Imaging Dyson Shell Civilizations with the James Webb Space Telescope

A Dyson shell is an enclosure around a star that captures most of its power for technological uses, described by Freeman Dyson in a letter to Science in 1960. Dyson's shell was later renamed a "sphere" by a careless journalist, though Dyson was agnostic about configuration . Most hypothesized designs are destructive or impossible. Some propose solid spheres, which are dynamically unstable. Some propose inhabitable shells, which are vastly more difficult.

This page proposes a Dyson shell composed of independently meter-scale "statites", capturing sunlight in a thin membrane composed of ice and nanometer-scale mechanisms, performing computation, emitting heat towards deep space, maneuvering in and supported by light pressure at 50 AU distances. This would be a vast undertaking, requiring thousands of years, but conforms to ordinary Newtonian physics. The goal is the efficient and safe use of 386 trillion terawatts of sunlight, ten billion times the power intersecting the earth, in a cold environment more conducive to efficient computation. The observational goal is finding other civilizations that have done this, which may be possible with existing telescopes.

I argue here that observing Dyson shells will be vastly easier than observing civilizations confined to planets. JWST may be able to image shells at the Nyquist limit out to 400 parsecs, and detect shell-like anomalies out to thousands of parsecs. Cold uniform Dyson shells, if they exist, will be easy to detect and differentiate from natural phenomena.

I am an engineer, not a physicist or an astonomer. For engineered objects to work, the physics must be correct, and I hope this page is acceptably so. While some of the best physicists and astronomers are also pretty good engineers, they are mostly watchers, not manufacturers. Observable Dyson shells will be manufactured by engineers, and though they evolve under other suns, their shells may best be understood by other engineers.

Why Dyson shells? Why not look for radio emissions or industrial gasses in planetary atmospheres?

Radio: Earth civilization passed through the peak "broadcast radio era" in less than a century. The galaxy is more than ten billion years old, three star generations, and life might originate anytime in the second or third generation where stellar metallicity is high enough - perhaps 6 billion years. Assuming a 0.1% chance of intelligent life originating around a G star (hotter stars burn too fast, red dwarfs will tide lock in the so-called-habitable zone) means 1e5 candidates in our galaxy, and a 100/6e9 chance of radio signals from their time window arriving here now. About a 0.17% chance - and that assumes we have enormous high gain dishes far from earth sweeping the sky, and a good way to differentiate intelligent noise (say, the randomly combined sync pulses of a thousand television stations) from all the other radio noise out there. The advantage of radio for SETI is that, away from molecular gas clouds, radio spectrum is quiet. That does not offset the enormous cost and low chance of success.

Gases: We will hopefully find life around other stars soon, spotting planetary oxygen atmospheres by some means. Water is everywhere; lifeless planets in our own solar system have water. But stars also have oxygen, and distinguishing planetary oxygen during a transit from the oxygen and noise of a star is problematic. A coronagraph might spot nearby planets with oxygen, but a star 10K parsecs away needs a telescope dish and a coronagraph with micro-arcsecond selectivity at >100 dB differential gain. That will only suggest life - not intelligent life. The hallmarks of industrial civilization are illusive: Fluorinated hydrocarbons? fossil carbon dioxide (lower-than-expected C14)? Spotting sure signs of life will be a great achievement, and is likely to happen with enough high-precision observation, but intelligent life will provide far fewer clues.

The De-Beneficiation Problem: Our own industrial civilization is dependent on our rapid consumption of ore bodies that took billions of years of plate tectonics, water/oxygen chemistry, life, and multitudinous geological events to beneficiate (concentrate by natural means). We are getting better at recycling, but eventually the "industrial atoms" disperse into an undifferentiated environment. Indeed, many industrial concentration processes for elements and compounds we value a lot result from entropy exchanges with more abundant compounds we value less. At some point, we maximize entropy, and we will need very expensive energy-intensive processes to reject that waste entropy away from the earth. This is not intrinsically impossible, but at some point it becomes practically too expensive and hazardous on one small planet. "One planet sustainability" is observationally identical to "no technology".

Visible Extraterrestrial Civilizations

Persistent civilizations (lifetimes on the order of gigayears) will be much more observable than kiloyear ones, so that even if they are rare, they will dominate what is available for us to observe. Persistent civilizations should be what we are looking for, not "civilizations like us right now". Those civilizations will not continue to accumulate pollution forever, so they will learn how to accomplish needful tasks without adding more pollution than they remove. They may use huge amounts of computation, because computation can substitute for resource use: a computer-designed smart phone uses less material than a Bell 500 desk set, telecommuting uses less transportation fuel, smart routing permits long-hull ton-mile-efficient container ships, industrial process control can maximize production efficiency and minimize wasteful defects better than humans pulling levers, etc.

