Stephen Hawking: Why Isn't the Milky Way Crawling with Life?

[Wyrm]

Ah, thank you. Do you have a source for that? I ask not because I doubt your word but because I'm curious to read it.

You've already read it.

It was in the paper you cited.

How much of a hurry? Regenerative braking systems on cars have charge put into them many times a day and only need replacement on a timescale of years. Flywheels often have quoted lifespans of 100K to 10 million cycles of use (ref - it's Wikipedia but it's sourced). Also, you can reduce wear on your energy storage systems by running your electronics directly off solar power when it's available (although that will create wear elsewhere; everything a machine does causes wear somewhere).

A good portion of the regenerative breaking on cars is actually wasted. The charging rate is limited by how fast charge carriers can migrate back to their original plates — if you recharge the battery too fast, you simply damage the battery. Rechargabe batteries are designed for a specific cycle: the lead-acid battery of your ordinary car is designed to spend most of its time at nearly fully charged, as is a UPS battery. Lithium-ion batteries, however, can be charged and discharged willy-nilly, up to a point, but their main disadvantage for us is that they have a limited shelf life — regardless of their treatment, their capacity inexorably decays as it gets older. By the time it reaches the system, a Li-ion battery would be dead as a doornail. Other batteries have a memory effect; frequent charging only partway causes the batteries to decrease capacity. Deep discharge tends to shorten life faster, as low as 200-1000 cycles at 100% depth for a NiCd. Also, batteries tend to lose charge even when not in use.

Actually, I think we can pretty much shelve batteries altogether. With any realistic STL speed, we've got a battery that has spent decades to centuries without a charge. Li-ions would be pretty useless, NiCd would be lazy beyond measure, and just about all batteries would have long self-discharged and decayed. Just about the only battery of any use would be a vanadium redox battery. The battery, however, must be kept warm, it requires pumps and sensors to keep the electorlyte flowing, and the battery's electrolyte is rather corrosive.

Flywheels are a more likely candidate, but remember that they too will lose energy with time, even in a vacuum. Also, their bearings will need replacing periodically. Magnetic bearings require power, as Earnshaw's theorem tends to spoil the party for permanent magnets, and limit acceleration.

The TOP500 system supercomputer eats 257 kW at peak processor power (source), so if we assume —for the moment— that the AI can run comfortably on it, the flywheel needed just to run the AI for a century (~811 TJ) is 1.62 million tonnes, and that flywheel will be spun down by the time it gets there. On the bright side, it'll probably keep everything toasty warm inside.

Kick-start from "battery" in the high MJ or low GJ range might not be infeasible. Modern FES systems can have energy densities in the 500 KJ/kg range (ref); 100 MJ could supplied by a 20 kg flywheel. Nevertheless, the problem you site is a valid concern. If one must keep the reactor constantly ticking over at high energy then probably be best strategy would be seek out relatively large icy bodies (in the range of several hundred meters to a kilometer or more). The decreasing surface to volume ratio of larger bodies means you will be able to get much more material from a somewhat larger body, and so will waste less time constantly looking for fresh deuterium sources.

Probably a good idea.

Thank you, it helps to actually have some numbers to work with. This is relatively encouraging - even at an order of magnitude less efficiency than the Vamork facility you'll still be getting a lot more energy out of the fuel than you put into extracting it.

Except that I couldn't even guess at the alpha (kg/(kg/s)) of an on-board deuterium extraction plant. It might be quite large to get even a measily 5 kg deuterium per year (remeber that you have to shove 8517 tonnes of water through this sucker to get it). The process also requires an electrolyte (sulfuric acid) to be added to the mixture to make it conductive, which must also be extracted. It will also require a non-trivial amount of sulfuric acid over 8517 tonnes of water processed.

(And I messed up the spelling: it's 'Vemork'.)

I think you misunderstand. I meant the probe replicates rapidly during the short period the comet is in sunlight once (i.e. within a time span of months), so by the time the comet gets too far away from the sun for solar power to be much use it has already finished.

Ummm, what? Aren't you here on this comet because you're going to harvest deuterium to get the seed energy you need so you can refine materials? To make more copies of yourself? How can you make more copies of yourself without materials?

True. On the plus side, you have access to a large heat sink, many thousands of tons of potential propellant, and abundant solar power (as long as you can build solar a power plant out of the locally available materials - structural strength will be a concern but not very much of one, as acceleration will be quite gentle). A torch drive is much more doable under these circumstances than it is for the probe itself. Solar powered mass drivers would probably be the way to go.

Okay, let's say we want to slow Hale-Bopp to have an aphelion of 20 AU within the six months either side of April 1, 1997, when it reached perihelion. With a semi-major axis of 186 AU and a perihelion of 0.914 AU, the velocity at perihelion is 44 km/s. The corresponding velocity with an aphelion of 20 AU is 43 km/s, so we need to have a delta-v 1 km/s over one year. Hale-Bopp's nucleus is about 60 km diameter of pure water, so this is a 1.13e14 tonne monster. The thrust required to affect this change over 1 year is 3.58 TN. If we suppose we want to use 1/4 of the mass to affect this change, we get an exhaust velocity of 721 m/s. This requires a power of 1.29 exowatts. This would require a solar panel 2274 km on a side (assuming that we begin thrusting at the orbit of Mars).

Where are you going to get a panel that big? The line losses will be tremendous! Also, that panel will be so fragile that even the piddly 0.0317 m/s² acceleration will probably shatter it. Furthermore, the only way you're going to get 3.58 TN of thrust in a reasonable mass is to go to a fusion engine. Furthermore, you'd need to mine ice at a rate of 896 million kg/s just to feed this monster.

If you're jumping onto and off of comets during these periods of the comets' orbits, then you're going to have expensive transfers.

If you can obtain your propellant readily from the comets, or if you're using a solar sail that doesn't need propellant, I don't see why this is a big problem.

Junghalli, 'obtain your propellant readily from the comets' skips over some very relevant details.

First, you're exploiting these comets because you need its fuel and mass. That means your tanks are nearly empty. You can't afford large delta-v maneuvers unless your engine has large specific impulse. Or a solar sail.

Second, you must locate the comet in time to catch it with your drive. Otherwise, the comet will escape before you can catch it. We on Earth have an advantage with millions of observers, amatuer and professional, scanning the skies for comets. Our probe is but one lonely craft, with all the attendant detection problems.

Third, you must then use your fuel or sails to match velocities with the comet, or you just crash into it and *pfft* goes your mission.

Let's say you detect a Hale-Bopp type comet with three months to match speed. If the comet intersects your orbit at 40° while at 1 AU, you need 60.1 km/s delta-v to catch it, which requires 0.0228 m/s² acceleration. If you use a VASIMR (~10 tonnes) at high gear (30 km/s @ 50 N), you need a mass fraction of 13% (87% propellant), and a engine fraction of ~1.5... which is impossible, even if we believe that we consider 87% propellant to be running on empty! A VASIMR at low gear (3 km/s @ 500 N) has a mass fraction of 1.9698e-9 and an engine fraction of 15%, also impossible, even if we believe the ludicrous mass fraction of 1.9698e-9! If we were to use a solar sail, the sail has to be 5.5346299938462e3 m² for each kg of probe mass. A 1000 metric tonne probe has to have a sail 74,000 km on a side. Ludicrous!

So, yeah. This is a big problem.

The best thing I could find was this paper which talks about propelling a spacecraft by accelerating many small lightsails with lasers and having them slam into a magsail as plasma (vaporized by an onboard laser on the ship). The light sails are on the order of 10 cm in diameter, have a mass of milligrams, and the paper talks about being able to launch one every few seconds. Since the sails must travel at .1 c to be able to catch up with the ship at the end of its acceleration this suggests acceleration of hundreds of thousands of G. They confirm in the paper they are talking about "a microsail capable of accelerating from rest to relativistic velocities in less than a second." This is using "only known physics and materials, although maximum system performance depends on improvements in materials, especially sail properties", and they recommend a diamond film sail. By a quick calculation, a 1 km sail might be able to take accelerations of around 30 G based on that.

Uhh... yeah. You realize the pulse laser destroyes the sail to send it out that quickly? The sail gets turned into a plasma, which the distant ship grabs onto with a magnetic field and uses as propulsion. As vaporization is part of these sails' operation, I don't think it applies to larger, nonvaporizing sails, which actually have to stay intact.

Yes, you need a start-up power source, but my point is you're only on a tight energy budget while you're still limited to it. Save for the ones that may be involved in building your power plant you can delay any high-energy processes you need to do until your power plant is complete.

And what makes you think building your power plant isn't itself a high-energy process, Tweedledum? Won't it need some form of refined materials to build 'em? They're not going to magically appear. Unless you're planning to build 'em out of dirt. Or ice. Or dirty ice.

Also, if you want to take a large flimsy construct with you, you can always limit acceleration to some small fraction of a G. Highly efficient rockets tend to have low thrust anyway. A bigger concern is the extra fuel you will need to accelerate a large power plant.

A large, flimsy construct will shatter to pieces even under the differences in pressure created by the solar wind, if it's large enough. Remember Mike's 'Size Matters' page? That's why ringworlds are generally regarded as impossible. A structure on the order of thousands of kilometers wide is not going to survive even gentle acceleration.

According to this site this company invests 45 kWh in the manufacture of a solar panel. A kilowatt-hour is 3.6 megajoules, so that's 162 megajoules to make a panel. They estimate 768.1 kWh with the sourcing and processing of raw materials. This probably involves shipping it over much of the planet, so it's probably pessimistic, but let's use that. We get 2.8 gigajoules per panel. Unfortunately it doesn't say how big their panels are, but 1 m^2 sounds reasonable. To manufacture 1 m^2 of panel per day would therefore require a generating capacity of 32.4 kW. According to Wikipedia, sunlight conversion rates of solar panels vary between 5-18%, let's use 10%, which gives us 137 watts/m^2. 32.4 kW would then require a square solar panel between 6-7 meters on a side. Our panel will require around 8 months to "pay back" the energy invested in its manufacture. Assuming speed and acceleration are no great priorities I see no reason the probe couldn't carry a "seed" panel array of considerably greater size than a 7 meter square. It might also be profitable to invest some of that energy in creating a large foil mirror to reflect more sunlight on the panel and speed up the work, depending on how much light your panels can take before it starts to damage them. Or, as a foil mirror could be quite large and light, the probe might already have one with it from the previous replication.

...

Junghalli...

Did you realize that the Genersys solar panel is a solar thermal panel? If you had looked in the materials manifest you'll find silicon strangely absent.

You are correct. I never said a self-replicating probe was a trivial thing to build. By our standards it would require incredible technological prowess. But then, by the standards of the Middle Ages a jet fighter would require incredible technological prowess. I tend not to see "requires incredible technological prowess" to be something that is likely to stop civilizations hundreds or thousands of years more advanced than ours. I realize this sounds like a vague appeal to superior technology, but I see no reason not to assume that what looks very difficult to us won't be quite doable to a civilization of sufficient technological sophistication. Exceptions, obviously, for things that just break physics or any remote degree of practicality, like FTL.

While I agree that it's impossible to know what the future will bring, its precisely that impossibility that's keeping me from saying that the desired AIs are possible, at the very least until we get an idea of the complexity of the required AI.

[Memnon]

Quick tidbit about regenerative braking: depending on the timescales we're talking about here, you can put the charge into capacitors. Since there's vacuum all around you, you can use that as insulation in the capacitor and thus hold the charge for a longish time. I'm not sure how long that is, but surely it's at least a few minutes. That way you can avoid damage to batteries also (but you have to use the charge almost right away - perhaps in more slowly charging the batteries).

[Junghalli]

The TOP500 system supercomputer eats 257 kW at peak processor power (source), so if we assume —for the moment— that the AI can run comfortably on it, the flywheel needed just to run the AI for a century (~811 TJ) is 1.62 million tonnes, and that flywheel will be spun down by the time it gets there. On the bright side, it'll probably keep everything toasty warm inside.

Why on Earth would you want to keep the AI running at full power while the probe is doing nothing but drifting through interstellar space? Especially if you didn't have a nuclear reactor? You would logically run the AI at a tiny fraction of its capacity or shut it down altogether while you were drifting between systems.

Ummm, what? Aren't you here on this comet because you're going to harvest deuterium to get the seed energy you need so you can refine materials? To make more copies of yourself? How can you make more copies of yourself without materials?

I think we've gotten off track. We were talking about exploiting the comet for propellant if you don't have a D-D reactor and are limited to sunlight for your power. If that's the case obviously, you don't need to worry about deuterium. If you do have a D-D reactor you don't need to worry about whether the comet stays in sunlight or not, you can easily work in the outer system. Provided you had the foresight to keep a fuel reserve to run your reactor while you refine more fuel, and given you only need to drag around kilogram quantities to do this you'd be an idiot not to.

