Load/Path

Work began on 12 October on a 290 km line between Porto and Lisbon, designed for trains at up to 300 km/h. The first 71 km, from Porto to Oiã, includes eight tunnels and 15 viaducts. One is a new road and rail bridge over the River Douro. A station will sit about 60 m under the centre of Vila Nova de Gaia, with 405 m platforms.

The line uses Portugal's wide 1,668 mm track gauge, so trains can also run on the existing network. The first section costs €2.65bn and should open in 2031. The rest should follow by 2032. Porto to Lisbon should then take 1 h 20 min instead of 2 h 48 min.

Mixing waste plastic into asphalt is popular, but it raises a worry: will the road shed microplastics? Penn State engineers mixed recycled low-density polyethylene, the plastic in carrier bags and film, into asphalt. They ran it through two standard lab tests. One rolls a loaded steel wheel over it to make ruts. The other pumps water through it to imitate rain and traffic.

The plastic mix usually did better than ordinary asphalt. It also released fewer particles that looked like the plastic. But the team tested only one mix, in the lab, not on a real road. The paper is in the Journal of Cleaner Production.

In most reactors, the fuel is solid pellets inside metal tubes. In the Molten Chloride Reactor Experiment, the uranium fuel is dissolved in a hot liquid salt. On 14 October, Southern Company and Core Power said the US Department of Energy had approved cost-shared funding for it. The sum wasn't disclosed.

It is a tiny reactor, 150 kW of heat, built in an old test cell at Idaho National Laboratory. It needs about 75 batches of fuel salt to reach criticality, the point where the chain reaction sustains itself. That could happen as early as 2027. It would be the first "fast" molten chloride reactor to do so. Fast reactors don't slow their neutrons down, which lets them burn more of their fuel.

Most of France's 699 AGC regional trains burn diesel for at least part of their journeys. SNCF and Alstom have taken three of them, removed the diesel engines and fitted four battery packs instead. The packs come from Swiss firm Leclanché. Cooling units on the roof keep the cells near 15 °C. By the end of October, all three will carry passengers in southern France.

They run about 80 km on battery. They recharge in about 20 minutes, even while moving, from overhead wires. The old fuel tanks stay on board, so the train weighs the same as before. The trial has cost €41m so far and will last two years.

Every transformer has a core made of grain-oriented electrical steel. It is iron mixed with silicon and rolled so its crystals line up in one direction. That cuts the energy lost each time the magnetic field flips, 50 times a second. Europe used 32% more of it in 2025 than in 2021. Over the same period, European mills' share of that market fell from 66% to 43%. Only two EU firms make it.

The European Commission has set temporary minimum prices for imported steel and transformer cores until February 2027. Steelmakers welcomed this. Transformer makers say the best grades still come from Japan and South Korea, so their costs will rise just as grids need more transformers.

US firm Sceye flew a solar-powered, uncrewed airship from New Mexico to Japan and back: nearly 30,000 km in 30 days. It flew in August and September. IEEE Spectrum reported the results this week. Over Shikoku island it held its position and worked as a 4G base station linked to SoftBank's network. People on the ground made calls, sent texts and streamed video through it.

The airship also relayed video from a drone to imitate surveying an earthquake zone. A server on board processed data on the spot. That cut delays by more than 40% compared with sending data to the cloud, SoftBank says.

Electric motors need strong magnets made with rare earths such as neodymium and dysprosium. China controls about 90% of the market for refined rare earths. A mineral called monazite could help other countries: it can be 45–70% rare earths by weight, and it turns up in beach and river sands.

The catch is thorium. Monazite usually contains this mildly radioactive element. Processing it leaves residues that must be stored safely, perhaps forever. A plant making 5,000 tonnes of rare-earth oxides a year from ore with 7% thorium would produce about 700 tonnes of thorium a year. Projects in Utah, Canada and Australia are trying to make the numbers work.

Perovskite solar cells are made by printing a liquid "ink" and letting it dry into crystals. The usual solvents, such as dimethylformamide, are toxic. That is a problem in a factory. University of Surrey researchers replaced them with ethanol and 2-methyltetrahydrofuran, a solvent made from farm and forestry waste.

