Industries / Mining and materials / Chemicals and petrochemicals / Industrial gases
Liquefaction and storage
Liquefaction is the work of cooling a gas until it turns to liquid, which takes up several hundred times less space, and storage is keeping that liquid cold in insulated vessels until it is used.
5
Milestones achieved, from 1892 to 1966.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1892
The vacuum flask
James Dewar at the Royal Institution in London sealed one glass flask inside another with a partial vacuum between them. The vacuum slows the heat getting in, so a very cold liquid inside boils away slowly. He first showed it on Christmas Day 1892 and never patented it; the Thermos company patented and renamed it for wider use in 1904.
Credit
James Dewar, Royal Institution
1895
Air liquefied by machine
Carl von Linde made liquid air in quantity on 29 May 1895 and registered his patent on 5 June. His machine compresses air, lets it expand through a valve so that it cools, and sends the cold expanded air back through a long coiled heat exchanger to chill the air coming in, until part of the air turns to liquid. His US patent, filed on 9 July 1895 and granted in 1903, also describes boiling the liquid off in stages, nitrogen first, to leave a liquid richer in oxygen, the start of air separation. In England William Hampson had registered a preliminary patent for a similar liquefier on 23 May 1895, and who came first was disputed.
Credit
Carl von Linde; William Hampson
1908
Helium turns liquid
On 10 July 1908 Heike Kamerlingh Onnes at Leiden University cooled helium to about −269 °C, a few degrees above absolute zero, and collected about a teacup of liquid. Leiden was the only place with liquid helium until 1923, and the cold it gave let Kamerlingh Onnes discover superconductivity, the loss of all electrical resistance in some metals, in 1911.
Credit
Heike Kamerlingh Onnes and his Leiden laboratory
1946
A liquefier labs could buy
Samuel Collins and colleagues at the Massachusetts Institute of Technology built the first Collins helium cryostat in 1946, the first helium liquefier to run without outside refrigeration; earlier machines needed liquid hydrogen and liquid air made separately to pre-cool the helium. Arthur D. Little, Inc. made it in quantity, and the first model produced about 2 liters an hour. A 1968 National Bureau of Standards book reports that more than 350 had been sold, and that nearly all of the 700,000 liters of liquid helium made in the US in 1958 came from Collins machines.
LessonCollins hung the heat exchangers in helium gas so that small leaks no longer mattered, and fitted the expansion engine with close-fitting nitrided steel parts; designing around the weak points of earlier machines, poor vacuum seals and worn engines, is what made the liquefier practical.
Credit
Samuel C. Collins (MIT); Arthur D. Little, Inc
1966
Helium shipped as liquid
By the 1960s helium producers found that the cheapest way to deliver helium was to liquefy it at the plant and ship it as a liquid, as had long been done for oxygen and nitrogen. The US made about 1.7 million liters of liquid helium in 1964, and by 1966 its liquefaction capacity passed 12 million liters a year. A helium plant near Otis, Kansas, that opened in 1966 was built to liquefy its whole output and ship it in semitrailers of about 10,000 gallons (about 38,000 liters), insulated by many thin reflective layers in a vacuum.
Credit
US helium producers, among them the Kansas Refined Helium Company; figures from the National Bureau of Standards (now NIST)
Watch for
Liquid air as grid storage
Highview Power is building a plant at Carrington, near Manchester, England, that stores electricity by using it to liquefy air, keeps the liquid in insulated low-pressure tanks, and later warms it so the expanding gas drives turbines. Highview says the plant will deliver 50 MW for up to six hours, about 300 MWh, and calls it the world's largest commercial plant of its kind. In March 2026 Highview wrote that it would be operating in late 2026, and UK Research and Innovation, which funded Highview's early work, also gives 2026. On 17 November 2026 Elexon's public register of units in Britain's balancing mechanism, the market the grid operator uses to buy and sell power at short notice, listed no unit for Highview Power.
Achieved whenthe plant is in commercial service and someone other than Highview reports its measured output and round-trip efficiency, the share of the electricity put in that comes back out.
Claimed by
Highview Power, March 2026
Open
01
A public record of boil-off
An open record of how fast liquid nitrogen actually boils away in the small vacuum-insulated vessels the industry delivers and refills, such as lab dewars and portable liquid cylinders of 160 to 250 liters. Each vessel is weighed automatically for months, and its measured daily loss is set beside the maker's rated loss, with its model, age and a thermal photo. A failing vacuum shows up as a higher loss, and often as a cold patch on the outer wall.
The briefCredit · unclaimed02
Measure liquid air storage
An open script that works out, month by month, how much of the electricity the Carrington liquid air plant draws to liquefy air comes back out as electricity, using public meter data from Britain's electricity market.
The briefCredit · unclaimed
Open 01 · the brief
A public record of boil-off
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open record of how fast liquid nitrogen actually boils away in the small vacuum-insulated vessels the industry delivers and refills, such as lab dewars and portable liquid cylinders of 160 to 250 liters. Each vessel is weighed automatically for months, and its measured daily loss is set beside the maker's rated loss, with its model, age and a thermal photo. A failing vacuum shows up as a higher loss, and often as a cold patch on the outer wall.
Why it is still open
A vessel's vacuum is usually judged one vessel at a time, by a boil-off test against the design value. NASA used thermal cameras at Kennedy Space Center to find a likely missing patch of insulation in one of its 1965 liquid hydrogen tanks and a cold region on a liquid oxygen tank, but noted that wind, temperature, humidity and cloud reflections skew such images, and released no data set. Searches in November 2026 found boil-off models for liquefied natural gas and liquid hydrogen, and patents for camera alarms on containers, but no public field measurements for small nitrogen or argon vessels.
Built with what exists
A platform scale or load cells under each vessel, read by an HX711 board and an ESP32 microcontroller that logs the weight every ten minutes; a phone thermal camera such as the FLIR One; Python to compute daily loss; a small vision model fine-tuned on one graphics card to sort thermal photos (see the Note "One GPU to fine-tune"); Zenodo to publish the data with a DOI.
First step
Put one 50-liter lab dewar of liquid nitrogen, already in routine use, on a logging scale for two weeks and compare its daily loss with the figure on the maker's datasheet.
Done looks like
At least 30 vessels logged for 60 days each, with measured loss per day as a percentage of contents compared with the maker's rating, published as open data; and a test of whether thermal photos alone pick out the vessels losing more than one and a half times their rating.
Open 02 · the brief
Measure liquid air storage
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open script that works out, month by month, how much of the electricity the Carrington liquid air plant draws to liquefy air comes back out as electricity, using public meter data from Britain's electricity market.
Why it is still open
The plant was not yet running in November 2026, and Highview's own March 2026 account of it gives no efficiency figure. Published efficiencies for liquid air storage come from models and design studies, not from the metered output of a grid-scale plant.
Built with what exists
Elexon's free Insights API, which lists each unit in the balancing mechanism and publishes its metered half-hourly volumes; Python and pandas; a public notebook. The job depends on the plant registering as a unit; if it does not, the data has to be requested from Highview or the regulator, Ofgem.
First step
Write and test the script this month on a battery storage unit already in Elexon's register, then watch the register for a new unit named for Highview or Carrington.
Done looks like
Twelve months of monthly round-trip efficiency, with energy in and energy out in megawatt-hours, and the code published so that anyone can rerun it.