Okay, let's say we want to slow Hale-Bopp to have an aphelion of 20 AU within the six months either side of April 1, 1997, when it reached perihelion. With a semi-major axis of 186 AU and a perihelion of 0.914 AU, the velocity at perihelion is 44 km/s. The corresponding velocity with an aphelion of 20 AU is 43 km/s, so we need to have a delta-v 1 km/s over one year. Hale-Bopp's nucleus is about 60 km diameter of pure water, so this is a 1.13e14 tonne monster. The thrust required to affect this change over 1 year is 3.58 TN. If we suppose we want to use 1/4 of the mass to affect this change, we get an exhaust velocity of 721 m/s. This requires a power of 1.29 exowatts. This would require a solar panel 2274 km on a side (assuming that we begin thrusting at the orbit of Mars).

Why would you use a giant comet like Hale-Bopp? It has way more volatiles in it than you'd ever need for propellant, and if you can replicate entirely off comets probably also way more carbon (I confess I haven't looked up the carbon fraction in a typical comet, but even if it's only .1% Hale-Bopp has 1.13e11 tons). If you were persuing this strategy you would logically pick a small snowball, in the range of hundreds of meters across, not a 60 km monster. By some quick napkin calculations a 100,000 ton snowball could have its orbital velocity altered significantly (by 10 km/s) in the space of 1.6 years by a single MAX mass driver listed on the Atomic Rocket engine list page (20,000 newtons of thrust). Said mass driver would require 350 MW. They also have exhaust velocities of 30 km/s, we may not need that much and could possibly trade fuel efficiency for greater thrust. And, of course, there's nothing stopping us from using multiple ones.

Let's say you detect a Hale-Bopp type comet with three months to match speed. If the comet intersects your orbit at 40° while at 1 AU, you need 60.1 km/s delta-v to catch it, which requires 0.0228 m/s² acceleration. If you use a VASIMR (~10 tonnes) at high gear (30 km/s @ 50 N), you need a mass fraction of 13% (87% propellant), and a engine fraction of ~1.5... which is impossible, even if we believe that we consider 87% propellant to be running on empty! A VASIMR at low gear (3 km/s @ 500 N) has a mass fraction of 1.9698e-9 and an engine fraction of 15%, also impossible, even if we believe the ludicrous mass fraction of 1.9698e-9! If we were to use a solar sail, the sail has to be 5.5346299938462e3 m² for each kg of probe mass. A 1000 metric tonne probe has to have a sail 74,000 km on a side. Ludicrous!

60 km/s delta V and .0228 m/s^2 for a spacecraft of significant mass is within the reach of some nuclear thermal gas core designs. Of course, such a design would require uranium, and you'd burn most of propellant getting to the comet, so it'd be pretty pointless. Good luck getting significant quantities of fissionables or precursor materials from a comet too. You'd be much better off prospecting asteroids for subsurface water if you used such a drive.

So granted this approach doesn't make much sense for refuelling.

Uhh... yeah. You realize the pulse laser destroyes the sail to send it out that quickly? The sail gets turned into a plasma, which the distant ship grabs onto with a magnetic field and uses as propulsion. As vaporization is part of these sails' operation, I don't think it applies to larger, nonvaporizing sails, which actually have to stay intact.

OK. Best thing I could find really quickly (I have to be somewhere) is this article which talks about a hypothetical spacecraft with a 1000 km lightsail. The craft would accelerate at .3 G. This paper talks about thermally limited acceleration of a saphire sail of 1000 m/s^2. It doesn't say whether they factored structural strength into it.

For what it's worth, it wouldn't surprise me if a solar sail could withstand high accelerations. Relative to the direction of thrust it already starts out as an incredibly thin pancake of material.

Did you realize that the Genersys solar panel is a solar thermal panel? If you had looked in the materials manifest you'll find silicon strangely absent.

No reason solar thermal is off the table but, yeah, doh. That's what happens when you're in a hurry.

This site estimates 6.4 years to "pay back" the energy invested in manufacturing a solar panel at 12.5% efficiency. A 12.5% efficient solar panel 1 m^2 across generates roughly 171 watts, so that would be 3.45 X 10^10 joules to manufacture it. You'd need a solar panel hundreds of meters across to generate that in a day. They give more calculations below but honestly I just did the above napkin calc because I am in a bit of a hurry.

This is not encouraging, but apparently there's at least one type of solar power system that takes a lot less energy to manufacture, and it is from a study done in 1977 (i.e. technology will be better even today, let alone hundreds or thousands of years from now).

While I agree that it's impossible to know what the future will bring, its precisely that impossibility that's keeping me from saying that the desired AIs are possible, at the very least until we get an idea of the complexity of the required AI.

As for as the AI goes, we know that a computer with human-level intellect which can be built from mostly carbon is possible. I speak of course of the human brain. Of course, for this task the human brain is decidedly sub-optimal, but there's no reason to think it's the best thing an advanced civilization could produce. It's the result of the slapdash design process of evolution, after all.

[Wyrm]

Why on Earth would you want to keep the AI running at full power while the probe is doing nothing but drifting through interstellar space? Especially if you didn't have a nuclear reactor? You would logically run the AI at a tiny fraction of its capacity or shut it down altogether while you were drifting between systems.

First off, that probably is the AI in low-power mode. You saw the "peak processor power" and didn't realize that that was for a modern supercomputer, which can't run any AI we know of even approaching the sophistication we need, so the power requirement I stated would be a horrible lowball.

Secondly, even if it could, it would probably be a good idea to spend some time surveying the system for as long as possible. Even if you do it in 'dumb mode', it's still going to take a hell of a lot of processor power to crunch all the data. You'll also spend some time during the final decade deliberating over the data, making decisions and planning your approach. That is, in 'smart mode'.

Active magnetic bearings take approximately 1.33 W/kg to run, so a 500 kJ/kg flywheel will lose 2.66e-6th of its energy every second, or more concretely, it will run down in 4.3 days. I think we have a serious stumbling block here.

I think we've gotten off track. We were talking about exploiting the comet for propellant if you don't have a D-D reactor and are limited to sunlight for your power. If that's the case obviously, you don't need to worry about deuterium. If you do have a D-D reactor you don't need to worry about whether the comet stays in sunlight or not, you can easily work in the outer system. Provided you had the foresight to keep a fuel reserve to run your reactor while you refine more fuel, and given you only need to drag around kilogram quantities to do this you'd be an idiot not to.

Okay. It'd probably be easier to simply flit about the Kuiper belt to refine your deterium rather than catching comets long-period comets, because the difference in delta-v is non-trivial.

Why would you use a giant comet like Hale-Bopp? It has way more volatiles in it than you'd ever need for propellant, and if you can replicate entirely off comets probably also way more carbon (I confess I haven't looked up the carbon fraction in a typical comet, but even if it's only .1% Hale-Bopp has 1.13e11 tons). If you were persuing this strategy you would logically pick a small snowball, in the range of hundreds of meters across, not a 60 km monster. By some quick napkin calculations a 100,000 ton snowball could have its orbital velocity altered significantly (by 10 km/s) in the space of 1.6 years by a single MAX mass driver listed on the Atomic Rocket engine list page (20,000 newtons of thrust). Said mass driver would require 350 MW. They also have exhaust velocities of 30 km/s, we may not need that much and could possibly trade fuel efficiency for greater thrust. And, of course, there's nothing stopping us from using multiple ones.

Atomic Rockets' MAX scores are pie-in-the-sky optimistic. Specific impulses and thrust should be reduced by 1/3 at the least. At 13.3 kN of thrust, this mass driver will take 2.4 years to do the same job.

For starters, that 10 km/s delta-v comes at the cost of a significant amount of mass. If your goal is to bring a comet with perhelion .95 AU and aphelion 100 AU down to an aphelion within 2 AU for convenient solar flux, then you're talking about a change in velocity better than 43 km/s. At an exhaust velocity of 20 km/s, only 11.6% of the mass will remain after this maneuver, or 11,600 tonnes. Eighty percent of it would be water, so the carbon-bearing portion is only 20% that, or a measily 2328 tonnes of carbon-bearing material — COx, essentially.

Furthermore, what's left isn't going to last long. A 11,600 tonne comet nucleus is a comet about to fizzle out. A comet forms that tail because the volitiles are sublimating off the comet, so it's getting smaller with each swingby of the sun. Comet Halley lost 2.8e14 g of its 2.2e17 g in its last flyby, or about 8.8 tonnes per second. If we scale down the rate of sublimation for our dinky little comet, then our dinky comet evaporates totally in seventeen years.

This may sound like a lot of time to construct another probe, but the most optimistic stellar layovers are measured in millenia for the production of a hundred new probes. If we use this as a boundary condition for the exponential growth function, 100 = exp(1000 yr/β), then β = ~217 yr. That is, the time scale of growth for a new probe is around 217 years. Your comet will be long gone.

That, partially, was the reason I picked a Hale-Bopp type comet. It'll last long enough to do something with.

60 km/s delta V and .0228 m/s^2 for a spacecraft of significant mass is within the reach of some nuclear thermal gas core designs.

If you consider a propellant fraction of 82% to be 'within the reach' of the engine, yes.

So granted this approach doesn't make much sense for refuelling.

Yep.

(i.e. technology will be better even today, let alone hundreds or thousands of years from now).

If you're talking about solar thermal, not necessarily. You're already bumping hips against thermodynamic limits in space. For instance, at 1 AU from Sol, you can't get the hot end much hotter than 393 K. (Why?) The cold end (radiator) must be hot enough to radiate away the waste heat to space. The hot end isn't very hot, and you're not going to get the cold end very cold, and with real devices more realistically modeled with endoreversable efficiencies (ie, η = 1 - √[Tc/Th] is a more realistic model of real-world efficiency), efficiency is not going to be good.

[Junghalli]

Let me take a break from the quote spaghetti for a moment to illustrate a point.

There's at least one kind of Von Neumann we know will work: organic life. Let's build an Astrochicken and see what its capabilities and needs might be.

Our Astrochicken's AI will consist of three redundant brains each the size of a human's, so it will have a mass of 4.2 kg. It will take energy from solar power, entering suspended animation as it travels through interstellar space. In addition to the brain our Astrochicken needs some basic organs like a heart, specialized organs to help it break down the materials in its environment, a variety of scientific instruments (telescopes etc.), specialized systems to allow it to survive in hard vacuum such as a pressure shell and complex system of heat pumps, heating elements, and radiators, and finally some sort of engine (probably a solar sail). Let's give it 1000X the mass of a human, or a mass of 68 tons.

A typical modern Westerner consumes something on the order of 2000 calories a day. Children don't generally eat massively larger portions than adults so this presumably covers energy for growth of child to adult as well as energy for the adult to maintain itself. A calorie is 4.2 KJ, so a human uses 8.4 megajoules per day, which translates to an energy output of 97 watts. At 1000X the mass of a human our Astrochicken will require approximately 100 kilowatts. Assuming it obtains energy from the sun and uses a solar collector that is 10% efficient (comparable to modern solar panels) at 1 AU from a star like Sol it will require a collector with a surface area of 73 m^2, which translates to a square roughly 8.6 meters on a side. This is a relatively pessimistic figure for a number of reasons. First of all, the human body is rather inefficient at making use of food energy; an engineered lifeform could probably do better. Second, the human brain consumes a disproportionate amount of our body's energy and our Astrochicken has a much lower brain to body ratio than we do. Granted, one imagines it may have huge banks of neurons that store mission data, but these would probably be kept mostly "asleep" during this stage of the mission to free up energy for growth. Third, our Astrochicken could probably get away with being a good deal less physically active than a human, instead investing much of that energy in growth. Still, for the sake of ease let's use 100 kilowatts.

Assuming it grows at the same relative rate as a human our Astrochicken can replicate once every 20 years or so. Its body is more-or-less made of CHON, so it should be able to live and grow fairly readily on what can be found in a typical C-type asteroid or dirty snowball, assuming it gets plenty of sunlight, although we may have to do some bioengineering to eliminate some trace elements Earth life needs that may be problematic.

Of course, 20 years is kind of long (although really not too onerous compared to the megayear timespans of the mission), so let's speed things up a bit, make it replicate in 1/100th that time or .2 years. It now needs 10 megawatts of energy instead of 100 kilowatts to fuel its growth. It now needs a solar panel with a surface area of 73,000 m^2, or roughly 270 meters on a side. This may increase its mass above the 68 ton figure (at 1 kg/m^2 the solar panel will mass 73 tons), but this will not increase replication time to more than half a year assuming the above figure. There are natural lifeforms like bamboo that grow at a vastly faster rate than human tissue, so a growth rate dramatically faster than a human is not beyond the possibility of biology.

As you can see, even using some relatively conservative assumptions organic biology, which we know to be possible, could build a fairly effective self-replicating probe as long as we could overcome the engineering challenges of getting it to survive in space. BTW note that at 68 tons this thing is much more massive than the 722 kg Voyager 1 probe, so it probably would have room for a fairly respectable science package. I fully expect a very advanced civilization will probably be able to do quite a bit better than organic biology.

