The hardest thing about Biosphere 2 was not the rainforest or the ocean. It was keeping the air in. Every rigid building breathes with the weather: as the air inside warms it expands, as the barometer outside falls the air inside pushes out, as humidity rises there is simply more gas in the room. William Dempster, the project’s director of systems engineering, calculated that if the structure had been sealed rigid, the pressure difference across its glass would have swung by about 5,000 pascals in a day, roughly a hundred pounds on every square foot. The frame would have burst or buckled, and long before that the breathing would have pumped air through every flaw in it. A sealed building that does not solve this problem is not sealed; it is a bellows.

The rehearsal

The solution was worked out in a smaller box first. The Test Module, built in 1985 on the Oracle site, enclosed about 480 cubic metres, a steel liner below ground and glass above, and was the largest closed ecological facility in the world until the big one went up beside it. It ran closures from the last day of 1986.

John Allen shut himself in it for seventy-two hours in 1988, and in November 1989 the botanist Linda Leigh spent three weeks inside, after a week in which the module was sealed with nothing living in it but plants, to check that nothing toxic built up. More than sixty person-days were logged in all. The module was the first facility to purify its air by biological means rather than catalytic burners, to cycle its water, and to recycle human waste through a planted marsh. Its leak was measured at about 24 percent of its air a year. That was a triumph for 1989 and far too much for what came next.

The envelope

The building that followed sits on a footprint of 1.27 hectares and holds about 180,000 cubic metres of air, 30,000 tonnes of soil and four million litres of seawater. Above ground it is laminated glass on a steel space frame: 16,000 square metres of panes, some 80 kilometres of caulked glazing joints.

Below ground it is a continuous sheet of stainless steel three millimetres thick, welded along 16 kilometres of seams, running under the soil and up the basement walls to meet the glass. Ordinary stainless corrodes in wet soil and seawater, so the liner is a high-nickel alloy chosen for toughness; a system of leak-detection channels was laid beneath it so that flawed welds could be found and repaired before closure. Every pipe and cable enters through that liner, welded airtight, because the designers had learned that a stranded electrical wire cannot be sealed at its jacket alone: air travels between the strands. There are five airtight doors, two of them inside vestibules that work as airlocks.

The lungs

The breathing problem was solved with two domes the crew called the lungs, connected to the main structure by underground air ducts. Each is a cylindrical tank with a flexible, impermeable membrane for a lid, weighted so that it rests on the air beneath it. As the air inside Biosphere 2 expands the membranes rise; as it contracts they fall. Together they can take up about 43,000 cubic metres of expansion against a fixed air volume of about 140,000, about 30 percent.

The membranes’ own weight would have kept the interior at a slight positive pressure, about 150 pascals, which would still drive air out through every pinhole, so each lung sits inside a weatherproof dome in which a blower lowers the pressure just enough to carry the membrane’s weight by suction. In that state the pressure difference between inside and outside falls to within eight pascals of zero, and there is almost nothing pushing air either way. Dempster patented the arrangement. It is, in his words, the enabling technology for everything else.

Measuring the leak

A leak rate is a claim, and Dempster measured it twice by methods that could not share an error. First, he pressurised the building to about 150 pascals and timed the lungs’ deflation as air was forced out through whatever holes existed: 65 percent of the atmosphere a year at that pressure. Since flow through small holes scales with the square root of the pressure difference, the same holes at the operating range of plus or minus eight pascals give less than ten percent a year.

Second, he spiked the air with gases that nothing inside would consume, sulfur hexafluoride, helium and krypton, and watched them thin as outside air crept in. The sulfur hexafluoride fell about eight percent a year between May 1992 and April 1993, on a clean straight line. Helium, a much smaller molecule, diluted at nearly the same rate, which told him the leaks were not a myriad of microscopic pores or seepage through the caulk but a few ordinary flaws. He later put a figure on them: across eighty kilometres of joints and sixteen of welds, the pinholes and construction faults added up to a single hole about 19 millimetres across.

For scale: the Test Module leaked 24 percent a year. NASA’s Biomass Production Chamber at Kennedy Space Center, the best sealed plant-growth facility in the United States at the time, exchanged between two and ten percent of its air a day. Biosphere 2, at under ten percent a year, was tighter than anything of its size by about two orders of magnitude, and nothing built since has matched it.

The number matters because it is the difference between an anecdote and a measurement. In a building that leaks a few percent a day, a slow change in the air is washed out by the outside before anyone can see it. In this one, Dempster showed afterwards by simulation, a leak as small as one percent a day would have hidden the oxygen decline entirely. The crew would have felt fine and learned nothing. Instead the oxygen fell on a curve that could be read, and when it had to be found, the books could be balanced because the atoms had nowhere to go.

Closed to matter, open to energy

The seal was for matter only. Sunlight came through the glass, about half of it; the rest was lost to reflection, absorption and the shadows of the frame, which is why the farm went hungry in two cloudy winters. Electricity came in by cable from an energy centre outside, and heat came and went by water in sealed pipes to 25 air handlers, because on a summer day the glass would otherwise have cooked the interior to 65 degrees.

The same air handlers, condensing moisture out of the humid air, were the water supply: 20,000 to 40,000 litres of condensate a day, clean enough to drink, with the whole stock of water passing through the crew, the crops, the marsh and the air in a few weeks. The carbon dioxide in the air turned over in about four days. A system that fast punishes every imbalance quickly, and that, too, was by design.

What the seal bought

Nothing else about the project is beyond argument: the science, the crew, the management, the money. The seal is. It held for two years at a standard that made the air inside an instrument, and it was the seal, not the sensors, that found the missing oxygen. Columbia University cut openings in it in 1996 to run the building as a flow-through greenhouse, and it has not been closed since. Dempster’s conclusion in the engineering journal was modest: that closed systems of many sizes would be needed for the work ahead, that their envelopes would be asked to do what nature does for free, and that he hoped the lessons would be used.

Sources

• William F. Dempster, “Biosphere 2 engineering design,” Ecological Engineering 13 (1999), 31–42. The volume, the liner, the glass, the lungs, the two leak measurements, the energy and water systems. ecotechnics.edu
• William F. Dempster, “Tightly closed ecological systems reveal atmospheric subtleties: experience from Biosphere 2,” Advances in Space Research 42 (2008), 1951–1956. The stranded-wire problem, the 19 mm hole, the pressure range, the NASA chamber comparison, and the simulation of leakier systems. ecotechnics.edu
• William F. Dempster, “Methods for measurement and control of leakage in CELSS and their application and performance in the Biosphere 2 facility,” Advances in Space Research 14 (11), 331–335, 1994. The original leak measurements.
• Mark Nelson and William F. Dempster, “Living in space: results from Biosphere 2’s initial closure, an early testbed for closed ecological systems on Mars,” in Strategies for Mars, American Astronautical Society Science and Technology Series vol. 86, 1996, 363–390. The Test Module’s size, closures and leak rate, the lungs’ combined volume, and the four-day residence time of carbon dioxide.
• Abigail Alling, Mark Nelson and colleagues, “Biosphere 2 Test Module experimentation program,” 1990 and 1993, for the 1988 and 1989 closures.