Space Energy: While some persistent civilizations will make do for a very long time with the energy resources of their birth planet, stars of our type ( G and K ) evolve into red giants and eventually cook their inner planets. Our Sun will cook the Earth within 100M to 500M years. Even if we managed to last that long, we would not survive on an unmodified Earth in an unmodified solar system for billions of years. On the other hand, the sun emits ten billion times as much energy as the Earth intercepts, and thirty quadrillion times as much energy as human brains use (emitting waste heat at 300K). That energy is going to waste now. A persistent civilization will learn to use some or all of that off-planet energy, without damaging the life-bearing home planet; perhaps even using space energy to build sunshields or create off-planet environments for long term survivability as the home star evolves.

Information can move with a tiny fraction of the power needed to create it - whether we ever use space energy beamed to Earth for industrial/technical use, we can use space energy to create and manage information in space and a tiny fraction of that power to exchange information with Earth (the subject of this Server Sky website). How much information will a persistent civilization accumulate? The answer so far, based on a single example, is "as much as possible." Huge quantities of bits require significant energy input to maintain, and vastly more to sort and select. The cost of a bit operation is proportional to ln(2)kT, where T is the system's ambient temperature; hot systems require more energy than cold ones for the same amount of information production and storage. More information can persist in colder systems, Arhennius acceleration of charge leakage and chemical degradation is exponential with temperature. That is why large data centers use a large fraction of their input power for cooling, and why nobody builds kilowatt CPU chips.

Singularities are Pathological: This flies in the face of "singularity philosophy", the "faster-hotter-denser" theorists whose prinicipal metric is the speed of light in uniprocessor systems computing non-parallelizable tasks. Computation in real systems is limited by heat out, not energy in. Most real tasks benefit from parallelization, the more the better, and intimate coupling with a cold heat sink. Below-ambient heat sinks require refrigeration power, leading to a bigger total heat sink. These Frankenstein monsters might amuse horror fiction readers and unscientific philosophers, but they are not useful or practical. Life evolved around parallel processing, and we should learn from the experts - especially if those experts are biological cells.

Dyson Shells (sometimes called spheres) collect all of a star's power and emit heat into 2.7K deep space, limited by the black body emission temperature on the outside of the shell. Shells constructed from objects in nested Keplerian orbits are heavy, and objects eclipse each other. Solid spheres are unstable to perturbations.

"Statite" sparse shells supported by absorbed light pressure against solar gravity have a surface mass density of 0.77 g/m2 = L / 4 π c μ, assuming equal infrared emissions on both sides and no additional incoming infrared from other statites. If the shell is optically dense and contains an infrared bath, with zero albedo Lambertian surfaces, then the total thrust increases by 5/3 to 1.28 g/m2. Since light flux and gravity are both inverse square, this ratio is fairly accurate beyond a few solar radii. At 50 AU, the flux is 0.55 W/m2, and with 75% emissivity on the outward side, the black body temperature (Stefan-Boltzmann law) is 60 Kelvin. At that temperature, ice is stable for geological time, and there appears to be enough ice in the Kuiper belt to build a shell of statites at this density and radius. Ice is the most common low-temperature structural material in our solar system, and in the universe we observe.

Infrared Filtering Inner Surfaces, like a grid of conductive wires spaced 30 μm apart, can greatly reduce emissivity into the inner system. This removes an important source of outward light pressure thrust, although the outwards emissivity and black body temperature remains the same. That can reduce the mass of the shell to 0.77/3 or 0.26 g/m2.

Path Dependence Versus Technological Convergence: At first glance, it seems silly to presume that all spacefaring technological lifeforms will, in a small fraction of a billion years, build such a shell around their star. But all do not have to. What is important is that civilizations that capture all their star's energy will be billions of times more visible than those that do not.

If they originate from a Neptune-temperature planet, the 60K infrared heat bath inside the shell will significantly heat their home planet. If they come from an Earthlike planet at 250 K, their planet will be heated 0.21 K by the extra background heat. An infrared-filtering statite inner surface can eliminate the heating, but will not change outward appearance - an efficient heat-emitting outer surface, radiating all the power of the enclosed star, as cold as possible.

