Nearly 70 percent of students are South Carolinians, and a significant majority of alumni choose to apply their skills within our state's borders.

The colleges thriving academic and research excellence continue to drive economic growth and spur social mobility in the Palmetto State. Leading in energy systems, advanced manufacturing and intelligent infrastructure research, the college will set a university record in fiscal year 2024, receiving over $70 million in sponsored awards.

Read more:

Introducing USC's Molinaroli College of Engineering and Computing - University of South Carolina

New method extracts lithium from seawater, to boost battery production – Interesting Engineering

Researchers have optimized a new method for extracting lithium from widespread sources such as seawater, groundwater, and flowback water (a byproduct of fracking and offshore drilling).

Developed by researchers at the University of Chicago Pritzker School of Molecular Engineering (PME), the method shows how certain particles of iron phosphate can most efficiently pull lithium out of dilute liquids.

The new method is expected to hasten an era of faster, greener lithium extraction.

Our method allows the efficient extraction of the mineral from very dilute liquids, which can greatly broaden the potential sources of lithium, said Chong Liu, Neubauer Family Assistant Professor of Molecular Engineering.

Right now there is a gap between the demand for lithium and the production. Our method allows the efficient extraction of the mineral from very dilute liquids, which can greatly broaden the potential sources of lithium.

The method isolates lithium based on its electrochemical properties, using crystal lattices of olivine iron phosphate.

Because of its size, charge and reactivity, lithium is drawn into the spaces in the olivine iron phosphate columns like water being soaked into the holes in a sponge. But, if the column is designed perfectly, sodium ions, also present in briny liquids, are left out or enter the iron phosphate at a much lower level, according to the study.

Researchers tested how variation in olivine iron phosphate particles impacted their ability to selectively isolate lithium over sodium.

When you produce iron phosphate, you can get particles that are drastically different sizes and shapes, said PhD student and first author Gangbin Yan.

In order to figure out the best synthesis method, we need to know which of those particles are most efficient at selecting lithium over sodium.

The study details how researchers synthesized olivine iron phosphate particles using diverse methods, resulting in particle sizes ranging from 20 to 6,000 nanometers. These particles were then grouped by size and used to construct electrodes for extracting lithium from a weak solution, as reported by Phys.org.

Researchers observed that overly large or small iron phosphate particles tended to allow more sodium into their structures, leading to less pure lithium extractions.

It turned out that there was this sweet spot in the middle where both the kinetics and the thermodynamics favor lithium over sodium, said Liu.

We have to keep this desired particle size in mind as we pick synthesis methods to scale up But if we can do this, we think we can develop a method that reduces the environmental impact of lithium production and secures the lithium supply in this country.

Amid a rising demand for electric vehicles, the demand for lithium the mineral required for lithium-ion batteries has also soared. However, current methods of extracting lithium from rock ores or brines are slow and come with high energy demands and environmental costs. In contrast, the new method is environment-friendly and faster than other current methods.

The study was published in the journal Nature on June 7.

NEWSLETTER

Stay up-to-date on engineering, tech, space, and science news with The Blueprint.

Prabhat Ranjan Mishra Prabhat, an alumnus of the Indian Institute of Mass Communication, is a tech and defense journalist. While he enjoys writing on modern weapons and emerging tech, he has also reported on global politics and business. He has been previously associated with well-known media houses, including the International Business Times (Singapore Edition) and ANI.

Continued here:

New method extracts lithium from seawater, to boost battery production - Interesting Engineering

Engineering Cancer’s End: Moffitt Scientists Say Bioengineering Will Change Our Ability To Research And Treat Cancer – Eurasia Review

Bioengineering is revolutionizing cancer research, andMoffitt Cancer Centeris at the forefront of this transformative movement. Moffitt is the first National Cancer Institute-designated comprehensive cancer center with a dedicated bioengineering department. This area of science integrates engineering and physical sciences with oncology to change how we understand and treat this complex disease. In anew commentary publishedinCancer Cell,W. Gregory Sawyer, Ph.D., andElsa R. Flores, Ph.D., share their visionary framework to accelerate cancer discovery and therapy breakthroughs through bioengineering.

Cancers complexity has been a formidable obstacle for researchers, said Sawyer, chair of MoffittsDepartment of Bioengineering. Traditional methods often struggle to capture the intricate interplay between cancer cells, the immune system and the surrounding environment. Cancer engineering offers a unique perspective by integrating these diverse fields, creating a powerful platform to develop next-generation solutions.

