The blue LED was supposed to be impossible—until a young engineer proposed a moonshot idea. Head to https://brilliant.org/veritasium to start your free 30-day trial, and the first 200 people get 20% off an annual premium subscription.

If you’re looking for a molecular modeling kit, try Snatoms, a kit I invented where the atoms snap together magnetically - https://ve42.co/SnatomsV

Huge thanks to the UC Santa Barbara Materials Dept (https://ssleec.ucsb.edu/) for taking us around.
Thanks to Álvaro Bermejillo Seco for reviewing the science.
Thanks to these especially helpful sources:
Nobel Prize Biography - Shuji Nakamura - https://ve42.co/NakamuraNobel 
Johnstone, B. (2015). Brilliant!. Prometheus Books. - https://ve42.co/Johnstone2015 
Nakamura, S., Pearton, S., & Fasol, G. (2010). The Blue Laser Diode: The Complete Story. Springer. - https://ve42.co/Nakamura2010 

▀▀▀
References:
https://www.youtube.com/watch?v=O8M2z2hIbag
https://www.youtube.com/watch?v=TGUteH93xNo
https://www.youtube.com/watch?v=idwKHQEw78g
https://www.youtube.com/watch?v=yoTALRhAqWc
Touchstone, L. A. (2022). Nick Holonyak Jr. University of Illinois. - https://ve42.co/Touchstone2022 
Perry, T. S. (1995). The Unsung Inventor. IEEE Spectrum. - https://ve42.co/Perry1995 
Chabay, R. & Sherwood, B. (2011). Matter & interactions (4th ed.), S2: Semiconductors. Wiley. - https://ve42.co/ChabaySherwood
How MOCVD Works via Aixtron - https://ve42.co/MOCVD 
Vangala, S. R., et al. (2019). Epitaxial growth of ZnSe on GaAs. Journal of Crystal Growth. - https://ve42.co/Vangala2019 
Nakamura, S. (1991). GaN Growth Using GaN Buffer Layer. JJAP. - https://ve42.co/Nakamura3rd1991 
Amano, H., et al. (1989). P-Type Conduction in Mg-Doped GaN w/ LEEBI. JJAP. - https://ve42.co/Amano1989 
Huang, M., et al. (2021). Defects in Mg–H‐Codoped GaN. Physica Status Solidi. - https://ve42.co/Huang2021 
Schubert, E. F. (2006). Light Emitting Diodes, Ch 4: LED basics. Cambridge University Press. - https://ve42.co/RPI-LEDs
Kitada, C. (2001). Blue About Japan. Japan Inc. - https://ve42.co/Kitada2001 
Whitaker, T. (2002). Nakamura loses Nichia patent battle. Optics.org. - https://ve42.co/NichiaSales3 
Pirates Osaka. (2014). Nakamura awarded Nobel Prize in Physics. Hatena Blog. - https://ve42.co/NichiaSales1 
Growth Bozu via Twitter. - https://ve42.co/NichiaSales2 
Rose, J. (2014). Blue LEDs – Filling the world with new light. The Royal Swedish Academy of Sciences. - https://ve42.co/Rose2014 
Pattison, P. M., et al. (2017). LED lighting efficacy. Comptes Rendus Physique. - https://ve42.co/Pattison2017 
Electricity pricing via EIA - https://ve42.co/ElectricityPricing
Lane, K., et al. (2023). Lighting. IEA. - https://ve42.co/LightingIEA 
LED Footprint via The Climate Group - https://ve42.co/ClimateGroupLED 
Nichia’s History via Nichia - https://ve42.co/NichiaHistory 
Shuji Nakamura via Wikipedia - https://ve42.co/NakamuraWiki 

Images & Video:
Lighting the World via UCTVInsight on YouTube - https://ve42.co/UCTVep2 & https://ve42.co/UCTVep3 
Palo Alto Times 1971 Article via Newspapers.com - https://ve42.co/Newspapers 
Nick Holonyak, Jr. and the LED via UIUC on YouTube - https://ve42.co/HolonyakIllinois 
The Original Blue LED via Science History Institute on YouTube - https://ve42.co/OGBlueLED 
Maxfield, M. (2022). Compound Semiconductors. EE Journal. - https://ve42.co/Maxfield2022
M. Stutzmann, et al. (2001). Playing with Polarity. pss (b). - https://ve42.co/Stutzman2001  
Isamu Akasaki in 1995 via Andrey Nikolaev on YouTube - https://ve42.co/AsakiNikolaev 
Pioneer TX-610 Stereo Tuner via Ian Marino on YouTube - https://ve42.co/StereoMarino 
Shuji Nakamura via EPO on YouTube - https://ve42.co/NakamuraEPO
Nichia Campus via Nichia on LinkedIn - https://ve42.co/NichiaHQ
Nichia via TDElektronik on YouTube - https://ve42.co/NichiaTDE 
Violeds Sterilization of COVID-19 via Seoul Viosys - https://ve42.co/SterilizationUV  