First off, that probably is the AI in low-power mode. You saw the "peak processor power" and didn't realize that that was for a modern supercomputer, which can't run any AI we know of even approaching the sophistication we need, so the power requirement I stated would be a horrible lowball.

That's missing the point. That power usage is for a modern supercomputer; why are you going to need to run any kind of advanced computer for 100 years while you're doing nothing but drifting through empty space? Why run any kind of computer when you're in a radically energy-poor totally empty environment, for that matter? The only reason you'd conceivably need it is meteorite avoidance, and I'm pretty sure significant size meteorites are very, very rare in interstellar space (I haven't found any figures but you're talking an enormous area for any potential impactors to disperse over), while a whipple shield or armor plate can handle the odd sand grain. If you are worried about significant-size impactors a passive defense system like a dust shield is probably going to be easier than trying to keep a computer running 24/7/365 for centuries in an energy-impoverished environment, unless you have a nuclear reactor in your probe. And finally, even if you do need an active avoidance system, why would it have to be as powerful as a modern supercomputer? All it has to be able to do is recognize when something is likely to hit the probe and give a short tap on the engine to move out of the way.

Secondly, even if it could, it would probably be a good idea to spend some time surveying the system for as long as possible. Even if you do it in 'dumb mode', it's still going to take a hell of a lot of processor power to crunch all the data. You'll also spend some time during the final decade deliberating over the data, making decisions and planning your approach. That is, in 'smart mode'.

If you need external energy to function decision-making is probably something that would best be left until after you're close enough to the target star to be able to run your computers off its energy. You might lose some probes this way by having them enter systems that don't have the raw materials they need to replicate or refuel but you can compensate for this by having them reproduce multiple times when they get the opportunity. The biggest problem with this might be capture by hostile aliens but honestly the probe isn't going to be particularly detectable as it drifts unpowered through the outer system, and if you're really paranoid you can set up a relay system for the probes' transmissions so you don't have to put the location of your homeworld in their memory banks. Not that they're realistically likely to be able to read your totally alien computer language anyway (or at least I don't think so, but my knowledge of how computer languages work is minimal)

Active magnetic bearings take approximately 1.33 W/kg to run, so a 500 kJ/kg flywheel will lose 2.66e-6th of its energy every second, or more concretely, it will run down in 4.3 days. I think we have a serious stumbling block here.

It would probably be best to design a solar powered probe so that it can drift completely unpowered through interstellar space. It can be awakened by sufficient energy hitting the solar panels.

Atomic Rockets' MAX scores are pie-in-the-sky optimistic. Specific impulses and thrust should be reduced by 1/3 at the least. At 13.3 kN of thrust, this mass driver will take 2.4 years to do the same job.

You'd probably need multiple ones to do the job. The point is it's only undoable if you insist on using a huge comet.

For starters, that 10 km/s delta-v comes at the cost of a significant amount of mass. If your goal is to bring a comet with perhelion .95 AU and aphelion 100 AU down to an aphelion within 2 AU for convenient solar flux, then you're talking about a change in velocity better than 43 km/s. At an exhaust velocity of 20 km/s, only 11.6% of the mass will remain after this maneuver, or 11,600 tonnes. Eighty percent of it would be water, so the carbon-bearing portion is only 20% that, or a measily 2328 tonnes of carbon-bearing material — COx, essentially.

If you need more you can use a bigger comet. There's plenty of range between a very small 100,000 ton comet (which wouldn't even be 100 meters across) and a giant like Hale-Bopp.

Furthermore, what's left isn't going to last long. A 11,600 tonne comet nucleus is a comet about to fizzle out. A comet forms that tail because the volitiles are sublimating off the comet, so it's getting smaller with each swingby of the sun. Comet Halley lost 2.8e14 g of its 2.2e17 g in its last flyby, or about 8.8 tonnes per second. If we scale down the rate of sublimation for our dinky little comet, then our dinky comet evaporates totally in seventeen years.

Solvable by building a sunshade between the comet and the sun. For bonus points make your sunshade out of solar panels so it can double as your power plant.

If you're talking about solar thermal, not necessarily. You're already bumping hips against thermodynamic limits in space.

The page in question was talking about photovoltaic panels, for which there is much room for technological improvement.

[Junghalli]

Doh! Again I seem to have messed up in one of my calculations: 100 kilowatts would require 730 m^2 of solar panel or a roughly 27 X 27 meter square panel.

[Wyrm]

(Skipping astrochicken for now...)

That's missing the point. That power usage is for a modern supercomputer; why are you going to need to run any kind of advanced computer for 100 years while you're doing nothing but drifting through empty space? Why run any kind of computer when you're in a radically energy-poor totally empty environment, for that matter? The only reason you'd conceivably need it is meteorite avoidance, and I'm pretty sure significant size meteorites are very, very rare in interstellar space (I haven't found any figures but you're talking an enormous area for any potential impactors to disperse over), while a whipple shield or armor plate can handle the odd sand grain. If you are worried about significant-size impactors a passive defense system like a dust shield is probably going to be easier than trying to keep a computer running 24/7/365 for centuries in an energy-impoverished environment, unless you have a nuclear reactor in your probe. And finally, even if you do need an active avoidance system, why would it have to be as powerful as a modern supercomputer? All it has to be able to do is recognize when something is likely to hit the probe and give a short tap on the engine to move out of the way.

Impactors have nothing to do with it. Here's the reasons to keep the computer active:

(a) You need to keep the lights on. Spreading to all systems in the galaxy requires fast STL drives, going small but significant fractions of c. They take decades to accelerate and decelerate. If you want to be captured by the target system's gravity, you need to start decelerating decades out. If you delay until the solar flux is sufficient for a solar panel of reasonable size to reactivate you, you're already on your way to the next star.

(b) You need to keep warm. Like us, electronics don't like being too cold. A probe that is space-cold is dead. It's one of the reasons why the Mars probes weren't simply rebooted every morning.

If you need external energy to function decision-making is probably something that would best be left until after you're close enough to the target star to be able to run your computers off its energy. You might lose some probes this way by having them enter systems that don't have the raw materials they need to replicate or refuel but you can compensate for this by having them reproduce multiple times when they get the opportunity.

See (a) above. Unless you start decelerating decades out, your success rate is zero percent.

Active magnetic bearings take approximately 1.33 W/kg to run, so a 500 kJ/kg flywheel will lose 2.66e-6th of its energy every second, or more concretely, it will run down in 4.3 days. I think we have a serious stumbling block here.

It would probably be best to design a solar powered probe so that it can drift completely unpowered through interstellar space. It can be awakened by sufficient energy hitting the solar panels.

See (a) above.

If you need more you can use a bigger comet. There's plenty of range between a very small 100,000 ton comet (which wouldn't even be 100 meters across) and a giant like Hale-Bopp.

Yes, but comets like Hale-Bopp are easier to find early, because they are so big and start shedding significant amounts of gas early, and their nuclei are more easily seen (and given they are amongst the darkest objects in the universe, you'll need as much light as you can get). That means you can hop onto them more easily and do your business. A small comet you won't see until it's already to late to hop onto it.

Solvable by building a sunshade between the comet and the sun. For bonus points make your sunshade out of solar panels so it can double as your power plant.

Okay, but that solar-panel sunshade needs refridgeration. Otherwise, it heats up and begins radiating itself. (Also, their efficiency goes down, which worsens the problem.) Sunshades in space usually consist of reflective material.

If you're talking about solar thermal, not necessarily. You're already bumping hips against thermodynamic limits in space.

The page in question was talking about photovoltaic panels, for which there is much room for technological improvement.

The paper talks about how thin you can make the panels, and there's a limit to how thin they can be made. The cells have to be thick enough so that an incoming photon has a good chance of knocking an electron into the n-type region, or knocking a hole into the p-type region. If the cell is too thin, the photon strikes the back contact and is wasted. The thinness of the cell has an absolute lower limit of about 100 micrometers (.1 mm) for silicon.

Also, the best types of cells are not conducive to simple manufacture, nor made with the common elements. For instance, the best top contacts are made of gold, which is only about a half-order of magnitude more plentiful than thorium. The best materials to make a solar cell are gallium arsenide and gallium indium arsenide. However, both are about five to seven orders of magnitude less abundant than silicon. The best class of cells are multijunction, where multiple layers of differing semiconductor (with different band gap energies) are deposited atop each other. The problem with these cells is that different semiconductors have different lattice parameters, and if they differ too much, the efficiency drops precipitously due to the abundance of lattice defects between junctions. There is a derth of any mention of using silicon in a multijunction photovoltaic cell. (Guess its lattice parameters are too out of whack.)

[Junghalli]

(a) You need to keep the lights on. Spreading to all systems in the galaxy requires fast STL drives, going small but significant fractions of c. They take decades to accelerate and decelerate. If you want to be captured by the target system's gravity, you need to start decelerating decades out. If you delay until the solar flux is sufficient for a solar panel of reasonable size to reactivate you, you're already on your way to the next star.

These sorts of drive systems are pretty much out of the question if we're talking the sort of ship that requires solar power to function. The solar panel would need to be infeasibly huge and massive, even if you ignore the problem that at reasonable acceleration it's going to be way out of the solar system before it can build up that kind of speed. You need a fusion drive or better (RAIR, Bussard scramjet) to get significant fractions of c. That implies the capability to give the ship on onboard reactor, and if your ship is self-replicating obviously your fusion system is practical to wilderness refuel, making this whole discussion mute.

Keeping power ticking over at reasonably low rates with fusion over centuries is doable. To keep a 50% efficient D-D reactor ticking over for 100 years at 1 MW would require 10.2 kg of deuterium.

If you are talking the sort of drives that can practically function off solar power you are going to be restricted to .001-.01 c and timescales of 5-100 myr to survey the galaxy. Such speeds do not require years of acceleration (save with very low thrust drives). Trying to explore the galaxy with that sort of technology is a game for the very, very patient. This is the scenario we've been talking about with all this talk of utilizing solar power. If you want to survey the galaxy at > .01 c you're going to need nuclear drives.

(b) You need to keep warm. Like us, electronics don't like being too cold. A probe that is space-cold is dead. It's one of the reasons why the Mars probes weren't simply rebooted every morning.

This is a legitimate concern. There are two possible solutions. One is to design electronics that can withstand deep cold. Maybe not something we can do, but a civilization with superior materials technology might be able to do it. The other is to use nuclear power, either fusion or fission breeder reactors.

Yes, but comets like Hale-Bopp are easier to find early, because they are so big and start shedding significant amounts of gas early, and their nuclei are more easily seen (and given they are amongst the darkest objects in the universe, you'll need as much light as you can get). That means you can hop onto them more easily and do your business.

This is true. The Hubble Telescope has detected comets as small as 340 meters (link), but only at a small heliocentric distance if I read the paper correct.

At any rate, you only need to follow this kind of strategy if you have to use solar power and need lots of water. There are lots of way out of that strategy. For propellant instead of using something that needs volatiles you can use a mass driver and feed it bullets forged from asteroidal metals. You will likely be able to obtain volatiles from some asteroids as well. You can try to build a nuclear reactor capable of wilderness refueling and then you won't need to worry about sun.

[Memnon]

Can I jump on the chicken? :P

Assuming it obtains energy from the sun and uses a solar collector that is 10% efficient (comparable to modern solar panels)

The theoretical maximum efficiency for photosynthesis is 25% on Earth. However, since this thing is going to be drifting quite a lot, and the distance is far between stars, it seems unlikely that it could even get 10%. That's assuming photosynthesis, however.
Also, photosynthesis is optimized for certain wavelengths of light which other suns would have in different amounts.

Assuming it grows at the same relative rate as a human our Astrochicken can replicate once every 20 years or so. Its body is more-or-less made of CHON, so it should be able to live and grow fairly readily on what can be found in a typical C-type asteroid or dirty snowball, assuming it gets plenty of sunlight, although we may have to do some bioengineering to eliminate some trace elements Earth life needs that may be problematic.

You will need magnesium for photosynthesis and iron for transport and calcium or something else for structure and phosphorous for DNA and plasma membranes and so on. It's not that easy to get rid of trace elements.
And cell respiration using the electron transport chain, one of the most efficient systems to get food to energy, needs some of those elements also. Even basic things like cell signaling using the phosphorylation cascade would need to be reworked.

There are natural lifeforms like bamboo that grow at a vastly faster rate than human tissue, so a growth rate dramatically faster than a human is not beyond the possibility of biology.

However, bamboo is also simpler than a chicken, and is porous/depends on a constant source of water. So that kind of growth rate is dependent on abundance.

As you can see, even using some relatively conservative assumptions organic biology, which we know to be possible, could build a fairly effective self-replicating probe as long as we could overcome the engineering challenges of getting it to survive in space. BTW note that at 68 tons this thing is much more massive than the 722 kg Voyager 1 probe, so it probably would have room for a fairly respectable science package. I fully expect a very advanced civilization will probably be able to do quite a bit better than organic biology.