Cancer engineering blends 12 key fields, including system dynamics, nanomaterials, robotics, and biofabrication, to tackle cancer from all angles. This powerful platform could lead to advancements in early detection with microfluidic devices and advanced imaging techniques. Additionally, nanomaterials engineered on a microscopic level could revolutionize drug delivery by transporting medications directly to cancer cells with minimal impact on healthy tissues.

The potential doesnt stop there. 3D bioprinting technology offers the potential to create customized tumor models, allowing researchers to test drug efficacy and personalize treatment plans for individual patients. Sophisticated mathematical modeling, informed by engineering principles, could provide a deeper understanding of cancers intricate biological processes, paving the way for developing more effective therapies.

The possibilities unlocked by cancer engineering are truly exciting, said Flores, associate center director ofBasic Scienceat Moffitt. We envision more universities and cancer centers following Moffitts lead and creating dedicated cancer engineering programs to foster collaboration and accelerate progress in the fight against cancer.

Here is the original post:

Engineering Cancer's End: Moffitt Scientists Say Bioengineering Will Change Our Ability To Research And Treat Cancer - Eurasia Review

Solvent engineering for scalable fabrication of perovskite/silicon tandem solar cells in air – Nature.com

Distinction of different alcohols as solvents

The perovskite films were fabricated by a two-step sequential deposition method based on previous work15,23. As depicted in Fig.1a, our process combines co-evaporation and blade-coating techniques to meet the requirements for large-area fabrication of the perovskite films. Supplementary Fig.2 shows the deposition of an inorganic framework on both glass and textured silicon substrates. It is worth noting that the second step was implemented in air to match the realistic production environment. However, ethanol and isopropyl alcohol, which are widely used as solvents of the organic salt in the second step, confront two major challenges in the natural environment: firstly, these solvents readily absorb environmental moisture38; secondly, the rapid evaporation rate of the solution will affect the film uniformity. Consequently, these challenges often result in inhomogeneous and poor perovskite films, adversely affecting the PCE and stability of the devices.

a Schematic of the hybrid two-step deposition method. b Physical parameters of different alcohols. c Images of organic salts used different alcohols and after exposed to air for 1h. d Images of perovskite films after blade-coating organic salts without gas-quenching and annealing. The direction of blade-coating is from left to right.

To address this issue, we carried out analysis and study on various alcohols with different saturated vapor pressures and polarities including ethyl alcohol (EA), isopropanol (IPA), n-butanol (nBA) and n-pentanol (nPA). The images of different solutions after adding organic salts to the alcohol are shown in Supplementary Fig.3. For ease of expression, we refer to the following solutions, films and devices fabricated with ethanol as EA solution, EA film and EA device, as well as for IPA, nBA and nPA. As the carbon chain is lengthened, both the polarity of alcohols and the saturated vapor pressure decrease, as illustrated in Fig. 1b41. The saturated vapor pressure reflects the evaporation speed of the solvent, while the dielectric constant is positively related to the polarity of the solvent. Following the principle that like dissolves like42, the mutual solubility of alcohols and waterand thus their capacity to absorb moistureis dictated by their polarity difference. Given waters high polarity, alcohols with greater polarity are more soluble in water, leading to increased water absorption.

To investigate the impact of moisture on these different alcohol solutions in air, we exposed a measured amount of each solution to open air and observed the changes. In air environment, moisture absorption leads to rapid oxidation of I to I2, manifesting as a yellowing of the solution43,44. As shown in Fig.1c, the EA and IPA solutions turned from colorless to light yellow after one hour of exposure, while nBA and nPA solutions exhibited no significant color change, underscoring the protective effect of low polarity solvents against moisture interference. Furthermore, we compared the films after blade-coating without gas quenching and annealing on the same substrate (glass/inorganic frame) and documented the changes photographically. Figure1d illustrates that EA and IPA volatilize fastly and completely after blade-coating, in contrast to nBA and nPA films, which shows a gradual darkening. This shift signifies a decrease in volatilization rate with increasing carbon chain length, affecting perovskite crystallization dynamics. However, the slower volatilization rates results in the lingering of residual organic salts, which continues to undergo dissolution-recrystallization reactions with the perovskite45. This leads to localized accumulations of organic salts, as evidenced in Supplementary Figs.4 and 5.

