AI in Africa: Basics Over Buzz.

When Buti Manamela visited Lengau, Africa's fastest supercomputer, he had more prosaic technology in mind: electricity. South Africa's Deputy Minister of Higher Education, Science and Technology was at the Center for High Performance Computing in Cape Town for what should have been a showcase tour of a facility providing the country with the computing power needed to run and analyze the kinds of complex models and huge datasets that underpin artificial intelligence (AI) and machine learning (ML). But Manamela was there to better understand the impact of South Africa's rolling power blackouts on the center's operations. Lengau, which means "cheetah" in Setswana, is one of the most important outposts in Africa's AI infrastructure landscape; yet, it is struggling to operate at full capacity because of unreliable power.

African energy transitions should be driven from the ground up.

At the UN climate summit (COP28) kicking off in Dubai next week, we can expect the nations of the world to issue a flurry of energy- and climate-related announcements, pledges, and plans. Like their global peers, African governments will be using COP28 to demonstrate their climate ambition, building on commitments they made at the inaugural African Climate Summit held three months prior in Nairobi.

Electrical percolation in quasi-two-dimensional metal nanowire networks for transparent conductors.

We simulate the conductivity of quasi-two-dimensional mono- and polydisperse rod networks having rods of various aspect ratios (L/D = 25-800) and rod densities up to 100 times the critical density and assuming contact-resistance dominated transport. We report the rod-size dependence of the percolation threshold and the density dependence of the conductivity exponent over the entire L/D range studied. Our findings clarify the range of applicability for the popular widthless-stick description for physical networks of rodlike objects with modest aspect ratios and confirm predictions for the high-density dependence of the conductivity exponent obtained from modest-density systems. We also propose a heuristic extension to the finite-width excluded area percolation model to account for arbitrary distributions in rod length and validate this solution with numerical results from our simulations. These results are relevant to nanowire films that are among the most promising candidates for high performance flexible transparent electrodes.

Integrating simulations and experiments to predict sheet resistance and optical transmittance in nanowire films for transparent conductors.

Metal nanowire films are among the most promising alternatives for next-generation flexible, solution-processed transparent conductors. Breakthroughs in nanowire synthesis and processing have reported low sheet resistance (Rs ≤ 100 Ω/sq) and high optical transparency (%T > 90%). Comparing the merits of the various nanowires and fabrication methods is inexact, because Rs and %T depend on a variety of independent parameters including nanowire length, nanowire diameter, areal density of the nanowires and contact resistance between nanowires. In an effort to account for these fundamental parameters of nanowire thin films, this paper integrates simulations and experimental results to build a quantitatively predictive model. First, by fitting the results from simulations of quasi-2D rod networks to experimental data from well-defined nanowire films, we obtain an effective average contact resistance, which is indicative of the nanowire chemistry and processing methods. Second, this effective contact resistance is used to simulate how the sheet resistance depends on the aspect ratio (L/D) and areal density of monodisperse rods, as well as the effect of mixtures of short and long nanowires on the sheet resistance. Third, by combining our simulations of sheet resistance and an empirical diameter-dependent expression for the optical transmittance, we produced a fully calculated plot of optical transmittance versus sheet resistance. Our predictions for silver nanowires are validated by experimental results for silver nanowire films, where nanowires of L/D > 400 are required for high performance transparent conductors. In contrast to a widely used approach that employs a single percolative figure of merit, our method integrates simulation and experimental results to enable researchers to independently explore the importance of contact resistance between nanowires, as well as nanowire area fraction and arbitrary distributions in nanowire sizes. To become competitive, metal nanowire systems require a predictive tool to accelerate their design and adoption for specific applications.