Cold, Vacuum, and a Hundred Thousand Joints

When I wrote about building a mile-wide habitat out of ten-inch printed steel panels, I spent most of the article on the ground: who prints them, who inspects them, how the joint locks. I said very little about what happens to a panel once it is in space, and a friend who has spent as long around steel as I have asked the obvious question on the drive home.

Material shrinks when it gets cold. Up there it is going to be very cold. What does that do to a joint that was cut to a few thousandths on a bench in somebody's garage?

That is the right question, and once you start pulling on it, other questions come with it. This article is the list. I am going to go through everything I could find that the space environment does to a structure like this one, put a number on it where I can, and say plainly where the number hurts. Some of it turns out not to matter. Some of it matters a great deal. And one item on the list is a correction to my own article, so I will start there.

First, the correction

In "A Thousand Garages" I wrote that a printed stainless part, after sintering, "comes out with the properties of ordinary three-sixteen-L stainless." That is too generous, and I should have checked it harder before I printed it.

The honest numbers are these. A part made by binder jetting — powder held together with a glue, then baked in a furnace until the steel fuses — comes out around three percent porous, with the pores clustered in layers where the print was laid down. Tested against wrought three-sixteen-L, it has about half the yield strength and about seventy percent of the ultimate strength.¹ Fatigue is worse. One study found the fatigue strength of binder-jetted parts well below both laser-printed and wrought material, and found that the rough as-printed surface alone knocked another twenty percent off it.² Parts made from metal filament on a hobby printer go through the same furnace step and land in the same neighborhood.

Laser-fused parts are a different story; those come out as strong as wrought or stronger.¹ But the laser machines are the sixty-thousand-dollar tier in my article, not the garage tier.

So the garage tier makes parts at roughly half strength with wide scatter. I could take the easy way out of that and say structural parts come from the sixty-thousand-dollar machines and the garages make brackets. I am not going to, because the whole point of the last article was ten thousand garages, and I am not giving that up on the first hard number.

Look at where the strength actually comes from. A garage printer does not make a steel part. It makes a green part: steel powder held in a glue, in the right shape, to the right dimensions. The strength is made afterward, in the furnace, where the glue burns off and the powder fuses. And the furnace is not in the garage. It is at the yard, or at whoever the yard contracts to run it. So the half-strength number is not a garage problem. It is a furnace problem, and furnace problems have known answers.

The first is hot isostatic pressing, which the aerospace industry has used for decades to rescue castings and, lately, printed parts: put the sintered part in a vessel, heat it, and squeeze it from every side under gas pressure until the pores close. That is how you take a three-percent-porous part and make it a dense one. The second is peening the surface, because a lot of the fatigue loss in a printed part lives in its rough skin, and one of the studies behind this article found that working the surface recovered a good share of it.² The third is the plainest move in engineering: size the panel for the steel you actually have. If the material is half as strong, put more of it in the tongue. The spin load is known to the pound.

So the correction changes the yard, not the network. The garages print every part, structural ones included, as green parts to a drawing. The yard runs the furnace, the press, and the coupon tests, and stamps what passes. That is more work at the yard than I described, and the sixty-thousand-dollar laser shops still have a place for the parts that need it most. But ten thousand garages printing structural panels is still the plan, and now it is a plan with the steps that make it true. I am leaving the original article as written, with a note pointing here, because that is what the sources say and I would rather you see me change my mind than pretend I never had to.

Now the cold.

What steel does when it gets cold

Stainless steel shrinks about nine and a half millionths of an inch, per inch, per degree Fahrenheit. That is the kind of number that sounds like nothing until you multiply it.

On a ten-inch panel, a swing of a hundred degrees moves the length about ten thousandths of an inch. A swing of four hundred degrees, which is a fair bounding case for a bare metal plate going from full sun to full shadow, moves it nearly forty thousandths. A good printer holds a few thousandths. So the cold moves the part ten times more than the tolerance it was made to.

Here is the part that saves the joint: every panel shrinks the same amount. Same alloy, same temperature, same geometry. The tongue shrinks, the slot it sits in shrinks, and the fit between them barely changes. A uniform temperature change makes the whole ring smaller, not looser. That is the answer to my friend's question, and it is a good one.

Then multiply it around the ring. A mile-wide circle is a little over sixteen thousand five hundred feet around. Four hundred degrees of shrink takes about sixty-four feet out of that circumference and about twenty feet out of the diameter. Twenty feet. A structure a mile across breathes twenty feet between full sun and full shadow.

I want to be careful with that number, because it is a bounding case and two things pull it down hard. The first is mass. A mile of steel does not chill four hundred degrees in an hour; the thin plate the number was calculated for does. The second is the spin, and I did not see this one coming.