Supplementary Fig.6 displays images of perovskite films fabricated using different alcohols, both in N2 and air environments. These images corroborate the notion that moisture positively affects the crystallization rate of perovskite films46, as inferred from the observable color changes. To further evaluate the effect of the volatilization rate of the solutions on the perovskite films formation, we compared the morphology and the structure of perovskite films by using a scanning electron microscope (SEM) and X-ray diffraction (XRD). Supplementary Fig.7 reveals a pronounced PbI2 signal in EA films before annealing, leading to a substantial amount of PbI2 at the bottom of the perovskite layer (Fig.2a and e). This indicates that the conversion from the inorganic framework to perovskite is incomplete. Such findings suggest that the delay of solvent volatilization rate is conducive to prolonging the reaction of inorganic frameworks with organic salt solutions in terms of promoting the transformation of the inorganic framework into perovskite.

ad Top-view and cross-sectional SEM images. e XRD patterns of perovskite films after annealing. f PL spectra of perovskite films with the emission from the glass side. g Time-resolved PL transients of perovskite films. For TRPL, double exponentials were used for fitting the curves. hj PL mapping of perovskite films with the active area for 1.5cm1.5cm.

Comparatively, perovskite films fabricated in air environment exhibit a heightened PbI2 signal (Supplementary Fig.8 and Fig.2e), demonstrating that the moisture absorbed during fabrication prompts the decomposition of perovskite films upon air annealing. Specifically, the IPA films show a strong PbI2 diffraction peak located at 12.6 (Fig.2e), which stemmed from the decomposition of perovskite films after air annealing at 35% relative humiditya finding consistent with SEM image of the IPA films (Fig.2b). Impressively, the nBA films exhibited the lowest intensity of PbI2 peak in Fig.2e, with minimal residual PbI2 particles observed on the surface (Fig.2c), indicating negligible perovskite decomposition. However, a strong PbI2 signal was found in the nPA films with the solvent volatilization rate further slowed down (Fig.2e), which was attributed to the destruction of perovskite structure by residual solution (Fig.2d). Despite the low polarity of the nPA, the reduction in solution evaporation rate inadvertently introduces excessive H2O into the perovskite films, exacerbating degradation during annealing34. The UV-vis spectra and Tauc-plots of perovskite films fabricated using various alcohols are shown in Supplementary Figs.9 and 10. Additionally, the UV-vis spectra of the inorganic framework are detailed in Supplementary Fig.9. These results elucidate both the polarity and evaporation rate of the solvent have a joint effect on H2O absorption levels. In this view, nBA emerges as the optimal solvent for our specific requirements.

To discern the impact of different alcohol solvents on the defect density of perovskite layers, we performed steady-state photoluminescence (PL) measurements on samples with the configuration of glass/perovskite. As shown in Fig.2f, for EA films, the PL emission peak of the glass side exhibited a blue shift by several nanometers relative to the others. This shift indicates a residual amount of PbI2 at the bottom of the perovskite, owing to the incomplete conversion of PbI2. Notably, the nBA films exhibited the highest PL intensity, surpassing both IPA and nPA films. This enhancement is attributed to the enlarged grain size and effective elimination of PbI2, which in turn reduces the density of grain boundaries and suppresses the non-radiative recombination. In addition, the time-resolved photoluminescence (TRPL) measurements further supported these findings, with the lifetime of each sample recorded at 136.3, 146.6, 350.7 and 142.9ns, respectively (Fig.2g). These results highlight the superior performance of nBA in minimizing non-radiative recombination within the perovskite bulk.

We then performed PL mapping test to investigate the homogeneity of the films, as shown in Fig.2hj. Given the significant amount of PbI2 in EA filmswhich will notably passivate the defects and enhance the PL signal strength (as detailed in Supplementary Fig.11)EA films were excluded from this part of the analysis. The nBA and nPA films demonstrated superior uniformity compared to the IPA films, a trait ascribed to their lower saturated vapor pressure. This characteristic, combined with the solvents extended chain length, leads to slower volatilization, while reduced polarity further restricts water ingress into the films. Both factors contribute to a diminished crystallization rate of perovskite, yielding films with enhanced homogeneity47. However, the slow volatilization rate of the solvent allows the residual solution to continue interacting with the perovskite through dissolution-crystallization reactions. This process tends to produce a non-optically active -phase and creates voids within the bulk48, culminating in a diminished PL mapping signal in nPA films. Overall, the nBA films demonstrated less non-radiative recombination and superior uniformity, making them conducive to the scale-up fabrication of perovskite films.

