most computer mice contained rubber balls. A change in a mouse’s position was detected by rollers that measured the ball’s rotation on the X and Yaxes. The technology worked—but if you owned a desktop PC back in the early 1990s, you probably remember how the mouse balls attracted dirt, hair, and other detritus.
HP had developed optical scanning technology that, in theory, could detect position changes by taking a high-resolution picture of the surface underneath a mouse hundreds of times per second, then comparing one image to the next to see which features had moved. The problem was that the technique was slow and processor-intensive.
Hartlove and his colleagues realized that there was no need to obtain a perfect image of the surface, or to try to find perfect correlations from one image to the next. They wrote new algorithms that oversampled the surface—capturing 2,500 frames per second—and then immediately threw out most of the data, looking instead for images with just enough in common to provide basic navigational cues. “It was much easier to oversample like crazy and look for good data than to take huge samples and do huge operations,” Hartlove says. “We were able to take that, myself and a couple of other guys, and turn it into a low-cost, high-volume component that replaced the ball and eliminated all the drawbacks that people found with that technology, such as cleaning and choking hazards. We shipped a billion units before I left the company.”
Hartlove applied this focus on practicality at Korea’s MagnaChip, which Francisco Partners and Citigroup Venture Capital had acquired from financially troubled Hynix in a 2004 leveraged buyout. “The task was to go in there and turn what had been a very internally focused semiconductor group into a multinational company that had customers like Motorola and Nokia, instead of specialty products done for local Korean customers,” he says. “It was similar to HP, in that there were a lot of great technologies that were completely undirected as far as where they were going and what kinds of markets they had.”
He applied the same approach at Nanosys. The company could have used its nanorod technology to go into the solar panel business, or its nanowire technology to become maker of thin-film transistors. But a far more practical and sustainable strategy, Hartlove says, was to identify niches where Nanosys could help other companies make better products. “The largest organization I ever ran was about 3,000 people at Agilent in Panang, Malaysia, and I know what that’s like,” he says. “To take this company and build that capacity is not really a very smart thing to try. Instead, I said we will look for these places where we can make our contribution in a way that no one else can, so we can be highly differentiated.”
One of those places, Hartlove decided, is the lighting industry—specifically, LED backlights for LCDs. Most cell phone screens, an increasing number of laptop screens, and a small but growing number of desktop monitors are now lit by arrays of LEDs rather than traditional fluorescent bulbs. LEDs are brighter and more efficient than fluorescent lights. But the brightest, most efficient types of LEDs are blue, meaning their light must be augmented with yellow light from Yttrium-Aluminum-garnet (YAG) phosphors to make white light. And the problem with YAG-augmented LEDs, Hartlove says, is that