The whole design turns on spinning the habitat for gravity. At a half-mile radius that is about one turn a minute. A body turning once a minute in sunlight is a rotisserie. No part of the rim faces the sun for more than half a minute before it turns away, and it comes back around half a minute later. The temperature evens out around the ring the same way the load does. The spin, which I already called the fastener, is also the thing that keeps one side from cooking while the other freezes. So in steady sunlight the ring settles to one temperature and stays there, and the twenty-foot number never shows up.

Where it does show up is the eclipse.

Two seasons of darkness

A geosynchronous orbit is a kind place to live, thermally. For most of the year the structure is in continuous sunlight. Only near the two equinoxes, for about forty-five days each, does it pass through Earth's shadow once a day, and the longest of those passes is about seventy-two minutes.³⁴ So instead of the sixteen sunrise-to-shadow cycles a day that a low-orbit station takes, this structure takes roughly ninety a year, and they come in two predictable seasons that every satellite operator already plans around.

That changes the problem from fatigue to survival of a known event. During those seventy minutes the rim cools, the ring shrinks by some amount between nothing and twenty feet depending on how much steel is behind the skin, and then the sun comes back and it grows again. Ninety times a year, for however many years the thing stands. It is not the cycling that worries me. It is the gradient.

The outer skin cools first. The structure behind it, shaded and heavy, cools later. For those minutes the skin is shorter than the frame it is fastened to, and that is the one loading a sliding dovetail does not want: a shear across the joint face, not a pull at right angles to it. On a steel-on-steel structure at a single temperature, the geometry handles it. During an eclipse, the design has to handle the skin and the frame being at two different temperatures for an hour, twice a year, forever.

I do not have a finished answer for that. I have a rule: keep the structure one material. The moment an aluminum bracket bolts to a steel panel, or a plastic tray sits in a steel frame, two things shrink at different rates every eclipse. Aluminum moves about twice what stainless does. Every mixed joint is a joint that opens or crushes ninety times a year. Steel to steel, all the way through, and put the aluminum and the plastic inside, where the crew keeps the temperature.

The other rule is about assembly. A dovetail driven home warm and then cooled forty thousandths will either be locked up tighter than it was designed for, or it will be a hair loose in the direction the spin is pulling. Which one depends on whether the tongue or the slot has more metal around it. That is a bench test, and I will come back to it at the end.

The joint that welds itself

The next item on the list is one I had heard of and never seen, and it is the one that scares me most.

In vacuum, two clean metal surfaces pressed together can simply become one piece. On Earth every metal surface carries a skin of oxide that keeps it from bonding to the next one. In space there is nothing to rebuild that skin once it is rubbed off. The effect is called cold welding, and it has already cost NASA a mission.

In April 1991 the Galileo spacecraft, on its way to Jupiter, tried to open its main antenna, a sixteen-foot dish that unfolded like an umbrella. Three of the eighteen ribs never came out.⁵ Engineers spun the spacecraft up, cooled it, warmed it, and pulsed the deployment motors more than fifteen thousand times trying to shake the ribs loose. Nothing worked.⁶ The accepted explanation is that the ribs' restraint pins had lost their lubricant during years of ground transport, and the vibration of trucking and launch had rubbed the clean metal surfaces together until they bonded.⁷ Galileo flew the whole mission on a backup antenna with a tiny fraction of the bandwidth. The pictures still came back, but they came back slowly, and a lot of the science was lost.

The European Space Agency spent years afterward testing which metals weld to which. The finding that matters here is that a static contact barely welds at all. What does the damage is fretting: small, repeated rubbing under load, the kind that strips the oxide off faster than vacuum can fail to replace it.⁸ Launch vibration is exactly that. And a stack of two hundred steel panels tongue-to-slot in a fairing, shaking for eight minutes, is exactly that too.

So here is the fork in the road, and I have not decided which way to go.

One way to look at it: a joint that is designed never to come apart, welding itself shut in vacuum, is a gift. The structure grows stronger with time for free. That is the version I like.

The other way: the panel has to slide into place first. If it welds to the one it is stacked against in the fairing, or seizes halfway into its slot the first time it hesitates, the robot placing it is stuck with a part it can neither seat nor remove. That is a Galileo rib, a hundred thousand times over.

The answers the satellite people use are coatings that leave a surface that does not bond, dry lubricants that survive vacuum, and mating different alloys, because like metals weld to like more readily than to something else. Any of those can be put on a tongue at the factory. None of them is free, and every one of them is a surface the yard has to inspect. I would rather know that now than find out from the first robot.

The radiation bill

Geosynchronous orbit does two favors for a steel structure and then charges for them.

The first favor: there is no atomic oxygen up there. In low orbit the thin edge of the atmosphere is mostly single oxygen atoms, and they eat exposed polymer at a measurable rate; the space station is coated, inspected, and patched against it. NASA's own materials guideline says geostationary spacecraft are not covered by the atomic oxygen requirements at all, because there is nothing up there to erode them.⁹ One whole failure mode, gone.