▀▀▀
Special thanks to our Patreon supporters:
Chris Harper, Max Paladino, Balkrishna Heroor, Adam Foreman, Orlando Bassotto, Tj Steyn, meg noah, KeyWestr, TTST, John H. Austin, Jr., john kiehl, Anton Ragin, Diffbot, Gnare, 
Dave Kircher, Burt Humburg, Blake Byers, Evgeny Skvortsov, Meekay, Bill Linder, Paul Peijzel, Josh Hibschman, Juan Benet, David Johnston, Ubiquity Ventures, Richard Sundvall, Lee Redden, Stephen Wilcox, Marinus Kuivenhoven, Michael Krugman, Sam Lutfi

▀▀▀
Directed by Emily Zhang
Written by Emily Zhang, Ricky Nathvani, and Derek Muller
Edited by Trenton Oliver
Illustrated by Jakub Misiek
Animated by Fabio Albertelli, Mike Radjabov, David Szakaly, Ivy Tello, and Alondra Vitae
Filmed by Derek Muller, Raquel Nuno, and Trenton Oliver
Additional research by Gregor Čavlović
Produced by Emily Zhang, Han Evans, Gregor Čavlović, and Derek Muller

Thumbnail by Ren Hurley
Additional video/photos supplied by Getty Images and Pond5
Music from Epidemic Sound

LEDs don't get their color from their plastic covers, and you can see that because here is a transparent LED that also glows the same red color.

The color of the light comes from the electronics themselves.

The casing just helps us tell different LEDs apart.

In 1962, General Electric engineer Nick Holonyak created the first visible LED.

A few years after that, engineers at Monsanto created a green LED.

But for decades, All we had were those 2 colors.

So LEDs could only be used in things like indicators, calculators, and watches.

If only we could make blue, then we could mix red, green, and blue to make white and every other color, unlocking LEDs for every type of lighting in the world, from light bulbs to phones to computers to TVs to billboards.

Throughout the 1960s, every big electronics company in the world, from IBM to GE to Bell Labs, raced to create the blue LED.

Despite the efforts of thousands of researchers, nothing worked.

10 years after Holignac's original LED turned into 20, then 30, and the hope of ever using LEDs for light faded away.

According to a director at Monsanto, these won't ever replace the kitchen light.

They'd only be used in appliances, car dashboards, and stereo sets to see if the stereo was on.

This might still be true today if not for 1 engineer who defied the entire industry and made 3 radical breakthroughs to create the world's first blue LED.

Shuji Nakamura was a researcher at a small Japanese chemical company named Nishia.

They had recently expanded into the production of semiconductors to be used in the manufacture of red and green LEDs.

But by the late 1980s, the semiconductor division was on its last legs.

They were competing against far more established companies in a crowded market.

Younger employees begged Nakamura to create new products, while senior workers called his research a waste of money.

And at Nishia, money was in short supply.

Nakamura's lab mainly consisted of machinery he had scavenged and welded together himself.

Phosphorus leaks in his lab created so many explosions that his co-workers had stopped checking in on him.

By 1988, Nakamura's supervisors were so disillusioned with his research that they told him to quit.

So it was out of desperation that he brought a radical proposal to the company's founder and president, Nobuo Ogawa.

The elusive blue LED that the likes of Sony, Toshiba, and Panasonic had all failed at.

What if Nishiya could be the 1 to create it?

After suffering loss after loss on their semiconductors for more than a decade, Ogawa took a gamble.

He devoted 500 million yen, or $3 million, likely around 15% of the company's annual profit to Nakamura's Moonshot project.

Everyone knew that LEDs had the potential to replace light bulbs, because light bulbs, the universal symbol for a bright idea, are actually terrible at making light.

They work by running current through a tungsten filament, which gets so hot it glows.

But most of the electromagnetic radiation comes out as infrared heat.

Only a negligible fraction is visible light.

In contrast, LED stands for light-emitting diode.