Essentially, some of the assumptions amount to creating an almost new kind of life - some sort of super-prokaryote or something. That civilization would have to do things like redesign DNA, the way the brain works (sodium and K are hugely important), and so on. In other words, this conjecture is not really that conservative. XD

[Junghalli]

The theoretical maximum efficiency for photosynthesis is 25% on Earth. However, since this thing is going to be drifting quite a lot, and the distance is far between stars, it seems unlikely that it could even get 10%. That's assuming photosynthesis, however.

It would enter suspended animation during the interstellar journeys. This is something some bacteria can do readily, and extreme hibernation strategies can also be seen in macroscopic lifeforms like resurrection plants and wood frogs, so while centuries of suspended animation for a macroscopic creature is probably unprecedented it is building on capability that already exists in naturally evolved lifeforms.

Also, photosynthesis is optimized for certain wavelengths of light which other suns would have in different amounts.

True. If we're using biological photosynthesis it might be a good idea to design several alternate photosynthetic pathways well-suited to different kinds of stars into the organism's genome and have it be able to activate one or the other depending on environment.

You will need magnesium for photosynthesis and iron for transport and calcium or something else for structure and phosphorous for DNA and plasma membranes and so on. It's not that easy to get rid of trace elements.
And cell respiration using the electron transport chain, one of the most efficient systems to get food to energy, needs some of those elements also. Even basic things like cell signaling using the phosphorylation cascade would need to be reworked.

According to this chart magnesium and (red) phosphorus should both be readily obtainable from a typical C-type asteroid at 120 million and 1.1 million ppb (parts per billion), respectively. Calcium is available at 11 million ppb. Iron, of course, is extremely common at 220 million ppb. Other elements I know to be necessary for life such as sodium (5.6 million ppb), potassium (710K ppb), and nitrogen (1.4 million ppb) are also readily available. Carbon is 15 million ppb and the main site suggests the possibility of water ice being up to 10% of the asteroid's mass. Even if there is no water ice C-type asteroids would probably have hydrated silicates, albeit that would require extra processing to get out the water, which for an organic creature which is mostly water would be a real pain. It looks like a C-type asteroid should have all or most of the basic elements necessary for life. Granted that's not an exhaustive list of all the elements we need.

However, bamboo is also simpler than a chicken, and is porous/depends on a constant source of water. So that kind of growth rate is dependent on abundance.

True.

Essentially, some of the assumptions amount to creating an almost new kind of life - some sort of super-prokaryote or something. That civilization would have to do things like redesign DNA, the way the brain works (sodium and K are hugely important), and so on. In other words, this conjecture is not really that conservative. XD

True if you want to work on a dirty snowball (I think, I admit I haven't looked up comet element abundances), but you should be able to get most of the elements terrestrial life needs from a C-type asteroid. And admittedly by conservative I was more referring to assumptions like "you won't be able to make the metabolism more efficient than a human's" and "it will need the same amount of calories as a human for no faster growth even though it's probably a good deal less physically active" and "its solar panels will be 10% efficient".

[Junghalli]

Incidentally, some material I turned up on the mechanics extracting water from clay minerals in C-type asteroids: link.

The water in carbonaceous asteroids is loosely bound in clay minerals, and can be released by gentle heating. A solar furnace that can produce temperatures of 250 to 300 C suffices to extract almost all of the water in a typical carbonaceous meteorite. Liquid water can be readily condensced from the emitted gasses.

A solar furnace is basically just a big parabolic mirror like this thing (which can achieve 10 times the necessary temperature). One imagines our Astrochicken might synthesize it biochemically from some sort of plastic, or one could use polished metal or glass (made from asteroidal silicates). Since you're setting it up in zero gee it can be made very thin and light.

[Wyrm]

This is a legitimate concern. There are two possible solutions. One is to design electronics that can withstand deep cold. Maybe not something we can do, but a civilization with superior materials technology might be able to do it. The other is to use nuclear power, either fusion or fission breeder reactors.

I'm going to tackle this point first because it's important, and will have consequences for the first point you raised. If you're talking about any kind of semiconductor, there is a minimum temperature where the semiconductor will be able to do its job. This is because the conductivity of a semiconductor is in part due to thermal excitations of electrons into the conduction band, or holes into the valence band. If the temperature drops too far, the semiconductor stops conducting so well and loses its ability to have its conductance fine tuned as is necessary for applications.

With that in mind, there are two important considerations for a semiconductor chip: thermal stress and thermal shock. Thermal stress is induced by the difference in thermal expansion coefficients in the various materials. Silicon is particularly vulnerable, owing to its low toughness, low thermal conductivity, and high thermal expansion coefficient. Even if we suppose that the thermal stress can be reduced by proper choice of materials, thermal shock concerns are going to keep the rate of warming low.

A silicon chip is not a superconducting magnet, but let's take the LHC's recent woes as an example of warming a very cold piece of technology to more earthly temperatures, only instead of warming from operating temperature to maintenence temperature, we warm from cold storage temperature to operating temperature. Let's suppose that this operating temperature is -40 °C (233.15 K). From 4 K, this is a 229.15 K change in temperature. If we assume that the computer can be approximated by 1 tonne of silicon, this gives us a heat-up power of 20.5 W over three months. If we start at 3 AU using 15% efficient cells, we need 2,000 m² solar panel area to gather the power that far out. Three months later, we can start to get to work.

But we came from outside the system! If we came in at 0.001 c (~300 km/s), then the semi-major axis for a hyperbolic trajectory is 1.474 million km. At 3 AU, the orbital velocity is still about 300 km/s. Let's suppose the perihelion grazes the sun at .1 AU. Then the eccentricity of the orbit is ~11.15, which basically means that the path is essentially straight. To transverse 6 AU at a speed of 300 km/s takes about thirty-five days. You won't even have a chance to thaw.

Furthermore, if this is your source of thawing power, then you need to keep it open as you travel, else you can't catch the light and thaw out to even command it to open. Since it must catch the light from the target star, you can't put it behind a whipple shield. Also, you cannot ensure that the spin on your probe is really zero rev/sec, which means you might be facing the wrong way when you finally near the system. You need a means to keep your panel pointed at the star, which requires you to... keep the lights on.

As to appealing to advanced materials, I am weary to do that, especially given the topic that we are discussing: We're not really talking about how to create a von Neumann probe, but rather why we haven't already been visited and/or wiped out by one. Although we are explicitly designing our probe to not interfere with planetary life, we cannot expect other civilizations to be this courteous.

These sorts of drive systems are pretty much out of the question if we're talking the sort of ship that requires solar power to function. The solar panel would need to be infeasibly huge and massive, even if you ignore the problem that at reasonable acceleration it's going to be way out of the solar system before it can build up that kind of speed. You need a fusion drive or better (RAIR, Bussard scramjet) to get significant fractions of c. That implies the capability to give the ship on onboard reactor, and if your ship is self-replicating obviously your fusion system is practical to wilderness refuel, making this whole discussion mute.

"Moot." The word you're looking for is "moot."

Keeping power ticking over at reasonably low rates with fusion over centuries is doable. To keep a 50% efficient D-D reactor ticking over for 100 years at 1 MW would require 10.2 kg of deuterium.

Why 1 MW? That figure assumes that the plasma really is fusing at a rate of 1 MW to sustain itself. This figure is very hard to take seriously.

Let's do a back-of-the-envelope calculation. Suppose we have a spheromak reactor confining a plasma of fusing deuterium. In order to remain in equilibrium, the heat loss of the plasma to various sources must be equal to that of heat gain via self-heating through fusion and external heating:

Ploss = Pself + Pext

Ploss is characterized by the confinement time, τth, via

τth = W/Ploss

where W is the energy content of the plasma. This energy content is equal to

W/V = 3 ne kT

where ne is the electron density of the plasma (equal to the nuclear density) and T is the temperature. The product ne τth must be greater than the Lawson criterion for D-D fusion, which is about

L = 4.5e21 s/m³

Because of the Lawson criterion we can calculate the minimum electron density via the confinement time, which we can assume for now hovers at around τth = 9.6 s. The minimum ne is 4.6875e20 m⁻³.

The power loss, Ploss, is equal to

Ploss = W/τth = 3 ne VkT/τth

Since kT for D-D fusion is at least 35 keV and the volume of the plasma is [4π/3] r³, Ploss comes to 430 MW.

The amount of self-heating the plasma gets is governed by

Pself = ƒch Pfus

where ƒch is the fraction of the energy that remains in charged particles (for D-D fusion, 82%) and Pfus is the amount of power actually generated in the fusion reaction.

The external power added to the plasma by the plant is governed by

Pext = Pfus/Q

where Q is the fusion energy gain factor. In waiting mode, the Q factor for the is approximately 20 by the following derivation

Q = Pfus/Pext = 1/[ηheat ƒrec ηelec (1-ƒch)]

where ηelec = 0.4 is the efficiency of converting the energy that comes to us via neutrons to electricity, ƒrec = 1 (idling mode) is the fraction recirculated to the plasma, and ηheat = 0.7 is the efficiency of electricity heating the plasma.

Putting it all together, we get

Pfus = Ploss/[ƒch + 1/Q] = 494 MW

The average energy per fusion in this cycle is 12.5 MeV. Therefore, 494 MW requires the consumption of 4.097e-4 moles/s, or 2.59 tonnes in 100 years.

If you are talking the sort of drives that can practically function off solar power you are going to be restricted to .001-.01 c and timescales of 5-100 myr to survey the galaxy. Such speeds do not require years of acceleration (save with very low thrust drives). Trying to explore the galaxy with that sort of technology is a game for the very, very patient. This is the scenario we've been talking about with all this talk of utilizing solar power. If you want to survey the galaxy at > .01 c you're going to need nuclear drives.

I think we're going to need nuclear drives, period. Otherwise, even 0.001 c is hopelessly optimisitc, because solar panels don't have a very good alpha (W/g). Suppose that our array provides enough juice to power our electric WHATEVER drive out to 3 AU. Outside that radius, the WHATEVER quits working. Suppose that our mass ratio is 1/20. That is, 95% of us is propellant by mass. Ignoring orbital mechanics bollocks, we want to make it out with a bit less than 300 km/s speed. This demands by the Tsiolkovski equation:

∆v = -u log R

which gives us u = ~100 km/s. The equation relating distance traveled, mass ratio, exhaust velocity, thrust and initial mass is

D = [Mu²/F][R log R - R + 1]

We already know D = 6 AU, R = 1/20 and u = ~100 km/s, so F/M = a0 = 8.94e-3 m/s². That is, each kilogram of wet spacecraft mass must produce 8.94e-3 N of thrust, or each kilogram of dry mass must produce 0.179 N of thrust. The minimum power required from each kilogram is 8.95 kW. If a 15% silicon solar panel must be .1 mm thick, then each square meter of panel is 232.9 g/m². At 3 AU, the square meter collects 10.4 mW. Thus, a solar panel sufficient size to power the WHATEVER drive is 859,750 m²/kg, or 200.2 tonnes per kilogram.

Ummm... no.

[Junghalli]

A brief digression from self-replicators. How hard would it be to explore the galaxy without them?

There are several hundred billion stars in the galaxy, let's say 200 billion probes. Let's assume each of them is 1000 tons. That's 200 trillion tons. PERMANENT estimates a 100 meter asteroid might have a mass of 3 million tons (ref), which gives a density of 5.7 tons/m^3 assuming a spherical shape. Our required 2 X 10^14 tons would be supplied by a spherical asteroid perhaps 44 km across. Assuming 1/8 of the asteroid's material is actually suitable for constructing the probes we'd need to deconstruct an 88 km asteroid. Assuming a mere 1/1000th the material is suitable we'd need a 440 km asteroid, or basically we'd need to take apart an asteroid about half as big as Ceres. Over 200 known asteroids are over 100 km, so the lower estimates could be achieved by deconstructing any one of these. Essentially, the project is well within the scope of the solar system's resources, and to a very advanced spacefaring civilization with self-constructing machinery and capable AI it should be doable.

Of course, 1000 tons may be an optimistic estimate. Although personally if anything I lean toward it being a pessimistic one. Such an advanced civilization could probably build highly miniaturized and lightweight probes and mount them on advanced solar sails. If they could achieve a velocity of .01 c the galaxy could be surveyed in 5-10 million years; a long time in our mayfly terms, but perhaps not unacceptable to immortal beings with the vastly increased time horizons that entails, and an eyeblink compared to its age. You could even easily cut down material needed since each probe could visit multiple systems as they need no internal propellant or fuel, at the expense of a slight increase in mission time. Of course, such minimalistic visitors pose no problem for the Fermi Paradox. And, of course, there is the challenge of making a machine that can last megayears, which is no small accomplishment.

If we assume that the computer can be approximated by 1 tonne of silicon, this gives us a heat-up power of 20.5 W over three months. If we start at 3 AU using 15% efficient cells, we need 2,000 m² solar panel area to gather the power that far out. Three months later, we can start to get to work.

Solar radiation intensity at 1 AU is 1366 watts so by my calculations a 15% efficient 1 m^2 solar panel at 4 AU would generate 12.8 watts. 2 m^2 of panels would suffice to generate 20.5 watts. A 15% efficient 2000 m^2 solar panel at 4 AU would generate 25.6 kilowatts. That power input would take 1.8 hours to raise the temperature of 1 ton of silicon (specific heat 712 j/kg) 229.15 degrees, assuming 100% efficiency and no energy losses. I must be misinterpreting your statement. Is the 3 month waiting period due to the need to not damage the material?

You mention silicon as particularly vulnerable to thermal stress, what about other materials, like carbon?

Since it must catch the light from the target star, you can't put it behind a whipple shield.

You could have a mechanism to fold up or move aside the whipple shield automatically activated by a smaller solar panel (perhaps shielded behind a whipple shield of clear or semi-clear material).

Also, you cannot ensure that the spin on your probe is really zero rev/sec, which means you might be facing the wrong way when you finally near the system. You need a means to keep your panel pointed at the star, which requires you to... keep the lights on.

Or you could include a mechanism to move your solar panels around if they're facing the wrong way. That shouldn't take much power; as long as your solar panel isn't pointed completely in the opposite direction it's supposed to be you can move them to a better position (and you can add redundant panelling on the opposite side of the panel just in case the worst case scenario happens). Or do a few taps with low power cold gas thrusters to just rotate the probe. This requires you to have a safety margin between minimum awakening energy and energy needed to actually do much useful stuff, but that should be doable.

As to appealing to advanced materials, I am weary to do that, especially given the topic that we are discussing: We're not really talking about how to create a von Neumann probe, but rather why we haven't already been visited and/or wiped out by one. Although we are explicitly designing our probe to not interfere with planetary life, we cannot expect other civilizations to be this courteous.

Nevertheless, arguments that rely on civilizations thousands of years more advanced than us being able to do little better than our present technology strike me as somewhat unconvincing, unless there's a very good reason that nothing much better than present technology could be possible (i.e. we're already bumping up against fundamental physics limitations in that area).

Pfus = Ploss/[ƒch + 1/Q] = 494 MW

The average energy per fusion in this cycle is 12.5 MeV. Therefore, 494 MW requires the consumption of 4.097e-4 moles/s, or 2.59 tonnes in 100 years.

That is hardly infeasible. 2.59 tons is not huge compared to the likely mass of an interstellar probe, and it requires processing perhaps 135,000 tons of water to get it. Even a very small KBO will be able to supply that easily.

I think we're going to need nuclear drives, period. Otherwise, even 0.001 c is hopelessly optimisitc, because solar panels don't have a very good alpha (W/g).

With future technology we may be able to produce much lighter and more efficient solar panels, but yes, even .001 c is problematic with solar powered electrical, thermal, or mass driver rockets. Advanced solar sails might be able to do up to .04 c with a .019 AU approach to the sun (ref - slide 14), but that would require fairly formidable engineering and they don't say the payload fraction - I wouldn't be surprised if it was assuming a very tiny probe, which would have a hard time being self-replicating unless it was of the "grows like a plant" variety. The near-term solar sail design by the same paper is capable of around .002 c.

[Wyrm]

Solar radiation intensity at 1 AU is 1366 watts so by my calculations a 15% efficient 1 m^2 solar panel at 4 AU would generate 12.8 watts. 2 m^2 of panels would suffice to generate 20.5 watts. A 15% efficient 2000 m^2 solar panel at 4 AU would generate 25.6 kilowatts. That power input would take 1.8 hours to raise the temperature of 1 ton of silicon (specific heat 712 j/kg) 229.15 degrees, assuming 100% efficiency and no energy losses. I must be misinterpreting your statement. Is the 3 month waiting period due to the need to not damage the material?

Yes, and I admit that I screwed up the calculation somewhere. At 3 AU out, only a .9 m² solar panel is required.

You mention silicon as particularly vulnerable to thermal stress, what about other materials, like carbon?

Carbon/diamond can be a semiconductor and is excellent in terms of thermal conductivity, has a moderately better coefficient of expansion, and amongst the toughness gemstones known. However, it's band gap is about five times larger than silicon, which has many consequences for its ability to be a switch, and practically none of them are good. And forget using them in solar panels. A band gap that large means that the longest waves they can catch is 225 nm, which is in the ultraviolet region and way past the maximum of a sun-like star (it's why diamond is clear) — it may harvest energy only from the ever-dwindling tail of the spectrum. A carbon solar panel would be much worse than a silicon one.

You could have a mechanism to fold up or move aside the whipple shield automatically activated by a smaller solar panel (perhaps shielded behind a whipple shield of clear or semi-clear material).

If you could protect a solar panel with a clear whipple shield, why not cover the primary panel? Well, a whipple shield works by shocking the impactor into disintegrating, and that works by putting a massive enough barrier in the way such to punch through needs a great impulse. Furthermore, the parts of the whipple shield that are not directly smacked into have to survive the experience largely intact, which basically means ductile failure. Clear things are not noted for their ductility.

Also, clear things have the annoying property that they reflect a certain portion of light as well as transmit it. Also, for a whipple shield, there is an angle where the plane completely reflects the incident light. To combat this, an thin, sub-wavelenth coating is used to cancel out this reflection property and allow mostly transmittence. However, these coatings are vulnerable to erosion. Over the course of a light year traveling at 30 km/s, the interstellar medium imparts about 1.4 kJ over each square centimeter. This is plenty of erosive effect, and enough to ruin any antireflective coating —or indeed any smooth, optically active surface— turning it into a pitted, opaque mess. If an impactor hits, then you get a spray of material coating the interior clear part to similar effect.

This will also happen to the coating of the solar panel. I think that this is further proof that a probe can't simply reactivate by being warmed by sunlight.

Or you could include a mechanism to move your solar panels around if they're facing the wrong way. That shouldn't take much power; as long as your solar panel isn't pointed completely in the opposite direction it's supposed to be you can move them to a better position (and you can add redundant panelling on the opposite side of the panel just in case the worst case scenario happens). Or do a few taps with low power cold gas thrusters to just rotate the probe.

In order for this scheme to work, the panels have to produce enough power for the logic to realize it's out of normal orientation, be specific enough to tell the probe how it is out of orientation, and then drive the actuators necessary to reorient itself. The problem is that solar panels are Lambertian, and there is reflection from the surface when the panel is rotated away from normal. Both of these effects have maxima at parallel orientation. At 57° away from normal, the amount of light getting to the solar cell is only 50% of what it is at normal. There is 22% chance that the probe will find itself in this orientation. There is only 41% chance that the probe will find itself within 80° from normal, where it will only get 10% maximum power availible at this distance.

Also, the logic has to be extremely simple and the actuator low power for this to work, because otherwise, you'd have to spend three months warming it up before you can do anything.

This requires you to have a safety margin between minimum awakening energy and energy needed to actually do much useful stuff, but that should be doable.

How much energy do you propose to use doing this?

Nevertheless, arguments that rely on civilizations thousands of years more advanced than us being able to do little better than our present technology strike me as somewhat unconvincing, unless there's a very good reason that nothing much better than present technology could be possible (i.e. we're already bumping up against fundamental physics limitations in that area).

Except that kind of argumentation forces you to the conclusion that we are completely and utterly alone in the galaxy, and the fact that we haven't been rolled over already is the proof.

That is hardly infeasible. 2.59 tons is not huge compared to the likely mass of an interstellar probe, and it requires processing perhaps 135,000 tons of water to get it. Even a very small KBO will be able to supply that easily.

First off, messed up there. There are 4.097e-4 moles/s of reactions, but each reaction eats up two deuterons, so 5.17 tonnes/cy.

Secondly, as expensive as they are, solar panels still should be much easier to process and faster to create than the 135,000 tonnes of water you need to process. Heavy water fractionation takes time. It takes a certain amount of time to start up, and an onboard plant has a certain throughput no matter how much power you have.

Thirdly, the fusion plant used in the example was just that, an example. The plant you take along will by necessity be lower quality than what you can make back home. We're not just talking about wilderness refueling here, but wilderness fabrication. Getting conductors of sufficient properties is going to be a task. It is not guaranteed that the probe engine's idling characteristics will be this good, for any given power characteristics.

With future technology we may be able to produce much lighter and more efficient solar panels,

With a single-junction silicon solar panel, a .1 mm thick panel with convesion efficiency 10-20% is about the best you can hope for. Any thinner and the silicon becomes progressively transparent. The 10-20% efficiency is due to the fact that the band gap of silicon is 1.1 eV. Diamonds would be worthless as solar panels, and the other elements we can make solar panels out of are quite rare, so forget about multijunction solar panels of any stripe.

This is not to say that it can't be used as an interrim power source. Silicon is cheap to manufacture in bulk, comparatively, and even wilderness fabrication you can expect a much higher acceptance rate than for a fusion plant (which probably requires some special materials with high purity and quality).

[Junghalli]

Carbon/diamond can be a semiconductor and is excellent in terms of thermal conductivity, has a moderately better coefficient of expansion, and amongst the toughness gemstones known. However, it's band gap is about five times larger than silicon, which has many consequences for its ability to be a switch, and practically none of them are good. And forget using them in solar panels. A band gap that large means that the longest waves they can catch is 225 nm, which is in the ultraviolet region and way past the maximum of a sun-like star (it's why diamond is clear) — it may harvest energy only from the ever-dwindling tail of the spectrum.

What about carbon nanotubes? According to Wikipedia they can be made semiconducting with a very small band gap.

Also, clear things have the annoying property that they reflect a certain portion of light as well as transmit it. Also, for a whipple shield, there is an angle where the plane completely reflects the incident light. To combat this, an thin, sub-wavelenth coating is used to cancel out this reflection property and allow mostly transmittence. However, these coatings are vulnerable to erosion. Over the course of a light year traveling at 30 km/s, the interstellar medium imparts about 1.4 kJ over each square centimeter. This is plenty of erosive effect, and enough to ruin any antireflective coating —or indeed any smooth, optically active surface— turning it into a pitted, opaque mess. If an impactor hits, then you get a spray of material coating the interior clear part to similar effect.

This will also happen to the coating of the solar panel. I think that this is further proof that a probe can't simply reactivate by being warmed by sunlight.

Hmm, what about using temperature change as a trigger? Maybe have a vessel in the probe filled with an ice or fluid that boils at very low temperature, and run a pipe into it with a miniature turbine in it. As the probe draws close to the target star (say, passing the orbit of Saturn) the sun warms it enough to boil the substance, which turns the turbine, which charges the mechanism that automatically makes the solar panels unfurl. Assuming the probe approached the target star at an angle so it isn't completely in the shadow of the forward shield that should work. The big problem I can see with that is surveying red dwarf star systems, where the probe would have to get very close to the star to be warmed appreciably, leaving it little time to slow down.

Another possibility is some kind of internal clock. It obviously can't need active power. Perhaps a radioactive decay clock? Although that would require radioactive materials. Or maybe a mechanical timer? Something along the lines of a very slow-flowing hourglass with some kind of pressure-sensitive mechanical trigger at the bottom (obviously it couldn't be exactly that, because that would need gravity to work, but you get the general idea). Hmm, would it be possible to build a wind-up mechanical clock that takes decades to centuries to run down? Since we were discussing a biological probe earlier, perhaps have a flask with some living tissue in it that metabolizes a nutrient over a very slow and predictable rate, and when the nutrient is depleted automatically triggers unfurling of the solar panels through some means.

To return to the Astrochicken idea, one possibility that occurs to me is to have the whipple shields themselves made out of thin slices of photosynthetic cells. In deep space they form spores covered with light-sensitive receptors that cause them to de-sporulate in the presence of sunlight, and then send impulses (chemical or electrical) to mechanisms deeper in the probe which trigger the next stages of reactivation. The fact the shield is a mess won't be such a big problem because such a mechanism could be made highly decentralized and still be functional even if only a small percentage of the tissue was still viable.

In order for this scheme to work, the panels have to produce enough power for the logic to realize it's out of normal orientation, be specific enough to tell the probe how it is out of orientation, and then drive the actuators necessary to reorient itself. The problem is that solar panels are Lambertian, and there is reflection from the surface when the panel is rotated away from normal. Both of these effects have maxima at parallel orientation. At 57° away from normal, the amount of light getting to the solar cell is only 50% of what it is at normal. There is 22% chance that the probe will find itself in this orientation. There is only 41% chance that the probe will find itself within 80° from normal, where it will only get 10% maximum power availible at this distance.

Also, the logic has to be extremely simple and the actuator low power for this to work, because otherwise, you'd have to spend three months warming it up before you can do anything.

