Ever get the feeling you've left the oven on? William Warren had that feeling a couple of years ago, and he was right. Someone in his semiconductor research group at Sandia National Laboratories had accidentally kept an experimental capacitor--a type of component in an electric circuit--roasting too long in the lab. When they rescued it, the group decided to run some tests just for the heck of it. When they exposed the capacitor to a positive voltage, its electrical properties were just as they had expected. But when they reversed polarity and exposed the chip to a negative voltage, it behaved rather strangely. What's more, the component continued behaving strangely even after they turned the voltage off. Stranger still, they could undo the change by reversing the voltage once again.
What happened? Apparently, the unorthodox heating had freed protons to move around within the capacitor. The positive voltage nudged these charged particles down to the capacitor's bottom surface, where they acted to change the capacitor's conductivity. They stayed there until a reverse voltage came along and forced them back to the top surface. Nothing quite like this had ever been observed in electrical components made of silicon, but even so, Warren didn't think of it as anything more than a curiosity. Until, that is, he sat down with other researchers at an electronics conference a few weeks later and someone blurted out that the component might be great for nonvolatile memory--computer chips that don't forget everything when the power's turned off.
Most portable electronic devices, such as cellular telephones and digital cameras, already have some nonvolatile memory chips in them, but they are expensive to manufacture and slow compared with conventional memory. Warren's chips, by contrast, are made the same way regular chips are, which means they should be just as cheap to manufacture and just as fast. The researchers made their first batch of transistors in April 1998, but, Warren cautions, they still have a fair amount of refining to do before they are ready to produce commercial-quality working chips crammed with thousands of transistors. "Chips with great electrical properties don't do you much good if they're based on 'bolonium' and 'unobtanium,'" he muses.