Against the Dying of the Light

Cacophonous are the voices decrying the implosion of “the U.S. Innovation Ecosystem,” as the Harvard Business Review puts it.

Do not hope, the voices warn, to see in our future anything like the miraculous progress of the last two centuries, or even the last 50 years. It’s done. Tighten your belts.

Supporting the doomsters are dozens of academic studies arguing that “big research,” whether at our great corporations or our universities, yields increasingly diminishing returns. Making headlines recently was a Nature study concluding, “We find that papers and patents are increasingly less likely to break with the past in ways that push science and technology in new directions. This pattern holds universally across fields…”

The world your authors live in, the companies we visit, the astonishing innovations we witness almost daily, all give stark witness that innovation continues aboundingly.

As to whether we should depend on big research as the source of progress, much could be said. Today, however, we offer a simple parable.

Blue Light

The light emitting diode or LED is at least seven times more efficient than the classic incandescent lightbulb. Globally, that represents potential energy savings of more than $1 trillion a year.

Naturally the semiconductor industry was excited when the first red LED was invented at GE in 1962, with Monsanto following with green the next year.

Neither, however, was a potential replacement for the lightbulb. To build a full color spectrum—including white for lighting—blue would have to be added to the mix.

This turned out to be fiercely difficult. By the 1980s, after 20 years of trying, every major firm had abandoned the attempt. Engineers at Monsanto declared it would never happen.

As told by Derek Muller in a brilliant new episode of his YouTube channel Veritasium, the challenge was finally met 30 years after that first GE red light. That success, Muller explains, was down to the efforts of two eminent Japanese scientists, Dr, Isumu Akazaki and Dr. Hiroshi Amano and one Japanese tinkerer named Shuji Nakamura.

As Muller shows, all three deservedly shared the Nobel Prize for Physics, but it was the tinkerer who ran the race to the finish.

The light emitted by a diode is a function of the “bandgap” of the material used. The bandgap is the spread in energy levels between the valence and conduction bands of the electrons zipping around the nucleus of the atom. To get to blue would require forging a diode from semiconductors with large bandgaps. The two leading contenders were gallium nitride (GaN) and zinc selenide (ZnSe)

For the blue LED, three problems had to be solved.

• An LED, like any functional semiconductor, must be an almost perfect crystal. Irregularities in the crystal would cause the electrons to bump about, dissipating their energy as heat rather than light.
• A diode requires the conjunction of an n-type and p-type semiconductor. No one had ever made a p-type in either material.
• A commercially viable blue LED required an output power hundreds of times greater than had ever been achieved with either substance.

In the mid-1980s, Nakamura was a hands-on engineer who built most of the equipment in his lab at Nishiya, a Japanese electronics firm. He had been assigned to come up with products to save the company’s ailing semiconductor division. Perhaps imprudently, given that most of the industry had long abandoned the idea, he picked the blue LED as his target. Perhaps perversely, he decided to work in gallium nitride precisely because it was the less favored material, which he saw as giving him more scope for research.

While Nakamura was just getting started with his tinkering, Akazaki and Amano, the world’s leading experts on gallium nitride, appeared to have solved the first challenge of growing a high-quality crystal in gallium nitride. Saphire crystals were the chosen substrate—in effect a pattern—for making crystalline GaN—but the match between the two crystals was imperfect, causing excessive defects in the GaN crystal.

Akazaki and Amano appeared to have solved that problem by interposing a layer of aluminum between the sapphire and the GaN. Unfortunately, though this worked on a small scale, the aluminum caused problems in the crystal growing furnace (known as an MOCVD for Metal Organic Chemical Vapor Deposition) meaning the process could not be used on a commercial scale.

Shop Class Rules

Here the tinkerer stepped in for the first but not the last time. Nakamura had spent almost a year visiting a U.S. firm that was using MOCVD. He went hoping to learn advanced techniques. Instead, scorned by his colleagues for lacking a PhD, he was put to work as the lab equivalent of a handyman. One of his assignments was to build his host company’s next MOCVD for those exalted colleagues to use.

Nakamura was furious but that experience would be crucial to the world getting the blue LED. Back in Japan, where Akazaki and Amano had hit their aluminum roadblock, Nakamura focused on the machinery.

MOCVD at the time had only one nozzle to introduce the gaseous material to be deposed. With GaN, this process worked poorly because too much of the GaN dispersed around the chamber rather than deposing onto the substrate. Nakamura rebuilt the machine, adding a vertical nozzle (previously omitted for fear of the turbulence it would cause) to blow the GaN directly down onto the substrate.

The result was a high-quality crystal without the aluminum interposer. In effect, the first layer of GaN served as its own buffer between the sapphire and the additional GaN layers above, reminiscent of the crucial discovery that silicon could form its own insulating layer as silicon oxide. That insulator enabled the silicon microchip. Nakamura’s GaN buffer would be similarly crucial to the blue LED.

It was an ingenious solution, but the skills required could have been learned in a shop class and would never be taught in a university.

For the next step, making a p-type semiconductor from GaN, Akazaki and Amano again took the lead. Using magnesium as a dopant, their first attempts failed. But blasting the doped crystals with an electron beam, they made it work.

Once again, however, they had come up with a solution that would not scale commercially.

Nakamura acknowledged the electron beam approach, but looking to simplify, he tinkered again. He tried baking the doped crystal in a furnace at 400 C. At that temperature, hydrogen molecules, which had been impeding the electron flow, were forced out of the crystal and the required “electron mobility” was achieved. Once again, shop class had saved the day.

By early 1992, Nakamura had a working protype of a blue LED, but it delivered only 42 microwatts of power, far short of the 1000 minimum needed for a commercial device. A standard method to increase electron flow is to use a thin layer of material at the P-N junction to slightly shrink the band gap. For GaN, the best material for this purpose was indium gallium nitride. Alas, indium nitride and gallium nitride do not like to mix. Akasaki and Amano could not make it work.

This Goes to 11

One more time, shop class came to the rescue. Tinkering again with his MOCVD reactor, Nakamura went to 11, power pumping much more indium into the mix, hoping sheer volume would do the trick.

Surprising everyone, including Nakamura, it worked. A few more adjustments were made. By the end of 1992, the world—and Nishiya—had a blue LED, and therefore a white LED and every color of the rainbow. Nishiya was transformed into a multi-billion-dollar business. Last year, the entire LED market accounted for $90 billion in revenue and is projected to pass $300 billion by decade’s end.

As our great mentor, Carver Mead—the groundbreaking physicist who did the science confirming Moore’s Law and cofounder of some 30 tech enterprises—recently told us, forget what you hear about how “‘our new great science has created the modern world.’ Engineers created the modern world,” starting with “physical pictures they went into the lab and worked them out. That’s what built the modern world.”

Shuji Nakamura would agree.