LEDs primarily create light, so they're far more efficient.

And a diode is just a device with 2 electrodes, which only allows current to flow in 1 direction.

When you have an isolated atom, each electron in that atom occupies a discrete energy level.

You can think of these energy levels like individual seats from a hockey stadium.

And all atoms of the same element, when they are far apart from each other, have identical available energy levels.

But when you bring multiple atoms together to form a solid, something interesting happens.

The outermost electrons now feel the pull not only of their own nucleus, but of all the other nuclei as well.

And as a result, their energy levels shift.

So instead of being identical, they become a series of closely spaced but separate energy levels, an energy band.

The highest energy band with electrons in it is known as the valence band.

And the next higher energy band is called the conduction band.

You can think of it like the balcony level.

In conductors, the valence band is only partially filled.

This means with a little bit of thermal energy, electrons can jump into nearby unfilled seats.

And if an electric field is applied, they can jump from 1 unfilled seat to the next and conduct current through the material.

In insulators, the valence band is full and the difference in energy between the valence and conduction bands, the band gap, is large.

So when an electric field is applied, no electrons can move.

There are no available seats to move into in the valence band, and the band gap is too big for any electrons to jump into the conduction band.

Semiconductors are similar to insulators, except the band gap is much smaller.

This means at room temperature a few electrons will have sufficient energy to jump into the conduction band.

And now they can easily access nearby empty seats and conduct current.

Not only that, the empty seats they left behind in the valence band can also move.

Well, really it's the nearby electrons jumping into those empty seats.

But if you look from afar, it's as though the empty seat or hole is moving like a positive charge in the opposite direction to the electrons in the conduction band.

By themselves, pure semiconductors are not that useful.

To make them way more functional, you have to add impurity atoms into the lattice.

For example, in silicon, you can add a small number of phosphorus atoms.

Phosphorus is similar to silicon, so it easily fits into the lattice, but it brings with it 1 extra valence electron.

This electron exists in a donor level just beneath the conduction band.

So with a bit of thermal energy, all these electrons can jump into the conduction band and conduct current.

Since most of the charges that can move in this type of semiconductor are electrons, which are negative, this sort of semiconductor is called n-type, n for negative.

But I should point out that the semiconductor itself is still neutral.

It's just that most of the mobile charge carriers are negative.

So there is also another type of semiconductor where most of the mobile charge carriers are positive, and it's called p-type.

To make p-type silicon, you add a small number of atoms of say, boron.

Boron fits into the lattice but brings with it 1 fewer valence electron than silicon, so it creates an empty acceptor level just above the valence band.

And with a bit of thermal energy, electrons can jump out of the valence band, leaving behind holes.

It is these positive holes which are mostly responsible for carrying current in the p-type semiconductor.

Again, the material overall is uncharged, it's just that most of the mobile charge carriers are positive holes.

Where things get interesting is when you put a piece of p-type and n-type together.

Without even connecting this to a circuit, some electrons will diffuse from n to p and fall into the holes in the p-type.

This makes the p-type a little negatively charged and the n-type a little positively charged.

So there is now an electric field inside an inert piece of material.

Electrons keep diffusing until the electric field becomes so large it prevents them from crossing over.

And now we have established the depletion region, an area depleted of mobile charge carriers.

There are no electrons in the conduction band and no holes in the valence band.

If you connect a battery the wrong way to this diode, it simply expands the depletion region until its electric field perfectly opposes that of the battery, and no current flows.

But if you flip the polarity of the battery, then the depletion region shrinks, the electric field decreases, and electrons can flow from N to P.

When an electron falls from the conduction band into a hole in the valence band, that bandgap energy can be emitted as a photon.

The energy change of the electron is emitted as light.

And this is how a light emitting diode works.

The size of the band gap determines the color of the light emitted.

In pure silicon, the band gap is only 1.1 electron volts.

So the photon released isn't visible, it's infrared light.

These LEDs are actually used in remote controls like for your TV, and you can capture them on camera.

Moving up the spectrum, you can see why the first visible light LEDs were red, and then green and why blue was so hard.

A photon of blue light requires more energy and therefore a larger band gap.

By the 1980s, after hundreds of millions of dollars had been spent hunting for the right material, Every electronics company had come up empty-handed.

But researchers had at least figured out the first critical requirement, high-quality crystal.

No matter what material you used for the blue LED, it required a near-perfect crystal structure.