We fabricated the single-junction perovskite solar cells with an architecture of Glass/ITO/NiO/SAM/1.68 eV-perovskite/C60/SnOx/Cu. The schematic structure is shown in Fig.3a while the detailed photovoltaic parameters of the devices with an active area of 0.049cm2 applying EA, IPA, nBA and nPA are summarized in Supplementary Table3 and Fig.3b. For further comparison, we constructed devices under two distinct conditions: an N2 environment and ambient air, with their respective photovoltaic parameters detailed in Supplementary Fig.12. Devices fabricated in air exhibit smaller VOC compared to those fabricated in N2 glove box, which can be attributed to moisture-induced films deterioration. More notably, air-fabricated devices generally suffered from pronounced efficiency losses, except for those using nBA solvent. This exception highlights nBAs resilience to air exposure during fabrication, with such devices achieving the highest conversion efficiency. In our champion devices, nBA devices displayed distinct advantages in VOC, JSC, and FF with a narrower distribution proving its higher repeatability, as shown in Fig.3b. According to Fig.3c and Supplementary Table4, the improvement of nBA devices in VOC and JSC compared with IPA groups was attributed to the lower non-radiative recombination loss and parasitic absorption caused by PbI2 in the surface and bulk, which was also beneficial to the cells' light stability (as shown in Supplementary Fig.13). The integrated JSC value from the external quantum efficiency (EQE) curve in Fig.3d was calculated to be 20.81 and 20.99mAcm2, respectively, corresponding well with the values obtained from JV measurements. Compared with the IPA devices, the nBA displayed improved charge collection, particularly between 400 and 600nm, due to the larger grain sizes minimizing recombination49. In order to prove the influence of uniformity on the performance of large-area devices, we compared the JV curves of devices with a 1.044cm2 aperture area fabricated by IPA and nBA (Fig.3e and Supplementary Figs.16 and 17), and the specific data are shown in Supplementary Table5. The nBA devices outperformed the IPA counterparts in terms of FF and JSC, attributed to superior uniformity. Additionally, from the EQE spectra of eight cells with a small area of 0.049cm2 (Supplementary Figs.14 and 15), we observed that the nBA devices exhibited a much narrower distribution of the corresponding integrated current. Furthermore, we fabricated PSCs with an area of 1.044cm2, producing 15 devices per type. The histogram of their PCE was displayed in the inset of Fig.3e. Moreover, we compared the photovoltaic parameters of devices fabricated by IPA and nBA in different humidity, and the XRD of films were shown as well (Supplementary Figs.18 and 19), which proved that nBA hinders the effect of moisture during the fabrication of devices.

a Schematic architecture of single junction. b Photovoltaic parameters for IPA and nBA devices. c JV curves of the champion opaque devices (0.049 cm2 aperture area). d EQE spectra of the champion device. e JV curves of the champion opaque devices (1.044cm2 aperture area); PCE distributions of 15 devices for each sample are shown inset. f QFLS values extracted from the PL spectra for neat perovskite, HTL/perovskite and HTL/perovskite/ETL. g EL spectra for IPA and nBA perovskite devices. h VOC evolution as a function of light intensities for the IPA and nBA perovskite devices.

We then carried out photoluminescence quantum yield (PLQY) measurements to quantify the quasi-Fermi level splitting (QFLS) in the neat perovskite layers and the stacks by different layers (Fig.3f)50,51,52. The implied VOC values estimated from the PLQY measurements were in good agreement with the values obtained from the JV results. The above results suggested that replacing IPA with nBA could promote the conversion of PbI2, thereby synergistically mitigating the non-radiative recombination losses both in the bulk and in the interface between hole-transport-layer (HTL) and perovskite. The VOC, indicative of the recombination rate within devices, was assessed through the EQE at short circuit current conditions, effectively modeling the device as a light-emitting diode36. Furthermore, under the injection current of 21mAcm2 (equal to short circuit current Jph), the electroluminescence (EL) efficiencies of IPA and nBA devices were 0.1% and 0.4% (Fig.3g), corresponding to the VOC loss of 0.180 and 0.144V, respectively. This result is almost consistent with the JV results, that is, the IPA and nBA devices showed a VOC of around 1.20V and 1.22V. To further study the carrier recombination behavior, we investigated the dependence of the VOC on the light intensity53, as shown in Fig.3h. The semilogarithmic relationship displayed follows the expression with a slope = nkT/q log10e, where n is the diode quality factor. The IPA and nBA devices exhibited n values of 1.633 and 1.481, respectively, indicating reduced trap-assisted recombination in the nBA device.

Considering the need for thicker perovskite layers when fabricating on textured silicon, relevant characterizations for perovskite films on both glass and textured silicon substrates were conducted, as shown in Supplementary Figs.2026. However, the limited solvent penetration depth of IPA led to a significant amount of unreacted PbI2 in the underlying layer and further degraded the performance of the devices. While complete conversion of the inorganic framework to perovskite is achievable through adjustments in parameters like quenching gas pressure and blade-coating rate54,55, such modifications can detract from film uniformity and device performance, as evidenced in Supplementary Figs.2730. Consequently, parameter tuning was not utilized to fully convert IPA films in tandem devices.