Their small modules, with 11.5 cm² of active area, converted more than 18.2% of sunlight into electricity. The recipe also skips a fiddly step where a second solvent is dripped on to trigger crystals. The work is in Nano Energy.

On 11 October, NASA's Perseverance rover passed 45.22 km of driving on Mars. That beats Opportunity's 45.16 km, built up over more than 14 years. Perseverance did it in under six.

The difference is autonomy. Perseverance plans its own route about 90% of the time, at up to 120 m an hour. Opportunity drove itself about 5% of the time. People at NASA's Jet Propulsion Laboratory still pick the long-range goals. They also flag ground the software can't judge, such as soft sand.

NASA has set out how it wants to start building a Moon base. It plans about 70 robotic landings over a decade, 45 of them by 2032. First it wants reliable landings, with seven in 2027 and seven in 2028. Then power and communications kit. Then longer astronaut stays. Each landing should eventually carry up to 8,000 kg.

The record so far is shaky. Firefly's Blue Ghost in 2025 is the only fully successful US Moon landing since Apollo. Two Intuitive Machines landers tipped over. Firefly has just opened a cleanroom big enough to build 12 landers at once.

In a war, GPS signals are often the first thing to be jammed. Engineers at the US Naval Air Warfare Center have flown a drone that landed on a ship moving at sea with no GPS and no pilot. The test was reported this week.

The drone's cameras recognise the ship's shape. Its motion sensors and software predict how the deck will rise and fall, so it touches down at the right moment. Normally this needs radar systems costing millions of dollars and changes to the ship. The software runs on ordinary drone computers. The Navy hasn't published landing accuracy or how many flights it made.

Forecasts for two to six weeks ahead are so poor that forecasters call it the "predictability desert". A team from Microsoft, MIT, Harvard and others added a machine-learning layer on top of an existing forecast model. It learns each model's past mistakes and corrects for them.

Paired with the European forecast centre's model, it roughly doubled accuracy at that range. It also improved forecasts of extreme weather, where some correction methods make things worse. The work is in Nature, and the code is open source.

At the Paris Motor Show on 12 October, Peugeot revealed the E-T16, a concept based on its small 208. It has three electric motors, two at the back and one at the front, making 600 kW together. Peugeot claims 0–100 km/h in 2.5 seconds. Software shares the torque between the wheels moment by moment, which is called torque vectoring.

The 85 kWh battery runs at 800 volts and cools its cells through their sides, not the floor. Peugeot claims a 10–80% charge in under 15 minutes. Steering is by wire, with no mechanical link to the wheels. It is a concept: Peugeot says some of its features will reach the next E-208, not which ones.

F1 only switched to new engines this year. Half their power comes from the combustion engine and half from the battery. Now FIA president Mohammed Ben Sulayem says V8 engines will return, "likely in 2030". He said it in Kigali, at the launch of Rwanda's planned African Grand Prix.

His goals: cars about 80 kg lighter, at least 50% less spent on engine development, and a better sound. The FIA hasn't written the rules. That includes how much electric help the V8 will get.

CMR Surgical, based in Cambridge, makes Versius, a robot for keyhole surgery on soft tissue. Its arms are separate units that can be wheeled between operating theatres. On 12 October, US medical giant Stryker agreed to invest about $400m (£300m). It will sell Versius in the US and has an option to buy the whole company.

In the US, Versius is currently cleared only for gallbladder removal. CMR plans to apply for more procedures. Analysts expect it to start in smaller day-surgery centres rather than taking on Intuitive's da Vinci robot in big hospitals.

Cornell researchers have built a chip that makes tiny bubbles exactly where and when they want. A grid of nine electrodes splits water into hydrogen gas. The longer the voltage stays on, the bigger the bubble, from 10 to 100 micrometres. Ultrasound then makes each bubble vibrate, which pushes the liquid around it.

Arranging bubbles in rows or an L-shape creates different flow patterns. The team hopes to use this to move microscopic robots, and perhaps one day to deliver drugs to a precise spot. For now it is a lab demonstration on one chip, with no testing in living tissue. The work is in Science Advances.