We cannot accurately guess what technologies will be used for sunlight absorption, energy conversion, information processing, information storage, repair, radiation protection, or heat transfer. Those are inside the heat emitting surface, which from a distance will look almost the same, regardless of the technology it contains. If the statites are large hexagons (for example), separated or overlapping but with just enough maneuvering capability to remain perpendicular to sunlight and separated from neighbors, there may be small light emitting gaps between them. That light would be impossible to directly image, but patches of trillions of identical statites might produce Fourier interference patterns (like a point spread function) that would be imageable from a long way away.

But these are small changes to the overall idea:

Temperature Range and Inclusions

The above assumes a uniform temperature range for the whole sphere - this may not be justified! Some statite sections may be farther in and higher temperature, some sections may be much further away and colder. The farther away sections will need "feathered edges"; because the sun's disk inside the shell has a finite size (about 39 arcseconds or 187μrad at 50 AU). A displaced section at 51 AU must have edges that taper from full mass to zero mass over 28,000 km - and it will only be 1% colder.

The easier way to produce "colder" is to make small holes (perhaps of complicated shapes) and float smaller statites in the diverging light and heat flux. The reflected heat flux from these smaller colder statites will heat up the area around the hole a bit, since they will be warmer than the 2.7K background. This should not be difficult to model, but I will not do so now.

These variable temperature inclusions just above the outer surface of the sphere will be small (like statites themselves), otherwise they will fold up around the center pressure gradient. From a distance, if there are many of them, they will spread out the black body spectrum, though they will be within a few kilometers of the main shell and will not affect the sharpness of the distant image. Again, I will not model this now, but it is an example of the kinds of engineering details we can learn when we image these shells precisely.

Imaging Disks with Diffraction Limits, Pixels, and Point Spread Function-Based Correction

These disks will have a star powering them; if they are F, G, or K stars, the images will be far larger and higher power than other 60K objects, and will appear uniformly illuminated across the disk. However, if they have sharp edges (say, plus or minus an AU), they will have lots of high frequency spatial information, which can alias down into the baseband. An imager like the JWST MIRI has a complicated Point Spread Function (PSF) that scatters across many pixels in the imaging array; deconvolving a Dyson Shell image by Inverse Fourier methods may create artifacts in the image. That is actually good - those artifacts (in an uncrowded star field) may be the most distinguishing characteristic of an imaged sphere.

More on this subject in "Astrophysical Techniques" C.R. Kitchin; section 2.1 p233 to 240 in my 5th edition.

James Webb Space Telescope, NASA, 2018?

Frontier Science with the James Webb Space Telescope - great slides!

MIRI - The Mid-Infrared Instrument

search the space telescope website for point spread function:

Point spread function modeling software paper

Point spread function paper

Hubble Exposure Time Calculators

A Giant Segmented Mirror Telescope: Synergy with JWST, GSMT AND JWST: Looking Back to the Future of the Universe

What can JWST resolve?

A Jansky is 1e-26 W / m2-Hz. The Hz bandwidth can be estimated from the wavelength and wavelength window: Hz = c Δλ / λ2 . The sensitivity of the F2550W MIRI instrument (10K second observation time) is 26.2 μJy, or 2.62e-31 W / m2-Hz, &lambda = 25.5 μm, &lambda/Δλ = 6. The bandwidth is 1.96 THz, so the power sensitivity with the F2550 filter is 5.13e-19 W/m2. The Dyson shell power is 6E24 W. 4π42 = 1.17E43 m2, r = 9.6e20 m or 31K pc. Note, the sensitivity may be exaggerated, because the sensor is more sensitive for shorter wavelengths, while the black body spectrum of a 60K source peaks at 48μm and plunges exponentially for shorter wavelenghths.

If we assume the sensitivity is linear to exposure time, and assume a ten second exposure (with JWST sweeping the sky), 1/1000 of the power means a 1 K pc range. My guess is that 6 readouts a minute might put too much clocking power into the CCD imager and heat it up; OTOH, the sensitivity is probably more like the square root of exposure time, RMS averaging of noise. We can locate candidates for future detailed observation by sweeping JWST across the sky.

Assume the (rare) Dyson shells enclose mostly G stars and some hotter K stars, 10% of all stars, 100 cubic parsecs per candidate star. Hotter stars probably burn too fast to evolve intelligence, colder stars may be difficult to see, the "habitable zones" of red dwarfs will be tide locked. 10% is a conservative estimate of suitable stars for the emergence of intelligence, though metallicity and planetary formation will restrict suitable candidates to a tiny fraction of 10% of the stars.

