So an inverter or NOT gate (not to be confused with the AC/DC power inverter) is an abstract electric component that takes one input at either a high or low voltage state, and has one output which is the opposite voltage state of the input. At least that’s how I understand it.
Normally you would use a transistor. I’m wondering how it can be made without transistors.
One idea I was thinking about is to use an old-fashioned relay with two contacts, the kind with a spring.
~Max
It really depends on what you are tryiing to achieve - does it have to switch a significant current, does it have to operate at a particular speed?
Certainly you can use relays as a form or logic - I used to work on clickety clack telephone exchanges that used such logic extensively.
OTOH a simple latching mechanical switch can do it to an extent.
Also do you want a latch system to retain the current condition or do you want it to return to a particular state when you lose power and then restore power - like a safety switch.
I’m not actually building one, just a question borne out of curiosity. But I had in mind using it for logic, as in simple appliances, not so much as a safety switch.
~Max
A transistor is by far the easiest way to do it.
Assuming that this is some sort of thought exercise, you can also create a logic inverter with vacuum tubes.
It would probably be a bit slow (even slower than a relay) but you could use a heating element and a bi-metallic disk, with a switch connected to the disk.
There are also mechanical solutions, such as using a lever to invert the position of a switch or lever, for example.
If you want something practical for an appliance, you’re better of sticking to transistors and relays. Appliance controls are much more sophisticated and reliable if you use a microcontroller to do all the fancy logic stuff.
How does that work?
~Max
Using valves is way faster than relays - the original electronic computor used valves.
You can very easily get in to the 10s of MHz switching - that would be trivial, you can go much much higher than that
Light bulb shining onto a photocell, where the latter has a pullup resistor. When the light bulb is on (1 or “high”), the photocell will conduct and produce a low voltage. When the light bulb is off (0 or “low”), the photocell will be high resistance and produce a high voltage.
Heater and NTC thermistor, where the latter has a pullup resistor. Same as above.
There are lots of ways of make a NOT function using heaters, relays, light bulbs, motors, etc.
It’s not terribly practical. As I said it’s more of a thought experiment.
If you heat up a bi-metallic disk it will change its shape as the different metals pull against each other at different rates (this is basically how your car’s thermostat works and how non-electronic furnace thermostat switches work). So your logic “input” heats up the heating element, and causes the disk to snap to its “hot” position, flipping the switch off. Turn off your logic input, the disk cools and snaps back to its “cool” position, and flips the switch on.
A motor with a normally closed centrifugal switch would be another thinking-outside-the-box type of solution (and has some practical applications as well).
Okay, that makes sense. I was looking it up and found all these bimetal thermostat components.
~Max
I was very surprised to learn many years ago that the first op-amps used vacuum tubes. I learned this because the device I’d “inherited” in our lab was built this way. The Op Amp was in a drawer that I had to slide out. According to the Wikipedia page on them, the first Operational Amplifier dates to 1941.
So my immediate thought, which has been stated many times here already, is to use vacuum tubes (“valves”).
Trained monkey who flips a switch when a light turns on?
It’s the same idea. It’s really the same as an electromagnetic relay, except that instead of a magnet it uses the bending force from differential thermal expansion.
You could build some fairly complex logic using relays. There used to be a tic-tac-toe playing machine at the Franklin Institute (back in the 1960s) that used relay logic. I believe my intro digital logic design textbook mentioned relay-based logic as well.
ETA: Professor teaching the course said that relay-based logic was used in elevator controllers back in the day.
Another thermal approach is a wax motor. They use thermal expansion to create linear actuators and are in used as thermostats.
You can build a fully-fledged computer, with floating-point math and other features, using just relays.
This guy made one fairly recently, inspired by the Z3.
The basic rule is that you can implement any digital logic, combinatorial or sequential, with only NAND gates. Not very efficiently, but theoretically possible. So you could (in theory) build a digital computer in any technology which can implement the NAND function.
(Note: only NOR gates would work as well.)
Resistor Diode Logic came after tubes and before Resistor Transistor Logic.
(After that came Transistor Transistor Logic, AKA ‘TTL’, and then CMOS.)
I don’t think you can implement a NOT gate with only diodes and resistors. At least I can’t think of a way.
~Max
Back in the 90’s I worked on some old chemical plants and oil platforms that had old pneumatic controls. Those were some fine devices. As part of the controller system, they did the NOT gates but also had pneumatic integration and differentiation 
They also recorded data on a circular graph paper.