I designed integrated circuits back when 1µm geographies were just coming into play and the industry believed that 1nm geometries were physically impossible due to gate widths getting close the the angstrom-dimensions of molecules. What did that and all the intervening experience teach me?
Humans suck at predicting the future.
But that's good news for you! Because the reality that we can build a 1nm transistor means you have several plausible options.
- @IDNeon mentioned the most likely solution: chemical communication. There's actually a lot of options here, but if your micromachines need mobility, that means that communication is either by leaving traces (like bees do with pheromones or ants do with a scent trail, just a bit more complex...) or through physical touch. You could credibly suggest that they leave globs of recombinant DNA. The disadvantages of these solutions (assuming that's not something cool in your world... weakness are as important as strengths) are:
If using something like rDNA or pheromones, then the information is almost always historical. A micromachine is leaving information that will be found by another machine at a later time. Leaving instructions for what to do in the future isn't impossible, but the latency is ugly.
If using something like touch and a direct chemical interaction, then much more present information can be transmitted — but it's slow as each machine must be touched to communicate its instructions or reports.
- But you could also get away with low-energy electromagnetic transmission (aka Radio). Oh, you'd be working at very high frequencies... but a (almost certain) limitation is that your transmitting over very short distances. Meters, at most (more likely centimeters). But that's not necessarily a bad thing. The problem with microscopic machines is that there's a veritable googleplex of them — and I think it is implausible to talk to all of them at once ... kinda.
Because what you could do is implement something like the Internet's communication protocol. Want to send out a global message? You send it (literally) to *.*.*.*.1 Each machine is programmed to repeat the message once (receiving a duplicate does not incur another repeat) to reasonably guarantee that all machines eventually get the message. Want to send it to everyone on the local subnet? You send it to (proverbially) 192.168.055.*. And if you want to communicate with just one machine, you send it to (again, proverbially) 192.168.055.215. If you're thinking, "that's just subnet masking!" You're on the right trail. Subnet globals and masking are a good starting place for describing how the little bounders can send out gazzillions of messages and coordinate themselves.
Because in the end, what you really have is a planet full of computers in the palm of your hand. And each one needs to pass all messages along while processing any message that meets the criteria of its addressing.
- But, let's introduce one more idea. This one's more science fiction than science... but it comes from that article I linked to earlier about the 1nm transistor. From that article we read...
You see, while the 7nm node is technically possible to produce with silicon, after that point you reach problems, where silicon transistors smaller than 7nm become so physically close together that electrons experience quantum tunneling. So instead of staying in the intended logic gate, the electrons can continuously flow from one gate to the next, essentially making it impossible for the transistors to have an off state.
But what if we discovered a way, not to insulate the gate to guarantee quantum tunneling doesn't occur, but to take advantage of quantum tunneling in a predictable way?
Today's science says it can't be done — but you need to realize, you really need to realize, that in the 90s we really, truly, and honestly believed a 1nm transistor was by every law of physics IMPOSSIBLE. Which is why I don't like being limited by today's science when it comes to answering WB questions (and why I think that coming up with all the gory details is generally a bad idea). Maybe what those little beasties are doing is using your world's version of Abrason's Law of Quantum Thermal Balancing (discovered in 2245) to allow electrons to predictably tunnel across greater (much greater) than atomic distances such that the resulting excitation state of the electron (using the atomic receptor technology first developed by Sariah Lehtonnen in 2082) can be used to modulate information.2
Obligatory YouTube video
1 I remember in the mid-80s where this literally could be done. I remember sending messages out to X.X.*.* such that every machine on the subnets would receive the message. I'm sure that can still be done in the UNIX world, but the "feature" has been heavily controlled by protocols since the late 80s and early 90s when SPAM moved from being a prank between friends/associates to the early versions of the very real problem it is today. But the idea supports the concepts you're creating for your world... n'est-ce pas?
2 Call this "technobabble" if you want... but a lot of science fiction today is using the premise of existing technology to suggest the possibility of Clarkean Magic. The quote from We are Legion, by Dennis E. Taylor, which I reference in this meta post is a great example.