The second favor is the eclipse arithmetic above.

The bill is radiation. Geosynchronous orbit sits in the outer Van Allen belt, and the dominant hazard there is not protons, as it is in low orbit, but trapped electrons, along with whatever the sun throws during a storm.¹⁰ Steel does not care; a few inches of it is shielding, not victim. What cares is anything organic. Plastics in a space environment take doses measured in megagray per year,¹¹ which is a number that turns a polymer brittle, then yellow, then to powder. The printed plastic prototypes I promised in the last article are bench prototypes. They are for the kitchen table, and nothing plastic goes on the outside of this structure, ever.

There is a related effect that turns out to favor the design. Surfaces in that electron environment charge up, sometimes to hundreds or thousands of volts, and then discharge in an arc. That is a real satellite killer, and the usual defense is to make sure every surface is conductive and tied to every other one. A structure made entirely of steel panels in metal-to-metal contact, with no insulating joints, is one continuous conductor by nature. It cannot build a potential across itself, because there is no across. That is a nice thing to get for free.

The rest of the list

There are four more, and I will be brief because I have less to say about them.

Micrometeorites and debris. A mile-wide structure is an enormous target, and there is a graveyard of dead satellites parked a few hundred kilometers above the geosynchronous belt. The good news is that everything at that altitude is moving the same direction at nearly the same speed, so a collision with a piece of dead satellite is a slow bump, not a shot. The bad news is that natural micrometeorites do not care what orbit you are in, and they arrive at tens of kilometers a second. The structural rim is not the pressure boundary; that is a separate hull, and I wrote about its wall thickness in "Sixteen Inches of Steel." But a puncture in a rim panel is a panel that now needs replacing, and that is a job for the robot. I will come back to the graveyard in its own article, because I think it is a supply yard, not a hazard.

Outgassing. A sintered part carries a little binder residue, and in vacuum it cooks off and plates onto anything cold. On a structure with no optics that is a nuisance, not a failure. Bake the panels before they ship.

Creep. Steel under constant load slowly stretches, but not at these temperatures. Below several hundred degrees the effect on stainless is negligible over decades. I looked for a reason to worry about this one and did not find it.

Inspection. This is the one I cannot put a number on, and it is the one closest to my own trade. On the ground, the yard inspects every panel before it flies. Once it is placed, nobody is going to walk a hundred thousand joints with a flashlight. The best answer I have is that the robot placing the panel is also the instrument: it knows how much force it took to seat the tongue, and how far it traveled, and a joint that seats short or seats easy gets flagged before the next course goes over it. The keystone in each ring, the one panel that does need a pin, can carry a strain gauge. That gives you one reading per ring, which is thin, but it is something, and it is what the spin is doing to the structure in real time. If somebody who does this for a living has a better answer, I want it.

What this changes

Not the design, yet. The panel is still ten inches, the joint is still a sliding dovetail, the courses are still staggered. What changed is my confidence in three places.

The garage-printed part is half as strong as I said it was, with wide scatter. The garages still print everything; the yard now has a furnace, a press, and a test lab to turn what they print into steel, and the panel gets sized for the steel that comes out.

The joint has to be designed for an eclipse gradient, not for a uniform temperature, and the structure has to be one metal throughout so the gradient is the only one it sees.

And the tongue needs a surface treatment before it ever goes in a fairing, because a hundred thousand joints that weld themselves in the stack are not a structure, they are a brick.

What I am doing next

The bench test I promised last time gets one more step. Print the panels. Slide them together warm on the table and note the force it takes. Then put the assembly outside. It is Alaska; in January I can get thirty below zero on the porch without any equipment at all, which is a hundred-degree swing from the kitchen and enough to move a ten-inch part ten thousandths. Bring it in. Slide it apart, if it will come apart. Note the force again. Then do it in the other order: assemble cold, bring in warm, and see whether the tongue crushes or the slot gaps.

That is a plastic part on a porch, not a steel part in orbit, and I know the difference. But the direction of the answer will be the same, and the direction is what I need before anyone winds a coil.

If you run a printer and you have a freezer, you have everything you need to do the same test. Write in.

Ideas in Star Bound are published open — take them, build them better. Our open proposal.

Comments

Star Bound is not a forum, so comments run the old way: if you have comments, please write to us at hello@starboundmagazine.com and they will be published here in the next update, with your name on them. Say if you would rather stay anonymous.

No comments posted on this one yet. Yours could be the first.

Share this article

Copy link Facebook X Email

A note on where this comes from. I am not a metallurgist, a spacecraft engineer, or a degreed engineer of any kind. But I am a practical one, with more than forty years in heavy industrial construction: welding, millwright work, hydraulics, mining, and thirty-three years of electrical construction and commissioning on the North Slope of Alaska, in a place that punishes anything built wrong. Much of that time I have worked as a field engineer, which means I am the one standing in front of the equipment when the drawing turns out to be wrong, and the fix has to be made on the spot, with what is on hand, and has to work the first time. Anyone who has done that long enough learns that a lot of engineering gets corrected in the field by people with more experience than the people who drew it. That is the eye I bring to these subjects: not what should work on paper, but what actually gets built, and what breaks. The facts I take from other people are footnoted so you can check them and read further. The ideas built on top of those facts are my own, and I would rather see them knocked down by somebody who knows better than left standing because nobody looked.

Sources

1. Binder-jetted 316L tested against wrought: roughly 3 percent initial porosity, pores clustered in layers perpendicular to the build direction, and a stress-strain curve that lies below wrought with about 50 percent lower initial yield stress and 30 percent lower ultimate strength; laser-fused (SLM) 316L, by contrast, can exceed wrought yield strength. Roth et al., "Rate-dependent strength and ductility of binder jetting 3D-printed stainless steel 316L," International Journal of Mechanical Sciences, 2021. ↩
2. Fatigue strength of binder-jetted 316L well below both laser powder bed fusion and wrought 316L; as-built surface roughness lowers fatigue strength by roughly 21 percent, and post-processing the surface (polishing, and in a related study severe shot peening) recovers part of the loss. "Effect of Surface Roughness on Fatigue Behavior of 316L Stainless Steel Produced by Binder Jetting Process," TMS 2022 proceedings, Springer. ↩
3. Geosynchronous eclipse seasons: two periods a year centered on the equinoxes, roughly 44 to 45 days each, with daily eclipses of up to 72 minutes. Space Norway, "The eclipse season and what it means for satellites." ↩
4. NOAA GOES eclipse schedule: batteries for up to 72 minutes a day, late February to mid-April and late August to mid-October. ↩
5. Galileo high-gain antenna anomaly, April 1991: the 4.8-meter antenna failed to fully deploy; a few (probably three) of 18 ribs remained closed. IEEE Aerospace Conference, "Open! Open! Open! Galileo High Gain Antenna anomaly workarounds." ↩
6. Attempts to free the ribs, including spin-up to 10.5 rpm and more than 15,000 motor pulses ("hammering"). JPL, "Galileo's High Gain Antenna FAQ." ↩
7. Michael R. Johnson, "The Galileo High Gain Antenna Deployment Anomaly," 28th Aerospace Mechanisms Symposium, NASA, 1994. ↩
8. Cold welding in hold-down points is driven by fretting under launch vibration, and omitting grease raises the risk; ESA test programs (Merstallinger et al.) characterized material pairs for it. Merstallinger et al., "Cold Welding in Hold Down Points of Space Mechanisms Due to Fretting When Omitting Grease," Lubricants, 2021. ↩
9. "Geostationary, deep-space, and planetary spacecraft in general are not covered by this specification since the oxygen populations in those environments are negligible." NASA/TP-1999-209260, Material Selection Guidelines to Limit Atomic Oxygen Effects on Spacecraft Surfaces. ↩
10. In geostationary orbit the primary radiation sources are solar protons and trapped electrons of the outer belt, not the trapped protons that dominate low Earth orbit. Space environment analysis using ESA SPENVIS, in "Proton induced Dark Count Rate degradation in 150-nm CMOS Single-Photon Avalanche Diodes," 2022. ↩
11. Irradiation of plastic materials for space and nuclear application takes place at doses on the order of megagray per year. "Space environment simulation at radiation test of nonmetallic materials," Nuclear Instruments and Methods in Physics Research B, 1999. ↩
12. Editor's arithmetic. Thermal expansion of 316 stainless taken as 9.6 millionths per inch per degree Fahrenheit (a commonly published value for the 32–212 °F range). Ten inches × 400 °F × 9.6 × 10⁻⁶ = 0.038 in. A 5,280-foot diameter × 400 °F × 9.6 × 10⁻⁶ = 20.3 ft change in diameter, 63.7 ft in circumference. Spin rate for one g at a half-mile radius: about 1.05 rpm. The green-part-at-the-garage, strength-at-the-yard argument, the one-material rule, the eclipse-gradient concern, the rotisserie argument, the surface-treatment recommendation, and the robot-as-instrument idea are my own and are offered to be argued with.

The thermal-expansion arithmetic, the spin-rate figure, and the ring-circumference numbers are my own hand calculation from a published expansion coefficient (note 12) and are meant to give scale, not to size anything. The green-part-at-the-garage, strength-at-the-yard argument, the one-material rule, the eclipse-gradient concern, the rotisserie argument, the tongue surface-treatment recommendation, and the robot-as-instrument idea are my own proposals. The correction to "A Thousand Garages" stands on notes 1 and 2.