Any defects in the crystal lattice disrupt the flow of electrons.

So instead of emitting their energy as visible light, it is instead dissipated as heat.

So the first step in Nakamura's proposal to Ogawa was to disappear to Florida.

He knew an old colleague there whose lab was beginning to use a new crystal-making technology called Metal Organic Chemical Vapor Deposition, or MOCVD.

An MOCVD reactor, essentially a giant oven, was, and still is, the best way to mass-produce clean crystal.

It works by injecting vapor molecules of your crystal into a hot chamber, where they react with a base material called a substrate to form layers.

It's important that the substrate lattice matches the crystal lattice being built on top of it to create a stable, smooth crystal.

The crystal layers often need to be as thin as just a couple of atoms.

Nakamura joined the lab for a year to master MOCBD.

He wasn't allowed to use the working MOCVD, so he spent 10 of his 12 months assembling a new system almost from scratch.

Even worse, his lab mates shunned him, because Nakamura didn't have a doctorate, nor any academic papers to his name as Nishiya didn't allow publishing.

His lab mates, all PhD researchers, dismissed him as a lowly technician.

Nakamura wrote, "'I feel resentful when people look down on me.

I would not allow myself to be beaten by such people.

He returned to Japan in 1989 with 2 things in hand.

1, an order for a brand new MOCVD reactor for Nishiya, and 2, a fervent desire to get his PhD.

At that time in Japan, you could earn a PhD without having to go to university simply by publishing 5 papers.

Nakamura had always known his chances of inventing the blue LED were low, but now he had a backup plan.

Even if he didn't succeed, he could at least get his PhD.

But now the question was, with MOCVD under his belt, which material should he research?

By this time, scientists had narrowed the options down to 2 main candidates, zinc selenide and gallium nitride.

These were both semiconductors with bandgaps theoretically in the blue light range.

Zinc selenide was the far more promising option.

When grown in an MOCVD reactor, it had only a 0.3% lattice mismatch with its substrate, gallium arsenide.

Therefore, zinc selenide crystal had about a thousand defects per square centimeter within the upper limit for LED functioning.

Its only issue was that while scientists had figured out multiple different ways to create N-type zinc selenide, no 1 knew how to create P-type.

In contrast, gallium nitride had been abandoned by almost everybody for 3 reasons.

First, it was much harder to make a high-quality crystal.

The best substrate for growing gallium nitride was sapphire, but its lattice mismatch was 16%.

This resulted in higher defects, over 10 billion per square centimeter.

The second problem was that like zinc selenide, scientists had only ever created N-type gallium nitride using silicon.

And third, to be commercially viable, a blue LED would have to have a total light output power of at least 1000 microwatts.

That's 2 orders of magnitude more than any prototype had ever achieved.

So between the 2 candidates, almost all researchers were focused on zinc selenide.

Nakamura surveyed the crowded field and decided that if he were going to publish 5 papers by himself, he'd better focus on gallium nitride, where the competition was much less fierce.

This material's main claim to fame was 1 development back in 1972, when RCA engineer Herbert Maruska made a tiny gallium nitride blue LED.

But it was dim and inefficient, so RCA slashed the project's budget, calling it a dead end.

20 years later, scientific opinion hadn't changed.

When Nakamura attended the biggest applied physics conference in Japan, the talks on zinc selenide had over 500 attendees.

2 of those 5 attendees were the world experts on gallium nitride, Dr.

Isamu Akasaki and his former grad student, Dr.

In contrast to Nakamura's academic background, they were researchers at Nagoya University, 1 of Japan's best.

A few years earlier, they had made a breakthrough on the first problem of high-quality crystal.

Instead of growing gallium nitride directly on sapphire, they first grew a buffer layer of aluminum nitride.

This has a lattice spacing in between that of the other 2 materials, making it easier to grow a clean gallium nitride crystal on top.

The only issue was that the aluminum caused problems for the MOCVD reactor, making the process hard to scale.

But Nakamura wasn't even close at this stage.

Back at Nishia, he couldn't get gallium nitride to even grow normally in his new MOCVD reactor.

After 6 months, desperate for results, he decided to take the machine apart and build a better version himself.

His 10 months spent putting together the reactor in Florida were suddenly invaluable.

He began following the same routine each day.

Spend the first half of the day welding, cutting, and rewiring the reactor.

Spend the rest of the day experimenting with the modified reactor to see what it can do.

At 7pm, go home, eat dinner, wash, and sleep.