Specifically, the illustration of the tandem device is demonstrated in Fig.4a and a broader area showing the top-view as well as cross-sectional SEM images of the bottom SHJ is seen in Supplementary Fig.31. The performance of SHJ cell with and without semitransparent perovskite as a filter is shown in Supplementary Fig.32 and Supplementary Table6. It can be clearly seen from Fig.4b that the textured surface with pyramid sizes of 23m was well-covered by the conformally coated perovskite films as well as other functional layers. The corresponding device performance is depicted in Fig.4c and d; a tandem solar cell with an active area of 1.044cm2 achieved a champion PCE of 29.4% (VOC=1.83V, JSC=20.45mAcm2 and FF=78.63%) under reverse scan and the stabilized PCE was observed to be 28.8%. Moreover, an independently certified efficiency of 28.7% was tested from Fraunhofer ISE (shown in Supplementary Fig.33).

a Schematic diagram of perovskite/SHJ tandem solar cell. b Cross-sectional SEM images of perovskite/SHJ (average pyramid size is 23m) tandem for nBA devices. c JV curves of the tandem device (1.044cm2 aperture area); the digital photo of a device is shown in the inset. d MPP tracking of the tandems; PCE distributions of 16 individual tandem devices for each type is shown in the inset. e EQE spectra of a current-matched fully textured monolithic perovskite/SHJ tandem cell. f JV curves of the tandem device (16 cm2 aperture area); the digital photo of a device is shown in the inset.

As shown in Fig.4d, the integrated JSC value of the front and back subcell from EQE spectra (Fig.4e) was 20.62 and 20.51mAcm2, respectively, which was in good agreement with the JSC value determined from the JV measurements considering the loss caused by Ag grid. We further evaluated the operation stability of encapsulated tandem solar cells by measuring the maximum power output under 1-sun-equivalent illumination in ambient air with a relative humidity of 3050%. The encapsulated device retained 96.8% of its initial PCE after 780h of maximum power point (MPP) tracking (Supplementary Fig.34).

To validate the applicability of our approach for scalable fabrication, we applied blade-coating to produce perovskite films on a 36cm glass substrate. Subsequent steady PL and XRD tests conducted on samples from different regions of perovskite films (Supplementary Figs.35 and 36) demonstrated superior uniformity in nBA films compared to IPA films. Furthermore, we fabricated 36 cm2 perovskite/silicon tandem cells (aperture area, 16 cm2) and achieved a conversion efficiency of 26.3% (VOC=1.815V, JSC=18.54mAcm2, FF=78.31%), which is among the highest PCE of large-area perovskite/silicon tandem cells11. The consistency of the EQE spectra at different regions suggested that the film exhibited excellent uniformity (Supplementary Fig.37).

For the further development of perovskite/silicon tandem solar cells, scaling up the size of the perovskite films to M6 (166mm*166mm) becomes essential, a goal that proves challenging with blade-coating due to issues with film uniformity. Therefore, slot-die coating, an expandable technology that allows continuous liquid injection, emerges as the preferable future method49,56. Critical to this method is the complete conversion of the inorganic framework into perovskite films, achievable through careful adjustment of precursor solution concentration, the injection rate of the precursor solution, the rate of slot-die, gap distances between the blade and substrate, and quenching gas pressure. Digital photographs of the perovskite films fabricated under these conditions are depicted in Supplementary Figs.3840. With the optimum slot-die coating parameters (1mL/min, 100 m and 30 PSI), we achieved perovskite films with excellent homogeneity (Supplementary Fig.41). The optimal device delivered a PCE of 25.9% for 16cm2 (VOC=1.823V, JSC=18.50mAcm2, FF=76.63%), as shown in Supplementary Fig.42. These results are anticipated to surpass the efficiency of devices fabricated by the blade-coating in the future.

Read the original:

Solvent engineering for scalable fabrication of perovskite/silicon tandem solar cells in air - Nature.com

US to test hypersonic missile tracking with space-based sensors – Interesting Engineering

The Unites States Missile Defense Agency (MDA) has stated it plans to test its space-based satellites equipped with hypersonic missile tracking sensors. The Hypersonic and Ballistic Tracking Space Sensor (HBTSS) was deployed into orbit in February of this year.

HBTSS is designed to enable the MDA to get an early warning of potential hypersonic missiles. Presently, ground-based systems, while sophisticated, are limited by the curvature of the Earth and the nature of hypersonic missile flight paths.

