A circuit component touted as the "missing link" of electronics is starting to give up the secrets of how it works.
Memristors resist the passage of electric current, "remembering" how much current passed previously.
Researchers reporting in the journal Nanotechnology have now studied their nanoscale makeup using X-rays.
They show for the first time where the current switching process happens in the devices, and how heat affects it.
First predicted theoretically in the early 1970s, the first prototype memristor was realised by researchers at Hewlett-Packard in 2008.
They are considered to be the fourth fundamental component of electronics, joining the well-established resistor, capacitor, and inductor.
Because their resistance at any time is a function of the amount of current that has passed before, they are particularly attractive as potential memory devices.
What is more, this history-dependent resistance is reminiscent of the function of the brain cells called neurons, whose propensity to pass electrical signals depends crucially on the signals that have recently passed.
The earliest implementations of the idea have been materially quite simple - a piece of titanium dioxide between two electrodes, for example.
What is going on at the microscopic and nanoscopic level, in terms of the movement of electric charges and the structure of the material, has remained something of a mystery.
Now, researchers at Hewlett-Packard including the memristor's discoverer Stan Williams, have analysed the devices using X-rays and tracked how heat builds up in them as current passes through.
The team discovered that the current in the devices flowed in a 100-nanometre channel within the device. The passage of current caused heat deposition, such that the titanium dioxide surrounding the conducting channel actually changed its structure to a non-conducting state.
A number of different theories had been posited to explain the switching behaviour, and the team was able to use the results of their X-ray experiments to determine which was correct.
The detailed knowledge of the nanometre-scale structure of memristors and precisely where heat is deposited will help to inform future engineering efforts, said Dr Williams.
He recounted the story of Thomas Edison, who said that it took him over 1,000 attempts before arriving at a working light bulb.
"Without this key information [about memristors], we are in 'Edison mode', where we just guess and modify the device at random," he told BBC News.
"With key information, we can be much more efficient in designing devices and planning experiments to improve them - as well as understand the behavior that we see."
Once these precise engineering details are used to optimise memristors' performance, they can be integrated - as memory storage components, computational devices, or even "computer neurons" - into the existing large-scale manufacturing base that currently provides computer chips.
"With the information that we gained from the present study, we now know that we can design memristors that can be used for multi-level storage - that is, instead of just storing one bit in one device, we may be able to store as many as four bits," Dr Williams said.
"The bottom line is that this is still a very young technology, but we are making very rapid progress."