Nakamura repeated this routine every single day, taking no weekends and no holidays except for New Year's Day, the most important holiday in Japan.

After a year and a half of continuous work, he came into the lab on a winter day in late 1990.

As usual, he tinkered around in the morning, grew a gallium nitride sample in the afternoon, and tested it.

But this time, the electron mobility was 4 times higher than any gallium nitride ever grown directly on sapphire.

Nakamura called it the most exciting day of his life.

His trick was to add a second nozzle to the MOCVD reactor.

The gallium nitride reactant gases had been rising in the hot chamber, mixing in the air to form a powdery waste.

But the second nozzle released a downward stream of inert gas, pinning the first flow to the substrate to form a uniform crystal.

For years, scientists had avoided adding a second stream to MOCVD because they thought it would only introduce more turbulence.

But Nakamura used a special nozzle so that even when the streams combined, they remained laminar.

He called his invention the two-flow reactor.

Now he was ready to take on Akasaki and Amano.

But instead of copying their aluminum nitride buffer layer, his two-flow design allowed him to make gallium nitride so smooth and stable, it itself could be used as a buffer layer on the sapphire substrate.

This in turn yielded an even cleaner crystal of gallium nitride on top, without the issues of aluminum.

Nakamura now had the highest quality gallium nitride crystals ever made.

But just as he was getting started, things took a wrong turn.

While he had been in Florida, Nobuo Ogawa had stepped back from Nishia to become chairman.

In his day, Nobuo had been a risk-taking scientist, designing the company's first products.

It's why he supported Nakamura's lofty plans all this time.

But in his place, his son-in-law, Eiji Ogawa, became CEO of the company.

And the younger Ogawa had a much stricter outlook.

1 Nishiya client said, he has a mind of steel and he remembers everything.

In 1990, an executive at Matsushita, an LED manufacturer and Nishiya's biggest customer, visited the company to give a talk on blue LEDs.

In it, he claimed zinc selenide was the way forward, declaring gallium nitride has no future.

That very same day, Nakamura received a note from Eiji, stop work on gallium nitride immediately.

Eiji had never supported the research and wanted to end what he saw as a colossal waste.

But Nakamura crumpled up the note and threw it away.

And he did so again and again when a succession of similar notes and phone calls came from company management.

Out of spite, he published his work on the two-flow reactor without Nishiya's knowledge.

With crystal formation settled, he turned to the second obstacle, creating p-type gallium nitride.

Here, Akasaki and Amano had again beaten him to the punch.

They had created a gallium nitride sample doped with magnesium, but at first it didn't perform as a p-type as they expected.

However, after exposing it to an electron beam, it did behave as a p-type.

The world's first p-type gallium nitride after 20 years of trying.

The catch was that no 1 knew why it worked, and the process of irradiating each crystal with electrons was too slow for commercial production.

At first, Nakamura copied Akasaki and Amano's approach, but he suspected the beam of electrons was overkill.

So he tried heating magnesium-doped gallium nitride to 400 degrees Celsius in a process known as annealing.

This worked even better than the shallow electron beam, which only made the surfaces of the samples p-type.

And simply heating things up was a quick, scalable process.

His work also revealed why the p-type had been so difficult.

To make gallium nitride with MOCVD, you supply the nitrogen from ammonia.

Where there should have been holes in the magnesium-doped gallium nitride, these hydrogen atoms were sneaking in and bonding with the magnesium, plugging all the holes.

Adding energy to the system released the hydrogen from the material, freeing up the holes again.

By now, Nakamura had all the ingredients to make a prototype blue LED, and he presented it at a workshop in St.

Louis in 1992, and received a standing ovation.

He was beginning to make a name for himself.

But even though he had created the best prototype to date, it was more of a blue-violet color and still extremely inefficient, with a light output power of just 42 microwatts, well below the 1,000-microwatt threshold for practical use.

At Nishia, the new CEO's patience had run out.

Eiji sent written orders to Nakamura to stop tinkering and turn whatever he had into a product.

But in Nakamura's own words, I kept ignoring his order.

I had been successful because I didn't listen to company orders and trusted my own judgment.

At this point, he only had the third hurdle left, getting his blue LED to a light output power of a thousand microwatts.

A known trick to increase the efficiency of LEDs was to create a well, a thin layer of material at the p-n junction called an active layer, that shrinks the band gap just a bit.

See all Veritasium transcripts on Youtube

Why It Was Almost Impossible to Make the Blue LED