This is true, but these don't seem to me like insurmountable problems. I'm no computer expert but I'm pretty sure a small fraction of the energy needed to run a highly intelligent AI would be needed to run a computer system capable of going "hey wait a minute, I'm supposed to have X watts available but I only have Y, let me check my sensor inputs, oh yeah I'm 45 degrees out of alignment, better fire emergency thrusters in X pattern to get myself back in alignment." Especially as these operations could easily be spread over hours; even at .01 c you'll only get a few tenths of an AU closer to the target star in that time, so such a safety margin should be readily doable.

How much energy do you propose to use doing this?

Figuring out what's wrong and realigning the solar panels over a period of hours? Well, let's go back to the Astrochicken projection. The fast-replicating version requires a solar panel capable of gathering 10 MW at Earth distance. Let's say it "wakes up" at 16 AU; at that distance its solar panel will be capable of gathering 39 KW if fully turned to the sun. Assuming the solar panel is turned so that 1/100th that much energy is actually getting to it it's getting 390 watts. Since the human body runs on average at around 100 watts (based on a 2000 a day calorie diet), if we assume the AI is as energy efficient as the human brain this should be plenty to run one of the three redundant AI cores. Actually the human brain consumes about 20% of the body's energy at rest (ref), so running one of the AI's three redundant cores should require 20 watts. This gives the Astrochicken human-equivalent intelligence, which is probably way overkill for the needs of the scenario, but let's be generous. That leaves 370 watts to run some sort of thruster, reaction wheel, or just change the orientation of the solar panel. I admit I have no idea how to calculate the energy this would take, but even if we must accelerate the probe's 141 ton mass by 1 m/s this would require a mere 191 seconds (3.2 minutes) of charging to build up the necessary energy. In fact the biggest problem will probably be the thaw time. Assuming the probe is made of water (it is organic) it would require 118 gigajoules to raise the temperature of the entire mass by 200 C. It would probably be best to make the correction system as self-contained and small as possible so that we can concentrate on thawing out a manageable kilogram quantities of mass instead of the whole probe. At 390 watts we could raise the temperature of 10 kg of material by 200 C in 6 hours. Of course, this does leave the problem of the solar panel. Unless it was incredibly light (on the order of a fraction of a gram per m^2) it would have to be designed to function effectively in deep cold or its multi-ton mass would require many days to thaw, during which time at .01 c the probe would have covered many AUs. Of course, if the probe is organic there is the possibility of having it store "fat" during its replication which it could burn in this period. To raise the temperature of a 73 ton solar panel (1 kg/m^2) 200 C would require 61 GJ, which is the equivalent to 7262 days of food on a 2000 calorie diet. Atomic Rocket life support page estimates a man will consume 2.3 kg of food per day, so the probe will therefore have to store 16.7 tons of "fat" to thaw out its solar panel with stored chemical energy. Note; I haven't looked up the background temperature at 16 AU and I admit 200 C is probably somewhat overoptimistic (I mostly used it because it was mathematically convenient). Still, it's probably not off by more than a factor of 2 or so.

Secondly, as expensive as they are, solar panels still should be much easier to process and faster to create than the 135,000 tonnes of water you need to process. Heavy water fractionation takes time. It takes a certain amount of time to start up, and an onboard plant has a certain throughput no matter how much power you have.

However, solar has many disadvantages to nuclear power for this application, as you yourself have pointed out.

With a single-junction silicon solar panel, a .1 mm thick panel with convesion efficiency 10-20% is about the best you can hope for. Any thinner and the silicon becomes progressively transparent. The 10-20% efficiency is due to the fact that the band gap of silicon is 1.1 eV. Diamonds would be worthless as solar panels, and the other elements we can make solar panels out of are quite rare, so forget about multijunction solar panels of any stripe.

What about advanced solar cell concepts like polymer solar cells, dye-sensitized solar cells, nanocrystal solar cells, and carbon nanotube solar cells?

[Wyrm]

What about carbon nanotubes? According to Wikipedia they can be made semiconducting with a very small band gap.

Then you run into the goldilocks problem. The band gap has to be small enough for thermal motion to electrons across the gap, but too small and the semiconductor's electrical properties become quite hard to control. In order for doping to work, the impurities have to be about two orders of magnitude or more above the number of charge carriers the substrate has as an intrinsic semiconductor. Too small a band gap, and the semiconductor has to be heavily doped in order to make it p- or n-type. Also, two small a band gap makes p-n junctions, the foundation of semiconductor electronics, have very narrow deplition regions and pose hardly a barrier at all for electrons, so they tunnel right on through.

Solar cells suffer a similar goldilocks problem; the energy by which the photon exceeds the band gap is wasted, so obviously you don't want the band gap too small either. Carbon nanotubes, however, can be used to reduce reflection losses at various angles, although Lambertian effects still hold.

Hmm, what about using temperature change as a trigger? Maybe have a vessel in the probe filled with an ice or fluid that boils at very low temperature, and run a pipe into it with a miniature turbine in it. As the probe draws close to the target star (say, passing the orbit of Saturn) the sun warms it enough to boil the substance, which turns the turbine, which charges the mechanism that automatically makes the solar panels unfurl. Assuming the probe approached the target star at an angle so it isn't completely in the shadow of the forward shield that should work. The big problem I can see with that is surveying red dwarf star systems, where the probe would have to get very close to the star to be warmed appreciably, leaving it little time to slow down.

That's actually rather clever. The sunlight doesn't have to go through an optically smooth surface. Suffers the problem that mechanical devices will probably be just as decrepit as the electronics, though.

Another possibility is some kind of internal clock. It obviously can't need active power. Perhaps a radioactive decay clock? Although that would require radioactive materials.

And a means of detecting the activity and when it has fallen below some critical value... which requires power.

Or maybe a mechanical timer? Something along the lines of a very slow-flowing hourglass with some kind of pressure-sensitive mechanical trigger at the bottom (obviously it couldn't be exactly that, because that would need gravity to work, but you get the general idea). Hmm, would it be possible to build a wind-up mechanical clock that takes decades to centuries to run down?

Very very difficult without some sort of energy input, and keeping energy over that period of time is something we both agree is not really plausible. Even the Atmos clock requires temperature changes to achieve zero winding.

Since we were discussing a biological probe earlier, perhaps have a flask with some living tissue in it that metabolizes a nutrient over a very slow and predictable rate, and when the nutrient is depleted automatically triggers unfurling of the solar panels through some means.

And how do you suppose to detect the complete consumption of the nutrient?

To return to the Astrochicken idea, one possibility that occurs to me is to have the whipple shields themselves made out of thin slices of photosynthetic cells. In deep space they form spores covered with light-sensitive receptors that cause them to de-sporulate in the presence of sunlight, and then send impulses (chemical or electrical) to mechanisms deeper in the probe which trigger the next stages of reactivation. The fact the shield is a mess won't be such a big problem because such a mechanism could be made highly decentralized and still be functional even if only a small percentage of the tissue was still viable.

Except for the fact that the entire shield was taking about even erosion from the interstellar medium. Each square centimeter takes 61 J of kinetic energy over the course of the century, with no opportunity to repair the damage. If we assume a spore of 100 microns wide and the density of water, then the accumulated radiation dose is 4 MGr. The radiation is primarily protons and a quarter alphas, which gives us a Q = 4.375, and assuming that these things are the hardiest of protozoans (N = 0.0003) gives us a equivalent dose of 6 kSv in a human being. This grossly exceeds the LD100/7 day limit. Your spores are fried.

This is true, but these don't seem to me like insurmountable problems. I'm no computer expert but I'm pretty sure a small fraction of the energy needed to run a highly intelligent AI would be needed to run a computer system capable of going "hey wait a minute, I'm supposed to have X watts available but I only have Y, let me check my sensor inputs, oh yeah I'm 45 degrees out of alignment, better fire emergency thrusters in X pattern to get myself back in alignment." Especially as these operations could easily be spread over hours; even at .01 c you'll only get a few tenths of an AU closer to the target star in that time, so such a safety margin should be readily doable.

So you plan to perform logic intended to get the energy before the semiconductors are warm enough to work by having the semiconductors perform calculations before they're warm enough to work? You don't have to be a computer expert to realize that a computer that's too cold to work is capable of no logic at all.

By "the logic", I mean a distinct electrical circuit that can actually work in low temperatures. I used 'realize' in a very loose sense, in that the way the circuits would respond to a different level of light hitting it would produce a particular behavior. The problem is that's very difficult to do without some sort of electrically driven switch. We may have to look into relays.

Of course, this does leave the problem of the solar panel. Unless it was incredibly light (on the order of a fraction of a gram per m^2) it would have to be designed to function effectively in deep cold or its multi-ton mass would require many days to thaw, during which time at .01 c the probe would have covered many AUs. Of course, if the probe is organic there is the possibility of having it store "fat" during its replication which it could burn in this period. To raise the temperature of a 73 ton solar panel (1 kg/m^2) 200 C would require 61 GJ, which is the equivalent to 7262 days of food on a 2000 calorie diet. Atomic Rocket life support page estimates a man will consume 2.3 kg of food per day, so the probe will therefore have to store 16.7 tons of "fat" to thaw out its solar panel with stored chemical energy. Note; I haven't looked up the background temperature at 16 AU and I admit 200 C is probably somewhat overoptimistic (I mostly used it because it was mathematically convenient). Still, it's probably not off by more than a factor of 2 or so.

So... you're assuming that the astrochicken is biochemically active enough to warm its wake-up solar panel to where it starts working, to enable that same solar panel to warm the rest of it up to the point where its biochemically active. Does anyone else see what's wrong with this scheme?

However, solar has many disadvantages to nuclear power for this application, as you yourself have pointed out.

Yes, it does. However, they have many advantages as well, including being more amenable to wilderness fabrication. We don't even know the alpha of heavy water fractionation aparatus with enough throughput to create enough heavy water fast enough to provide a fusion plant with deuterium as fast as it is burnt, let alone let it build a stockpile.

What about advanced solar cell concepts like polymer solar cells, dye-sensitized solar cells, nanocrystal solar cells, and carbon nanotube solar cells?

The problem with all of your proposals is that, in order to be practical for wilderness fabrication, they have to be so much more awesome than silicon solar that its worth it to mine for the rarer elements and drag along the necessary machinery to make them, rather than simply making some extra area in silicon. Current dye-sensitized solar cells, for instance, require ruthenium dyes, and ruthenium is six orders of magnitude less abundant than silicon. The tellurium used in 3D carbon nanotube solar panels is almost as rare as ruthenium, but also include extra steps that would not be needed in an ordinary silicon solar panel.

At best, each can improve on the efficiency of silicon panels by at most five times. But to get the equivalent power, you just need to construct about five times the amount of silicon panels.

[Junghalli]

And how do you suppose to detect the complete consumption of the nutrient?

One possibility that had occurred to me was to have the specialized cells produce some sort of chemical which would build up over time and, once a certain threshold was reached, would initiate some sort of reaction that begins reactivation. One possibility might be to have the flask's interior covered with neurons, and the specialized cells produce a neurotransmitter that causes the neurons to fire when their receptors are saturated, the firing initiating a cascade of events that begins reactivation. The big challenge would be keeping the flask warm enough for metabolism to continue. A good idea would probably be to wrap it heavily in insulators to conserve what little heat it produces. If we could get it to keep itself warm producing a mere 1 watt we would require less than a ton of "fat" to keep it going over a 100 year voyage. Of course this, in turn, opens up the question of being able to exploit that "fat" when it's frozen solid. You'd have to work out some system of thawing out a very small amount at a time as it's consumed. Of course, if we're not assuming an Astrochicken we might have more efficient energy storage options than stored biochemical energy.

Except for the fact that the entire shield was taking about even erosion from the interstellar medium.

This would mostly happen on the side of the probe facing into the "wind" (i.e. the forward portion) wouldn't it? What about using the same idea but placing the photosynthetic cells on the probe's sides instead? As long as the probe's body isn't completely shielded from sunlight by the forward shield I'd think that would work better.

So you plan to perform logic intended to get the energy before the semiconductors are warm enough to work by having the semiconductors perform calculations before they're warm enough to work? You don't have to be a computer expert to realize that a computer that's too cold to work is capable of no logic at all.

You'd need enough of a safety margin that you'd be able to thaw out some small part of the computer even with only a small fraction of the solar panel's rated output at the distance from the local star at which the probe is supposed to "awaken", yes.

So... you're assuming that the astrochicken is biochemically active enough to warm its wake-up solar panel to where it starts working, to enable that same solar panel to warm the rest of it up to the point where its biochemically active. Does anyone else see what's wrong with this scheme?

One solution that occurs to me is to use a cascade effect. You store the "fat" inside the cells of the Astrochicken's solar panel. The first handful of cells thawed by the sun go into metabolic overdrive, burning their "fat" to produce heat which thaws out their neighbors. The neighbors, when thawed, do the same thing, in turn warming up their neighbors. Eventually you thaw out the whole panel by chain reaction.