To this end, sensors located in orbit will have an unobstructed view, enabling more accurate and timely interception. MDA serves as the Defense Departments executive agent for hypersonic defense.

It is racing to stay ahead of threats from Russian and Chinese development efforts. Tracking hypersonic missiles from space is necessary to allow interceptors more time to lock on.

Air Force Lt. Gen. Heath Collins explained at a discussion at the Center for Strategic and International Studies on June 6th that the major challenge with hypersonic missiles is that they re-enter the atmosphere before ballistic missiles are detected, leaving a very small window for interception due to their high speed.

And so instead of being down, looking up to find a hypersonic, you want to be high, looking down to track hypersonic, he continued.

Thats what Hypersonic and Ballistic Tracking Space Sensor is all about.

Although no specific date for the first test has been announced, Lt. Gen. Collins stated that it would occur in approximately a week. The test will involve a dummy target traveling at hypersonic speeds within the satellites field of view.

He said that the test would assess the sensitivity, timeliness, and accuracy of the two systems to meet the demonstration objectives for HBTSS and potentially inform changes or confirm the systems effectiveness, ultimately contributing to the Space Development Agencys future plans.

A second test is also planned for later in the year following the initial test.

The tests are key steps in our ability to prove out that we can close a hypersonic fire control loop from space, Lt. Gen. Collins said.

We are in lockstep working this with the Space Development Agency, and they are alreadyplanning HBTSS-like sensorsin their future tranches of the Proliferated Warfighter Space Architecture to start filling out that truly global hypersonic kill chain.

However, detection and tracking are only part of the solution. The ability to physically intercept and destroy hypersonic missiles is equally important.

Collins explained that the MDA is focused on finding alternative near-term capabilities for the interceptors as it aims to develop the GPI as quickly as possible. Congress is urging the agency to expedite the new interceptors field readiness.

NEWSLETTER

Stay up-to-date on engineering, tech, space, and science news with The Blueprint.

Christopher McFadden Christopher graduated from Cardiff University in 2004 with a Masters Degree in Geology. Since then, he has worked exclusively within the Built Environment, Occupational Health and Safety and Environmental Consultancy industries. He is a qualified and accredited Energy Consultant, Green Deal Assessor and Practitioner member of IEMA. Chris’s main interests range from Science and Engineering, Military and Ancient History to Politics and Philosophy.

Follow this link:

US to test hypersonic missile tracking with space-based sensors - Interesting Engineering

Remembering ‘Doc’ Helms: beloved mentor and pioneer in architectural lighting – University of Colorado Boulder

Professor Ron Helms (right) in CU Boulder'sphoto- metric lab in 1967.

In 1973, Illumination 1, which teaches the fundamentals of illuminating engineering, became a required course for all CU Boulder architectural engineering students and has been taught ever since.

Doc nurtured and inspired so many engineers, designers, educators and manufacturing professionals, said Cheryl English, one of Helms students from 1977-81 and a retired lighting executive. We remember Doc for his keen sense of humor and dedication to his students, whom he referred to as his family.

Helms received his undergraduate and masters degrees from the University of Illinois and PhD from Ohio State University. He left CU Boulder in 1981 to become the head of University of Kansas architectural engineering program, where he established the Bob Foley Illumination laboratory in 1985. He then went on to North Carolina Agricultural and Technical State University to establish a new illuminating engineering program; he retired in 2006.

Helms published many papers, authored three textbooks and presented technical programs.

"He wasa strong advocate for the recognition of architectual engineering and illuminating engineering ascredible disciplines in the industry,"English said. "Heworked to establish that recognition in professional organizations and engineeringaccreditations."

Helms leaves behind his four children, eight grandchildren, extended family and "countless lighting professionals who had the privilege of benefiting from his mentorship or attending his education programs, English said.

A memorial service was held on June 1 at Westminster Presbyterian Church in Greensboro, North Carolina.

Read the rest here:

Remembering 'Doc' Helms: beloved mentor and pioneer in architectural lighting - University of Colorado Boulder

Super impressive wall of wind turbines yield 2,200 kWh of quiet energy – Interesting Engineering

A wind fence developed by New York-based designer Joe Doucet is set to bring clean energy production into urban landscapes. The fence consists of vertical wind turbines, is modular, and, most importantly, is pleasing to the eye, making it more likely to be adopted in hotels, corporate buildings, and residential units.

Wind energy is an important component of the renewable energy mix that countries have adopted as they aim for a future away from fossil fuels. To achieve maximum energy gain and efficiency, original equipment manufacturers (OEMs) build bigger turbines every year for large installations.

