Why Blue Was the Hardest LED to Make

Every pixel on your screen is built from three small lights: red, green and blue. Two of them were easy. The third took the best semiconductor labs in the world a generation to get right.

When the Royal Swedish Academy gave the 2014 Nobel Prize in Physics to Isamu Akasaki, Hiroshi Amano and Shuji Nakamura, the citation read "for the invention of efficient blue light-emitting diodes which has enabled bright and energy-saving white light sources." The committee chair, Per Delsing, put it more bluntly. "A lot of big companies tried to [develop blue LEDs] and they failed," he said, "but these guys persisted and eventually they succeeded."

This is the story of why blue was so stubborn, and what its arrival did to the colour of the lights around you.

Why white light needs three lights

White light made from LEDs needs red, green and blue. Mix all three and your eye sees white. Without blue, you can make orange, yellow and a muddy kind of lime, and that is about it.

Red and green diodes had been around for decades. Blue was the missing primary. It is also the hardest to make, for a physical reason. Blue photons carry more energy than red or green ones, so you need a semiconductor with a wide band gap. Gallium nitride (GaN) has one. Its large band-gap energy puts its glow in the ultraviolet, which makes it a natural candidate for blue.

The catch is that gallium nitride is a nightmare to grow and a nightmare to dope. You will see both problems in the rest of this story.

The crystal nobody could grow

In 1971, a team at RCA made the first GaN LED. Herb Maruska, Jacques Pankove and Ed Miller got it to glow, but it was far too weak to sell. A record device in 1978 reached an external efficiency of 0.12 percent. By the mid-1970s, many researchers had walked away. The crystals were poor, and they carried so many stray electrons that the material could not be made p-type.

That second phrase matters. An LED is a junction between two kinds of semiconductor. One side (n-type) has spare electrons. The other (p-type) has spare "holes" for them to fall into. When an electron drops into a hole, it gives off a photon. No p-type layer, no LED.

Akasaki did not walk away. He grew single-crystal GaN in 1975, then moved from one growth method to another, switching to MOCVD in 1979. Amano joined the effort, and early results in the 1980s were poor.

The cushion that fixed the crystal

The first breakthrough was about the surface underneath. GaN has no natural base to grow on. In 1985, Akasaki's group grew high-quality GaN single crystals by laying down a thin low-temperature aluminium nitride buffer layer first. Think of it as a cushion between a mismatched floor and the crystal on top. Dislocation density dropped to around 10^8 to 10^9 per cubic centimetre, and electron mobility reached 700 cm²/V·s.

Better crystals still did not give p-type material. In 1987 the team saw that zinc-doped GaN glowed more after being hit with an electron beam, but it was not p-type. In 1988 they switched to magnesium. In 1989, magnesium-doped samples treated with the electron beam finally conducted as p-type, and the first GaN p-n junction LED followed at roughly 0.1 percent efficiency.

It worked. It was also slow, because the electron beam treated a small area at a time. Nobody was going to light a city with it.

The factory fix: bake it

Shuji Nakamura worked at Nichia, a Japanese chemical company known for phosphors. He solved the crystal quality problem independently, and he explained why the electron beam worked. Then he found a better route. By annealing magnesium-doped GaN in a nitrogen atmosphere, he produced low-resistance p-type films with plain heat, something a factory could repeat.

The numbers tell the story. Efficiency in the Akasaki group's devices reached 1.5 percent in 1992. In 1993, Nichia released a commercial blue LED at 2.7 percent. Late that year, the company announced an AlGaN-InGaN-GaN double heterostructure LED with a luminous intensity of 1 candela. For the first time, a blue LED was bright enough to be useful.

Price followed. Nichia's blue LEDs sold for $15 each in reasonable quantities in 1994. By 1995 volume pricing had fallen to $3.

Blue plus yellow equals white

Here is the twist. Today's white LEDs mostly do not use three separate diodes at all.

According to the history Nichia has published, customers kept asking for white light, and someone inside the company suggested phosphors, which was the firm's core business. In 1996, Nichia introduced a white LED that used a 455-nanometre blue chip to excite a cerium-doped YAG phosphor. The phosphor turns some of the blue into a broad band of yellow. Your eye adds the leftover blue to the yellow and reads white.

That is why a blue chip sits at the heart of nearly every white LED. It is also why early white LEDs had a cold, bluish cast. A white built from blue and yellow alone has a gap where red should be. Fixing that means adding red and green phosphors to the blue chip, which gives a higher colour rendering index.

From a $15 curiosity to the ceiling light

Once blue existed, the rest was engineering. Hewlett-Packard introduced the first 1-watt white LED in 1998. In 2002, Lumileds launched white LEDs with a colour temperature below 4,000 K. By 2008 the best commercial white LEDs reached 100 lumens per watt.

Over one decade of that period, the light output of a one-watt package rose from 15 lumens to 130, while the price fell from $300 per kilolumen to $11. The industry calls this pattern Haitz's Law. Cost per lumen falls 90 percent per decade, while output per lamp rises by a factor of 20.

That curve explains a lot about modern life. It explains why the phone in your hand has a bright backlight, and why cities began tearing out sodium lamps. It also explains a fight you may have witnessed on your own street. Early LED streetlights leaned toward cool, blue-rich white because those chips were efficient, and many towns discovered the cost in complaints. We covered the fallout in why cities keep ripping out the streetlights they just installed.

What designers can take from a crystal

You probably do not grow gallium nitride for a living. But the history has three habits worth borrowing.

• Know where your white comes from. On screens, white is a mix of three primaries, and on a lamp it is often blue plus a phosphor. Both versions are a negotiation, not a neutral. Our colour mixer lets you see how the primaries add.
• Check the temperature. A cool white and a warm white change how every colour in a room reads. Read more about colour temperature before you pick a bulb or a background.
• Test your blues. Dark blues on dark backgrounds fail contrast checks all the time. Run yours through the contrast checker.

Blue was the last colour LEDs learned to make, and it is now the one they lean on most. The lights in your ceiling, your pocket and your desk are all, in some small way, a bet that three researchers placed on an unfashionable crystal.

Sources

• Isamu Akasaki, Hiroshi Amano and Shuji Nakamura win 2014 Nobel Prize for Physics, Physics World
• Blue LED discovery wins Physics Nobel, Nature News Blog
• 20 years and counting for the GaN LED, Compound Semiconductor
• LEDs: Two Decades of Glorious Growth, Compound Semiconductor
• Blue LED + Phosphor = the Creation of the White LED, LEDinside