[Wyrm]

Of course, if we're not assuming an Astrochicken we might have more efficient energy storage options than stored biochemical energy.

Don't knock chemical energy. Unlike mechanical or electrical energy, chemical energy can be very stable over long periods of time, especially if it's cold. Most foods spoil only because some other critter has gotten to it first, not because of any inherent lifetime of the food.

As to your timing device, it's a maybe. It's exceedingly low Q, but as long as it has an accuracy of a few months over the century.

This would mostly happen on the side of the probe facing into the "wind" (i.e. the forward portion) wouldn't it? What about using the same idea but placing the photosynthetic cells on the probe's sides instead? As long as the probe's body isn't completely shielded from sunlight by the forward shield I'd think that would work better.

...

You'd need enough of a safety margin that you'd be able to thaw out some small part of the computer even with only a small fraction of the solar panel's rated output at the distance from the local star at which the probe is supposed to "awaken", yes.

"Safety margin"? You realize that the solar panels, being made of silicon, have to warm up themselves before they start working. Problem is that they are at thermal equilibrium with their surroundings. They're also nearly black. This makes calculating their equilibrium temperature at a given distance duck soup:

σT⁴ = W/R²

where W is the solar constant (1366 W/m²), and R is the distance in AU. When we put T = -40 °C (233.15 K), R = 2.85 AU. There will be no power coming from the solar panels until the probe gets within 2.85 AU of the target star, if the sunlight is coming in approximately normal. It doesn't matter how big the solar panel is (notice area is not part of the equation), it will be useless any distance further out than 2.85 AU.

Let's say the probe is coming in on a hyperbolic trajectory with a hyperbolic excess of 30 km/s and perihelion of 0.7 AU. Then it attains 39 km/s at 2.85 AU from the star. The true anomaly at this point is approximately -101.3°, before the perihelion. A rough determination of the true anomaly three months later, when the computer finally thaws, is 61.6°, after the perihelion, and leaving the probe about two months to slow down before it passes out of the point where the solar panels are useless.

This is for a solar panel pointed approximately at the sun, that is, in the direction it's headed. Since solar panels suffer from Lambertian effects, any solar panel that is angled away from the probe's direction of motion will catch less sunlight over its entire area, which changes the distance where the equilibrium temperature is -40 °C, and it will never be less than 2.85 AU.

One solution that occurs to me is to use a cascade effect. You store the "fat" inside the cells of the Astrochicken's solar panel. The first handful of cells thawed by the sun go into metabolic overdrive, burning their "fat" to produce heat which thaws out their neighbors. The neighbors, when thawed, do the same thing, in turn warming up their neighbors. Eventually you thaw out the whole panel by chain reaction.

Assuming you don't send it into thermal shock, but where are you going to get the oxygen to burn the fat? Also, "thaw" implies that the cells are frozen (water ice), which doesn't melt until you're within 1 AU.

[Junghalli]

"Safety margin"? You realize that the solar panels, being made of silicon, have to warm up themselves before they start working.

In which case you'd need to use stored energy to warm them up (which gets you back to needing an "alarm clock" of some sort). Unless you could produce very low temperature solar panels (how feasible would that be anyway?). Or you could wait until you are within 2.85 AU of the star to begin reactivation as you indicated below. Assuming a solar sail you would do most of your acceleration/decelleration close to the star anyway; whatever energy boost you could get beyond 2.85 AU would be very marginal anyway and probably scarcely worth fretting over its loss. A solar sail gets 9 newtons/km^2 at 1 AU (ref), at 2 AU it will be a mere 2.25 newtons/km^2, beyond 2.85 AU it will be even worse. Given the way the effectiveness of a solar sail increases exponentially as you get closer to the star you can probably leave a fairly decent safety margin in the event of the panel pointing the wrong way (at a little close than Mars's distance from the sun you will be getting 4 times as much energy over the solar panel as the minimum necessary to thaw it).

The downside is you will have problems with red dwarfs, where you will have to get incredibly close to the star to get any decent energy from it, but solar power is a big problem with those in general. The fact that a solar powered or propelled probe would have serious trouble with most of the stars in the galaxy is one of the best arguments for wanting to use nuclear power.

Assuming you don't send it into thermal shock, but where are you going to get the oxygen to burn the fat? Also, "thaw" implies that the cells are frozen (water ice), which doesn't melt until you're within 1 AU.

Oxygen (bonded in other chemicals like water) is readily available in carbonaceous asteroids, and could be stored away during replication (perhaps in some hemoglobin-like compound). Thawing is a more significant problem. One possibility that occurs to me is to place a lens above the solar panel which could focus sunlight down onto a small area and thaw it out (basically like burning ants with a magnifying glass). Of course, if the probe is misalligned the lens would not be focusing on the correct area, so you'd need a lot of different lenses pointing in different directions. Alternately, perhaps you could rig up a small secondary panel capable of operating at low temperatures which could direct energy into the surrounding tissue.

[Wyrm]

In which case you'd need to use stored energy to warm them up (which gets you back to needing an "alarm clock" of some sort).

(1) Using an "alarm clock" to heat up the solar panel defeats the purpose of the whole exercise. The point was having the first current from the solar panel being the first in a chain of events that wakes up the probe. If you're setting an alarm clock, you don't need to go through all this rigamarole.

(2) If you're using onboard heat to warm the solar panel to operating temperature, you'd best be saving that energy for some other use, because it is a net loss. The solar panel needs to be closer than 2.85 AU because that's where the equilibrium temperature is greater than the operating temperature. The panel is receiving power at W/R² per square meter from the sun, but it is radiating power at σT⁴ per square meter. The power we can get from the panel is limited by the former, and the power we must put back into the panel is equal to the latter. If the quantity W/R² - σT⁴ < 0, then you are putting more energy into the panel than you can get out. Past 2.85 AU, monojunction silicon solar panels are worthless.

It gets worse. The power you use from the panels cannot be used to heat the panels. Therefore, the balance equation looks more like

(1-ϵ)σT⁴ = W/R²

where ϵ is the real efficiency (the power harvested and not going back to the panel for heating vs the incident power), which is limited to the efficiency of the panel (15%). If you harvest energy at 15% incident power right at 2.85 AU, the panel cools down and stops working again until it warms back up. As the distance decreases, you can draw more power from each square meter of solar panel until you reach maximum efficiency.

Unless you could produce very low temperature solar panels (how feasible would that be anyway?).

Not very. You need a semiconductor that has a narrower band gap so that thermal motion can make carriers with the right impurities. However, that narrow band gap has three detrimental effects. The first is that, although more photons can be harvested, more energy is wasted for each photon. The second is that the potential difference across the junction in volts is exactly the band gap in electron-volts (which shouldn't be surprising). The less the potential difference, the less energy the collected current carries. The third is that the band gap is the barrier that keeps the charges on the "wrong side" once it gets popped over by a photon. However, if the band gap gets to narrow, tunneling becomes a problem, and the semiconductor begins to behave more like a straight conductor, which is NOT what you want.

For Sol's light, the optimal efficiency is 1.4 eV, and with a 1.1 eV band gap, silicon is already on the descending arm.

Or you could wait until you are within 2.85 AU of the star to begin reactivation as you indicated below.

Uh, no. That was not the point. When the solar panel first begins to be active, it starts reactivating the probe immediately. It's just that it takes three months to do it; reactivating from dead cold takes time, and when it finally does it's already 1.04 AU past perihelion and on its way out — I stated the true anomaly at that point as 61.6°, which is after perihelion.

Assuming a solar sail you would do most of your acceleration/deceleration close to the star anyway; whatever energy boost you could get beyond 2.85 AU would be very marginal anyway and probably scarcely worth fretting over its loss. A solar sail gets 9 newtons/km^2 at 1 AU (ref),

9.1130 N/km² maximum at 1 AU. The thrust of a solar sail is F = [(1+R)P/c] cos² θ, where P is the power incident over the sail, R is the reflectivity of the sail (usually close to 1) and θ is the angle from normal at which the sun strikes the sail — both Lambertian and law of reflection effects are at work. Work it out and it comes to approximately 9.1130 N/km².

at 2 AU it will be a mere 2.25 newtons/km^2, beyond 2.85 AU it will be even worse. Given the way the effectiveness of a solar sail increases exponentially

No. Thrust is as the inverse-square of distance. Inverse-square is not exponential.

as you get closer to the star you can probably leave a fairly decent safety margin in the event of the panel pointing the wrong way (at a little close than Mars's distance from the sun you will be getting 4 times as much energy over the solar panel as the minimum necessary to thaw it).

If the probe is coming in on the trajectory I used as an example, then at 2.85 AU, the probe is going at 39 km/s. At this speed, it will take at most 0.90 year to cross to the other side of the trajectory. The escape velocity at 2.85 AU is 25 km/s, so over the course of a year, the probe must accumulate a delta-v of 14 km/s. If the average incident flux is 432.28 W/m², this gives us 2.88 N/km². If the probe masses at 1000 tonnes, this works out to a delta-v of 8.197e-2 m/s per km² of sail area. To get a delta-v by the required 14 km/s requires at least 17,079 km² of sail area, 400 km on a side. And this is just to be barely captured by the star.

On the other hand, a 17,079 km² sail, if kept open for the entire journey pointed at the star, gets impacted with 61 J per square centimeter of its area, or 1.0418342724744e17 J over the entire sail, so your sail is either eroded away, or it has robbed the probe of kinetic energy about the same order of its own and probably dulled so that it's no use as a solar sail. That means the sail has to be closed during its journey, with all the attendant problems.

The downside is you will have problems with red dwarfs, where you will have to get incredibly close to the star to get any decent energy from it, but solar power is a big problem with those in general.

All the above problems are magnified by a red star, although such a star would have lower mass and therefore more time to get things done.

The fact that a solar powered or propelled probe would have serious trouble with most of the stars in the galaxy is one of the best arguments for wanting to use nuclear power.

Wholehearted agreement there!

Oxygen (bonded in other chemicals like water) is readily available in carbonaceous asteroids, and could be stored away during replication (perhaps in some hemoglobin-like compound).

I meant that more in the sense that, in order to burn fat, you have to have the oxygen readily available to the cell, to the tune of about 4 times the amount of fat by mass, and only a small portion of it is heated to where dissolved oxygen would have any mobility. Also, water has this annoying property that it rejects dissolved gasses as it freezes. That's why ice you make at home is cloudy.

Thawing is a more significant problem. One possibility that occurs to me is to place a lens above the solar panel which could focus sunlight down onto a small area and thaw it out (basically like burning ants with a magnifying glass).

And you're right back to the effect that ramming into the interstellar medium at 30 km/s does to optically perfect surfaces.

Of course, if the probe is misaligned the lens would not be focusing on the correct area, so you'd need a lot of different lenses pointing in different directions.

That'd be a metric shitload of lenses, Jung.

Alternately, perhaps you could rig up a small secondary panel capable of operating at low temperatures which could direct energy into the surrounding tissue.

Then you inherit all the problems with using solar panels to wake up the probe.

[Junghalli]

Uh, no. That was not the point. When the solar panel first begins to be active, it starts reactivating the probe immediately. It's just that it takes three months to do it; reactivating from dead cold takes time, and when it finally does it's already 1.04 AU past perihelion and on its way out — I stated the true anomaly at that point as 61.6°, which is after perihelion.

Why do we need three months? At 4 AU a 10% efficient solar panel will generate 8.5 watts/m^2. To simplify the math let's assume we're talking about a star that emits less energy the solar panel can exploit than our sun but about as much heat (totally unrealistic, but even this pessimistic calculation should illustrate my point) where this is the level of sunlight at 2.85 AU. Assuming the computer is 1 ton of silicon it will require 178 megajoules to thaw from 250 degrees cold. Assuming we must revive the probe by the time it reaches 1 AU we need to thaw it in one day. We require 2.06 kilowatts. That should be generated (with surplus) by a solar panel 16 X 16 meters; large but doable. Of course, this is probably a pessimistic figure as between 2.85-1 AU the probe will have shrunk the distance by a factor of more than 2 and hence increased the sunlight hitting its panels by a factor of more than 4. I stress again, this is a dramatically over-conservative calculation, the actual energy hitting the solar panel at 2.85 AU around a Sol-like star will give something closer to 13-17 watts of exploitable energy.

A decelleration within 1 AU is not particularly terrible for a solar sail probe like the advanced design suggested in the paper I linked to a while back, which boosted from .019 AU from the sun and would do most of its acceleration/decelleration quite close to the sun. Simply taking the roughly 9 newtons/km^2 it gets at 1 AU from a star like ours and applying the inverse square law to it is illustrative.