This has been preventing wind energy from participating in distributed energy generation, much like solar panels can be installed on rooftops, in gardens, and now even on balconies.

In 2021, Doucet was researching distributed energy products for wind energy and found that few good options were available. So, the designer did what he could best: design a new product that was both efficient and scored on aesthetics.

Doucets original design was called the Wind Turbine Wall. Over the last two years, the designer has developed and tested the concept several times, with the majority of changes affecting the shape and size of the blades.

With the team at Airiva, a company he co-founded with energy industry veteran Jeff Stone, Doucet put 16 designs of vertical turbine blades through the motions to arrive at three final versions that made it to the wind tunnel testing.

After rigorous testing at two facilities in the US, the team was convinced that the helical structure of the turbine blades was the most efficient. This isnt the first time someone is working with helically shaped blades in a vertical turbine. However, where Airiva claims to have made good progress is how to get maximum benefit by placing multiple blades operating simultaneously.

In a standard setup, where eight helical blades are precisely arranged, the Wind Fence generates about 2,200 kilowatts of energy annually.

From the output from a single unit, an average US household would need five Wind Fence units to remove its dependence on the grid completely. This might not sound too much until you realize that each unit measures nearly 14 feet (4.2 m) by seven feet (2.1 m).

Airivas team isnt seeking residential customers to buy their product. Since the concept was first unveiled in 2021, corporations, public institutions, and real-estate firms have been keen to install it on their premises.

The noiseless movement of the wind turbines makes clean energy generation aesthetically pleasing and helps make a statement about the companys transition to a greener planet. Airiva plans to use 80 percent recycled material in its production.

The solution is modular, and one can install an array of units to increase energy production at a facility. Even then, the Wind Fence wouldnt match the energy output of a massive horizontal turbine. But that isnt a target Airiva is trying to beat either.

The advantage of distributed energy systems is that they see fewer energy losses during transmissions since they are generated so close to the site of usage.

The company is still some distance away from installing its units commercially, though. Custom pilots could happen later this year, and the first orders will come in 2025.

NEWSLETTER

Stay up-to-date on engineering, tech, space, and science news with The Blueprint.

Ameya Paleja Ameya is a science writer based in Hyderabad, India. A Molecular Biologist at heart, he traded the micropipette to write about science during the pandemic and does not want to go back. He likes to write about genetics, microbes, technology, and public policy.

More:

Super impressive wall of wind turbines yield 2,200 kWh of quiet energy - Interesting Engineering

Automating Prompt Engineering with DSPy and Haystack | by Maria Mestre | Jun, 2024 – Towards Data Science

Teach your LLM how to talk through examples Photo by Markus Winkler on Unsplash

One of the most frustrating parts of building gen-AI applications is the manual process of optimising prompts. In a publication made by LinkedIn earlier this year, they described what they learned after deploying an agentic RAG application. One of the main challenges was obtaining consistent quality. They spent 4 months tweaking various parts of the application, including prompts, to mitigate issues such as hallucination.

DSPy is an open-source library that tries to parameterise prompts so that it becomes an optimisation problem. The original paper calls prompt engineering brittle and unscalable and compares it to hand-tuning the weights for a classifier.

Haystack is an open-source library to build LLM applications, including RAG pipelines. It is platform-agnostic and offers a large number of integrations with different LLM providers, search databases and more. It also has its own evaluation metrics.

In this article, we will briefly go over the internals of DSPy, and show how it can be used to teach an LLM to prefer more concise answers when answering questions over an academic medical dataset.

This article from TDS provides a great in-depth exploration of DSPy. We will be summarising and using some of their examples.

In order to build a LLM application that can be optimised, DSPy offers two main abstractions: signatures and modules. A signature is a way to define the input and output of a system that interacts with LLMs. The signature is translated internally into a prompt by DSPy.

When using the DSPy Predict module (more on this later), this signature is turned into the following prompt:

Then, DSPy also has modules which define the predictors that have parameters that can be optimised, such as the selection of few-shot examples. The simplest module is dspy.Predict which does not modify the signature. Later in this article we will use the module dspy.ChainOfThought which asks the LLM to provide reasoning.

Things start to get interesting once we try to optimise a module (or as DSPy calls it compiling a module). When optimising a module, you typically need to specify 3 things:

When using the dspy.Predict or the dspy.ChainOfThought modules, DSPy searches through the training set and selects the best examples to add to the prompt as few-shot examples. In the case of RAG, it can also include the context that was used to get the final response. It calls these examples demonstrations.