Two pixels needed for Nyquist deconvolution of point spread function. 0.11" (arcsecond) pixels means 0.22 as minimum for estimating size (more is better) A 100 AU round shell is 0.22" across at 450 pc. The galactic disk is 300 pc thick. A 450 pc radius, 300 pc thick cylinder is 1.9e7 pc3, 190,000 candidate stars, of which a tiny fraction will have intelligent life evolving into a Dyson shell. The size of that tiny fraction cannot yet be estimated accurately. These are the Dyson shell stars that may show an imagable disk with JWST.

There is already a good synoptic study by the WISE Wide Field Infrared Survey Explorer that may tell us where to look - the candidates will be very bright, compact, and obvious.

I am not optimistic about the density of long-lived civilizations out there - there are way too many ignored factors in the Drake equation. IMHO, the current search for "shirtsleeve" extraterrestrial planets is ludicrously optimistic. The planetary systems we can see (outstanding accomplishment!) vary way too much. Planetary formation is chaotic - we are here because of a long string of fortuitous accidents, accumulated chance collisions in a cosmic pinball machine.

However, what the planet-searchers are doing is explainable to the taxpayers, and what I write about here is not. If the exoplanet-hunting astronomers are not drinking too much of their own koolaid, perhaps they can think through what they are actually looking for, and use the superb JWST instrument to look for Dyson shells, not biology-hinting fractional petawatts from tiny naked mirror earths next to 400 yottawatt stellar furnaces.

I fear that intelligent life is too rare and too suicidal to find Dyson shells within the range of JWST, but if we survive long enough to build a spacefaring civilization, we will someday build much larger and more capable instruments in space (not just unfold them from a rocket payload) and will be able to image shells in nearby galaxies (a side view is easier than edge-on through gas clouds). Will we be able to spot such shells in our own galaxy, through gas clouds and past a dense disk of intervening stars? I don't know, but we can image the galactic center in infrared, so with multi-AU sized interferometers, perhaps someday we will eventually be able to survey most of the candidate stars in our galaxy.

If we do spot a shell, it will be much too far away for a bidirectional conversation. But we might learn a lot about how it is put together, helping us build our own, as well as a gigayear-duration civilization to make use of it. Today, we are agents of entropy, using up ore bodies concentrated by billions of year of plate tectonics, building ephemeral structures and devices that rapidly corrode, scattering the atoms into homogenous high-entropy waste. Stellar-sized Dyson shells will generate a prodigeous amount of entropy, and it remains to be seen whether the entropy difference between 5778K sunlight and 2.7K deep space can support a computation shell over the long term. By studying the emissions of a distant shell, and the operations going on around it, we can reverse-engineer their overall process. The gigayear survival of intelligence in the solar system may depend on it.

Other Infrared Space Telescopes

WISE Wide Field Infrared Survey Explorer, NASA, 2009-2013 10 month cold mission

What can WISE resolve?



guesstimated response to 60K 3.8e26W Dyson shell




20 μm




21 μm




22 μm




23 μm




24 μm




25 μm




26 μm




weighted total:


MoreLater - test the idea better.


Spitzer Infrared Satellite, NASA, 2003-present (now "warm")

What can Spitzer resolve?

Herschel Infrared Satellite, ESA, 2009-2013

What can Herschel resolve?

IRAS Infrared Astronomical Satellite, NASA+UK+Neth, 1983, 10 months

What can IRAS resolve?

Infrared Space Observatory, ESA, 1995-1998 (28 months)

What can ISO resolve?

ESA Planck, 2009-2013


SOFIA airborne observatory, DLR/NASA, 2013 to 2033?


What can SOFIA resolve?

The power of a 60K 3.86e26 W black body emitter between 35.5 μm and 38.5 μm is about 1.7e25 W.

Using the 37 μm dichroic filter, 3.5 μm wide, 900 second exposure, the bandwidth is 766 GHz. The power sensitivity is 0.42 Jy * 766 GHz or 3.22 e-15 W/m2.

4πR2 = 1.7e25 W / 3.22 e-15 W/m2 = 5.28e39 m2, R = 2.05e19 m. A parsec is 3.0857e16 meters, so R = 660 pc.


Infrared Ground Telescopes

UKIRT, 1979-present

NASA IRTF, 1979-present

ESO 3.6 m Telescope, 1977-present

Gran Telescopio Canarias, partially complete


JWST (last edited 2015-10-10 19:19:46 by KeithLofstrom)