2 AU - 2.25 N/km^2
1 AU - 9 N/km^2
.5 AU - 36 N km/s^2
.25 AU - 144 N/km^2
.125 AU - 576 N/km^2
.0625 AU - 2304 N/km^2
.03125 AU - 9216 N/km^2

The probe will do most of its decelleration quite close to the sun where it gets a great deal of impulse from intense light pressure. It gets an average of 64 times as much thrust between .25-.125 AU as between 1-2 AU, which compensating for the fact that portion of the acceleration track is 8X shorter means the speed sacrificed by losing the 2-1 AU acceleration track is 1/8 what is obtained in the .2-.1 AU track, ergo having to stop acceleration at 1 AU (so you can restart decelleration there) is not likely to be a very major loss. Of course, such a probe will have to be able to withstand intense heat.

Of course, the above analyses does ignore the safety margin required in case the probe is improperly alligned. To go back to the Astrochicken illustration, that probe would have a 73,000 m^2 solar panel, which at 285+ times the area of the panel in my beginning calculation provides a fairly comfortable safety margin right there. A 73,000 m^2 panel at 2.85 AU would gather sufficient energy to thaw 1 ton of silicon within 5 minutes (granted, how quickly can you thaw silicon before thermal damage becomes a problem?).

This seems to me to be one of the more sensible strategies, if one wished to avoid all the problems of having to keep the probe active during the interstellar journey.

All the above problems are magnified by a red star, although such a star would have lower mass and therefore more time to get things done.

The dimness of stars tends to drop off a lot faster than their mass though (Proxima Centauri, for instance, is .1 Sol masses but .000053 times Sol's luminousity.

PS, hope all the math's right, it's late and I'm tired.

[Wyrm]

Why do we need three months?

To thaw out the computer without breaking its little chips to pieces with thermal shock. Do try to keep up.

At 4 AU a 10% efficient solar panel will generate 8.5 watts/m^2.

Not if it's below operating temperature. And if you need to warm it up to operating temperature, you're wasting energy.

Assuming the computer is 1 ton of silicon it will require 178 megajoules to thaw from 250 degrees cold. Assuming we must revive the probe by the time it reaches 1 AU we need to thaw it in one day. We require 2.06 kilowatts.

Warming up a piece of supercold, fragile equipment takes on the order of months, no matter how much energy you have availible.

2 AU - 2.25 N/km^2
1 AU - 9 N/km^2
.5 AU - 36 N km/s^2
.25 AU - 144 N/km^2
.125 AU - 576 N/km^2
.0625 AU - 2304 N/km^2
.03125 AU - 9216 N/km^2

This is only the thrusts normal to the surface if you are pointing your solar sail head on to the star. However, you cannot shed speed like this. For instance, at perihelion, the thrust is perpendicular to your speed, so you will not shed any speed like that. The ideal angle of attack is half the angle between the star and your direction of motion.

2 AU - 2.24 N/km²
1 AU - 8.95 N/km²
0.5 AU - 35.5 N km/s²
0.25 AU - 139.8 N/km²
0.125 AU - 539.2 N/km²
0.0625 AU - 1972.8 N/km²
0.03125 AU - 4607.9 N/km²

The probe will do most of its decelleration quite close to the sun where it gets a great deal of impulse from intense light pressure. It gets an average of 64 times as much thrust between .25-.125 AU as between 1-2 AU, ergo having to stop acceleration at 1 AU (so you can restart decelleration there) is not likely to be a very major loss. Of course, such a probe will have to be able to withstand intense heat.

Impulse is thrust integrated over time. The probe spends proportionately less time nearer the sun than it does further out.

Velocities

2 AU - 42.3 km/s
2-1 AU - 47 km/s
1 AU - 51.7 km/s
1-0.5 AU - 59.2 km/s
0.5 AU - 66.7 km/s
0.5-0.25 AU - 78.1 km/s
0.25 AU - 89.4 km/s
0.25-0.125 AU - 106.2 km/s
0.125 AU - 122.9 km/s
0.125-0.0625 AU - 147 km/s
0.0625 AU - 171.1 km/s
0.0625-0.03125 AU - 205.7 km/s
0.03125 AU - 240.2 km/s

Transit times

2-1 AU - 3.1829787234043e6 s
1-0.5 AU - 1.2635135135135e6 s
0.5-0.25 AU - 4.7887323943662e5 s
0.25-0.125 AU - 1.7608286252354e5 s
0.125-0.0625 AU - 6.3605442176871e4 s
0.0625-0.03125 AU - 2.2727272727273e4 s

Average impulses

2-1 AU - 1.7808765957447e7 kg m/s
1-0.5 AU - 2.8081587837838e7 kg m/s
0.5-0.25 AU - 4.1973239436620e7 kg m/s
0.25-0.125 AU - 5.9780131826742e7 kg m/s
0.125-0.0625 AU - 7.9888435374150e7 kg m/s
0.0625-0.03125 AU - 7.4780681818183e7 kg m/s

As you can see, the change in impulse is almost linear with each step, although each step is half the distance of the last. The combined impulse is 3.0231284225098e8 kg m/s per km² area, and imparted on 1000 tonnes (1e9 kg) this translates into .302 m/s·km². Solar sails are not really effective on the outbound transit, so we may safely ignore it. To be captured by the star, we need to shed 14 km/s over the course of the thrusting, you need a sail of about 47,000 km² in area.

Of course, the above analyses does ignore the safety margin required in case the probe is improperly alligned. To go back to the Astrochicken illustration, that probe would have a 73,000 m² solar panel, which at 285+ times the area of the panel in my beginning calculation provides a fairly comfortable safety margin right there. A 73,000 m^2 panel at 2.85 AU would gather sufficient energy to thaw 1 ton of silicon within 5 minutes (granted, how quickly can you thaw silicon before thermal damage becomes a problem?).

I thought we already agreed that it would take on the order of months to thaw a silicon computer, so you're in for a long wait if you want to do it without damaging the circuits. If a chip cracks, your probe is dead.

That's why the needed startup solar panel is so small. The energy to heat up the computer is spread over a long time, which makes for low power.

The dimness of stars tends to drop off a lot faster than their mass though (Proxima Centauri, for instance, is .1 Sol masses but .000053 times Sol's luminousity.

3.16227766e-4 Sol's luminosity. Luminosity is as approximately 3.5th power of mass. The principle problem, however, is the low temperature of the star means that most of its energy is going to be in the infrared.

[Starglider]

Why do we need three months?

To thaw out the computer without breaking its little chips to pieces with thermal shock. Do try to keep up.

I had to stop following this thread for a while, but this caught my eye as up to your best standards of hilarious inaccuracy.

Overclockers regularly use liquid nitrogen to cool contemporary, consumer grade processors, and some current processors will in fact cold-start just fine from being soaked at −150°C (some are limited to -100°C ish due to internal temperature sensor issues), and climb up to an operating temperature of -60°C to -30°C in a few minutes with no problems. Cold soaking with liquid helium (which a few overclockers have actually tried) will prevent operation of normal processors, but they can still be warmed up and brought online in less than an hour, and for some other promising designs 4 K is actually the required operating environment.

Anyway, contemporary space probes are brought out of cruise mode gradually and carefully because there's no reason not to be as cautious as possible, but this is not any sort of fundamental constraint.

As for 'we need at least the world's best supercomputer to run an AI', that is of course completely unfounded speculation, since very few contemporary AI researchers (with working prototypes - not armchair futurists) claim hardware capacity as the primary limiting factor on capabilities (basically just the detailed brain simulation teams). Even if this assumption was correct, considering that a current workstation has about the processing power of a late 90s supercomputer, you probably won't have to wait more than a couple of decades for it to be usable on a laptop.

[Junghalli]

To thaw out the computer without breaking its little chips to pieces with thermal shock. Do try to keep up.

I had to stop following this thread for a while, but this caught my eye as up to your best standards of hilarious inaccuracy.

Overclockers regularly use liquid nitrogen to cool contemporary, consumer grade processors, and some current processors will in fact cold-start just fine from being soaked at −150°C (some are limited to -100°C ish due to internal temperature sensor issues), and climb up to an operating temperature of -60°C to -30°C in a few minutes with no problems. Cold soaking with liquid helium (which a few overclockers have actually tried) will prevent operation of normal processors, but they can still be warmed up and brought online in less than an hour, and for some other promising designs 4 K is actually the required operating environment.

Being relatively ignorant of this subject I'll wait and see how this works itself out...

This is only the thrusts normal to the surface if you are pointing your solar sail head on to the star. However, you cannot shed speed like this.

Why not? If you're headed straight for the star its light pressure is applying force directly opposite your direction of motion.

Velocities

2 AU - 42.3 km/s
2-1 AU - 47 km/s
1 AU - 51.7 km/s
1-0.5 AU - 59.2 km/s
0.5 AU - 66.7 km/s
0.5-0.25 AU - 78.1 km/s
0.25 AU - 89.4 km/s
0.25-0.125 AU - 106.2 km/s
0.125 AU - 122.9 km/s
0.125-0.0625 AU - 147 km/s
0.0625 AU - 171.1 km/s
0.0625-0.03125 AU - 205.7 km/s
0.03125 AU - 240.2 km/s

Is this the velocity gained or lost during each phase of acceleration/decelleration?

If it is I get a delta V of 1427.3 km/s if you start decelleration beyond 2 AU and 1338 km/s if you start decelleration at 1 AU.

As you can see, the change in impulse is almost linear with each step, although each step is half the distance of the last. The combined impulse is 3.0231284225098e8 kg m/s per km² area, and imparted on 1000 tonnes (1e9 kg) this translates into .302 m/s·km². Solar sails are not really effective on the outbound transit, so we may safely ignore it. To be captured by the star, we need to shed 14 km/s over the course of the thrusting, you need a sail of about 47,000 km² in area.

Is 1000 tons neglecting the mass of the sail itself? According to this an advanced carbon nanotube solar sail might be made as light as 30 kg/km^2. Assuming that weight-bearing structure doubles this a 47,000 km^2 sail might have a mass of 2820 tons, without this assumption we get 1410 tons. Obviously with a solar sail it would be good to keep the payload mass as small as possible.

As for 'we need at least the world's best supercomputer to run an AI', that is of course completely unfounded speculation, since very few contemporary AI researchers (with working prototypes - not armchair futurists) claim hardware capacity as the primary limiting factor on capabilities (basically just the detailed brain simulation teams). Even if this assumption was correct, considering that a current workstation has about the processing power of a late 90s supercomputer, you probably won't have to wait more than a couple of decades for it to be usable on a laptop.

The human brain manages to run a human-level intelligence on around 20 watts and 1.4 kg of mass, and it's probably not the most efficient computing system possible by a long shot...

[Wyrm]

I had to stop following this thread for a while, but this caught my eye as up to your best standards of hilarious inaccuracy.

And had you continued to follow the thread, you'd know that I never pretended it was anything other than a not-unfounded educated guess on my part, Junghalli found no reason to disagree, and we continued to use that assumption thereon. Do try to keep up.

Overclockers regularly use liquid nitrogen to cool contemporary, consumer grade processors, and some current processors will in fact cold-start just fine from being soaked at −150°C (some are limited to -100°C ish due to internal temperature sensor issues), and climb up to an operating temperature of -60°C to -30°C in a few minutes with no problems.

Okay.

Cold soaking with liquid helium (which a few overclockers have actually tried) will prevent operation of normal processors, but they can still be warmed up and brought online in less than an hour, and for some other promising designs 4 K is actually the required operating environment.

Reading your link, I noticed that while it could be used with conventional fabrication technology, it requires the use of superconductors. This requires either the use of cooling to very low temperatures, or some sort of exotic superconductor. Keeping electronics cool close to a star requires power, supercool even more so. We don't even know if superconductivity at practically high temperatures is possible, and so far such high-temeprature superconductors are made out of elements that are not abundant. Your low-temperature probe would have to hang out in the outskirts of the system, which would be fusion powered anyway, and quite irrelevant to a probe venturing into the inner system.

Of course, if you had been keeping up, you would've known that we were keeping pretty in the pipeline of current research. Do try to keep up.

Anyway, contemporary space probes are brought out of cruise mode gradually and carefully because there's no reason not to be as cautious as possible, but this is not any sort of fundamental constraint.

Then my assumption of the probe gradually coming out of cold storage is not unfounded. Thank you.

As for 'we need at least the world's best supercomputer to run an AI', that is of course completely unfounded speculation, since very few contemporary AI researchers (with working prototypes - not armchair futurists) claim hardware capacity as the primary limiting factor on capabilities (basically just the detailed brain simulation teams).

Again, completely unreferenced. Vague mumblings of experts in the field are still vague mumblings no matter how many experts mumble them.

Even if this assumption was correct, considering that a current workstation has about the processing power of a late 90s supercomputer, you probably won't have to wait more than a couple of decades for it to be usable on a laptop.

Can such a supercomputer be fabricated from a common element like silicon? Will it be able to manufactured in situ by machinery that may be transported to another star system on a probe with a limited payload with competing components, survive harsh cosmic radiation, and work with reagents that are easily synthesized from the surroundings? If not, it's not very relevant to this discussion, is it?