Silicon is the main material used in both computers and solar panels.

Sand is silicon.

Deserts have a lot of sand and sun.

This may be an oversimplified way of looking at things, but could we create a self-reproducing robot (nano or otherwise) that, placed in a desert, could spread using nothing but the materials that could easily be found in a desert and the energy from the sun? Or would additional resources be required?

If additional resources are needed, what is the best way to accomplish this with minimum human effort?

Basically, I'm looking for an artificial "plant" that would grow into a silicon "forest" which could capture most of the solar energy that falls on the desert, and could be used as a power source for humans.

Time is not a factor (this is a long-term project).

EDIT: Since the AI is currently beyond human technology, but is within the realm of plausibility, the hard-science part specifically applies to the materials and energy requirements. Assume that computers, assemblers, solar panels, etc. require the same materials that they use today.

Example: If rare elements not readily available in a desert are required to construct a computer, we have a problem (though if the quantity-per-unit is small enough that they can be provided with the initial "seed", you may mention that.)

Example 2: If solar power is insufficient to perform some of the operations used in the construction, we also have a problem. Solar collection does not need to be entirely photovoltaic (i.e. using solar heat directly is fine) as long as there is a method of constructing the solar-powered device using the materials and energy available.


4 Answers 4


Sand is mostly made of $SiO_2$, while silicon wafer used in solar cells are made by pure Si.

Therefore this thing has to supply the energy needed to convert the $SiO_2$ into $Si$, according to the reaction path $SiO_2 + \Delta H = Si + O_2$.

The $\Delta H$ can be calculated by the standard enthalpy of formation of the two components, being $1279 \ kJ/mol$

Solar cells and electronic circuits contain, other than silicon chips, also III or V group dopants like boron and phosphorus, used to adjust the properties of the silicon as needed. For the sake of this exercise, since they are present in traces, we can assume that their energy formation requirement is the same as silicon.

Assuming that the solar panel so produced has an efficiency comparable to the state of the art solar panels we have today, it will be around 20%, meaning that to produce 1 mol of Si the thing will have to harvest $1279/0.2=6395 \ kJ/mol$.

Considering an average power density of solar radiation on the ground to be $800 \ W/m^2$, the thing could be producing $800/1279\cdot 10^3=6 \cdot 10^{-4} mol /m^2s$.

Considering the atomic weight of silicon is 28, that would turn into an accretion rate of $0.017 \ g/m^2s$.

  • $\begingroup$ What about the solar generators that use mirrors to reflect sunlight onto a transparent bulb full of water until it boils and drives a turbine? That way you wouldn't need to turn SiO2 into Si, you'd just need to melt it into glass and something reflective to coat the mirrors with. $\endgroup$
    – nick012000
    Commented Jul 7, 2019 at 13:47
  • $\begingroup$ @nick012000, OP specified photovoltaic solar panels $\endgroup$
    – L.Dutch
    Commented Jul 7, 2019 at 13:53
  • $\begingroup$ He didn't, actually, just ones made out of silicon, and guess what glass is typically made out of. ;) $\endgroup$
    – nick012000
    Commented Jul 7, 2019 at 14:29
  • $\begingroup$ @nick012000, glass is made by silicon dioxide. $\endgroup$
    – L.Dutch
    Commented Jul 7, 2019 at 14:33
  • 1
    $\begingroup$ And what is silicon dioxide made out of? Silicon and oxygen. Thus, glass is made out of silicon. $\endgroup$
    – nick012000
    Commented Jul 7, 2019 at 14:34

I'm ignoring the hard science tag. We don't have the expertise yet.

There is substantial engineering required here. Let's do it in steps:

Step 1 is to make a fully automated solar cell plant. No humans involved.

Step 2 is to make an autonomous mine that can mine and refine silicon to the required purity.

Step 3 is to make both of these mobile.

Colateral thread:

Step 4a Make a machine that can assemble another copy of itself.

Step 4b Make a machine that given suitable resources, can make all the parts that compose it.

Step 5. Make the 4b machine good enough to make all the parts from material found in the desert. This is a challenge: It will have to be able to make VSLI circuits on a scale big enough to support it's own AI.

The first successful one will be a 'hive' creature. Many of the components will be incapable of reproduction -- worker bees. You will have some form of processing plant, and smaller critters will bring it materials. Materials will be made into solar cells, and solar cells will be mounted to provide power.

The production facility will create parts that can be used to make another production facility. Robots will move over some distance -- 20 km or so -- and start over. Once a production unit has made sevaral daughter plants, it will 'senesces' into a smaller plant that makes replaclement parts for the surrounding area. Meanwhile the daughter units expand the perimeter.

The AI has to be clever enough to come up with alternate construction methods based on what is available. I suspect that a lot of foamed rock will be used as support material, and tempered glass for structural members.

Don't hold your breath for the first one to be installed.


I’m going to base my answer off of the conclusion L.Dutch made. If we assume that a solar panel can be directly made from silicon without any energy use, we can figure out how much the solar farm would generate “compound” panels. Solar panels are about $1.635 m^2$ and about $18144 g$. $18144 g/ 0.017g(m^2s) = 1.07x10^6 m^2s$ or $296m^2h$. Converting this to solar panel hours gives us 181 solar panel hours to create another solar panel. Assuming your desert is dark 1/3 of the time, this gives us 272h per solar panel to make a solar panel. It’s not bad, but you may need to be concerned about other draws of power, like robots to get the sand off of the solar panels.


Sand also contains carbon (in the form of calcium carbonate), which can be used to make graphene. Graphene is a much more versatile substance, and so would be a superior choice. A thin layer of graphene would be super-strong, flexible, superconducting, capable of storing electricity, and and more efficient. The sand on the ocean floor could even be used for robots to build a superconducting conduit between continents around the world, so that all of the world's deserts could sell energy in a global superconducting power-grid, such that deserts in all time-zones would supply energy as was most efficient. This would also provide a new "gold mine" for such countries which have little agriculture, by allowing them to harvest and sell solar energy.

In any event, such developments would provide R&D to deploy them on other planets and moons in the solar system, with the maximized solar efficiency and graphene development, enabling the production of solar-powered spacecraft using huge unfolding graphene solar-collectors. Planets could be rapidly transformed into habitable biospheres capable of housing trillions of people, with limitless energy from solar power-- as well as fusion, if and when it becomes available. (However some places, like the moon, might depend on silicon for materials).


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .