Fellow JT experimenter Quantsuff has been experimenting with a Joule Thief using the 2SK170 JFET, and it runs at 131 millivolts (0.131V), powered by a solar PV cell lit by a table lamp. This is a very good JFET for low voltages; I made a similar circuit a while back and it ran at several tens of millivolts, but I had to put several JFETs in parallel to get enough current to increase the LED brightness.
Some things I learned (more like a large dose of reality).
Let’s say, for example, that the solar PV cell is putting out 150 millivolts or 0.15 volt. This is one tenth of the 1.5V AA cell. If the average Joule Thief uses 75 milliamps at 1.5V, then at 0.15V, it will need ten times as much current, or 750 milliamps, to put out the same light. But in order to draw ten times as much current at one tenth the supply voltage, the total JT resistance must be reduced to 1/100.
This means that all the wiring must be heavier gauge wire, and the wires should be kept short. The coil should have less turns of heavier wire. QS used a proto board for his circuit. Each contact in these boards may have a resistance of several milliohms, so the voltage drop (AKA IR drop) across the contacts can add up to a large percentage of the supply voltage. Changing to mechanically sound joints that are soldered good will greatly reduce the voltage drops.
The 2SK170 JFET handles up to 20 mA, but the 2SK170BL that I have are all rated 6 to 12 mA, so on average somewhere around 8 milliamps. I could see that this was going to be a problem because it will take almost a hundred of them in parallel to get the 750 milliamps that I mentioned above. That’s why the circuit board in the picture in the link above showed six of them in parallel, which was just a start.
Remember that when you’re dealing with voltages this low and currents this high, every little bit helps.