You also need to specify the type of optimiser you want to use to search through the parameter space. In this article, we use the BootstrapFewShot optimiser. How does this algorithm work internally? It is actually very simple and the paper provides some simplified pseudo-code:

The search algorithm goes through every training input in the trainset , gets a prediction and then checks whether it passes the metric by looking at self.metric(example, prediction, predicted_traces). If the metric passes, then the examples are added to the demonstrations of the compiled program.

The entire code can be found in this cookbook with associated colab, so we will only go through some of the most important steps here. For the example, we use a dataset derived from the PubMedQA dataset (both under the MIT license). It has questions based on abstracts of medical research papers and their associated answers. Some of the answers provided can be quite long, so we will be using DSPy to teach the LLM to prefer more concise answers, while keeping the accuracy of the final answer high.

After adding the first 1000 examples to an in-memory document store (which can be replaced by any number of retrievers), we can now build our RAG pipeline:

Lets try it out!

The answer to the above question:

Ketamine inhibits the proliferation of rat neural stem cells in a dose-dependent manner at concentrations of 200, 500, 800, and 1000M. Additionally, ketamine decreases intracellular Ca(2+) concentration, suppresses protein kinase C- (PKC) activation, and phosphorylation of extracellular signal-regulated kinases 1/2 (ERK1/2) in rat neural stem cells. These effects do not seem to be mediated through caspase-3-dependent apoptosis.

We can see how the answers tend to be very detailed and long.

We start by creating a DSPy signature of the input and output fields:

As we can see, we already specify in our description that we are expecting a short answer.

Then, we create a DSPy module that will be later compiled:

We are using the Haystack retriever previously defined to search the documents in the document store results = retriever.run(query=question). The prediction step is done with the DSPy module dspy.ChainOfThought which teaches the LM to think step-by-step before committing to the response.

During compilation, the prompt that will be optimised to look like this:

Finally, we have to define the metrics that we would like to optimise. The evaluator will have two parts:

Our evaluation dataset is composed of 20 training examples and 50 examples in the devset.

If we evaluate the current naive RAG pipeline with the code below, we get an average score of 0.49.

Looking at some examples can give us some intuition on what the score is doing:

Question: Is increased time from neoadjuvant chemoradiation to surgery associated with higher pathologic complete response rates in esophageal cancer?

Predicted answer: Yes, increased time from neoadjuvant chemoradiation to surgery is associated with higher pathologic complete response rates in esophageal cancer.

Score: 0.78

But

Question: Is epileptic focus localization based on resting state interictal MEG recordings feasible irrespective of the presence or absence of spikes?

Predicted answer: Yes.

Score: 0.089

As we can see from the examples, if the answer is too short, it gets a low score because its similarity with the ground truth answer drops.

We then compile the RAG pipeline with DSPy:

After we do this and re-evaluate the compiled pipeline, the score is now 0.69!

Now its time to get the final optimised prompt and add it into our Haystack pipeline.

We can see the few-shot examples selected by DSPy by looking at the demos field in the compiled_rag object:

There are 2 types of examples provided in the final prompt: few-shot examples and bootstrapped demos, like in the prompt shown above. The few-shot examples are question-answer pairs:

Whereas the bootstrapped demo has the full trace of the LLM, including the context and reasoning provided (in the rationale field below):

All we need to do now is extract these examples found by DSPy and insert them in our Haystack pipeline:

Our new pipeline becomes:

Lets check the same question we tried before.

Question:

What effects does ketamine have on rat neural stem cells?

Before:

Ketamine inhibits the proliferation of rat neural stem cells in a dose-dependent manner at concentrations of 200, 500, 800, and 1000M. Additionally, ketamine decreases intracellular Ca(2+) concentration, suppresses protein kinase C- (PKC) activation, and phosphorylation of extracellular signal-regulated kinases 1/2 (ERK1/2) in rat neural stem cells. These effects do not seem to be mediated through caspase-3-dependent apoptosis.

After:

Ketamine at higher concentrations inhibits the proliferation of rat neural stem cells, while not affecting apoptosis. Additionally, it decreases intracellular calcium concentration and suppresses PKC activation and ERK1/2 phosphorylation in these cells.

Job done!

In this post, we have used DSPy to optimise the prompt used in a Haystack RAG pipeline. We have done so by using a custom metric based on Haystacks evaluation framework that penalised the LLM for long answers while keeping the similarity with the correct answer high. With this approach, we have managed to improve our performance by almost 40% without having to do any manual prompt engineering.

More:

Automating Prompt Engineering with DSPy and Haystack | by Maria Mestre | Jun, 2024 - Towards Data Science