Lasers Get the Green Light

What the world needs now is a semiconductor laser that’s good, cheap, long-lasting, powerful, and truly green. Such a device could revolutionize information display, improve certain ophthalmological therapies, and give us affordable televisions with bigger, more dazzling pictures than the best available today.

For display, it would make possible a full-color projector small enough to fit inside a mobile telephone. It may even one day be possible to project ultravivid images directly onto your retina.

It could also cut the cost of a major eye treatment, because green light is ideally suited to burn thousands of spots into the retina, stopping the proliferation of new blood vessels and ameliorating diabetic retinopathy, one of the main causes of blindness in Europe and North America. Green light propagates to greater distances through water than any other color, so it would improve underwater communications. You may also find it in laser light shows, industrial process control, and one day, in DNA sequencing machines.

But the biggest market of all would arguably be for use in television sets. Laser TVs are already available—Mitsubishi began selling a 65-inch model in the United States in 2008, for US $6999. The eye-watering price, which has since dropped hardly at all, reflects the high costs of the green and blue lasers (there’s a red one, too, of course, but it’s a relatively inexpensive descendant of the semiconductor laser chip in DVD players). The blue is a longer-wavelength sibling of the ultraviolet laser used in Blu-ray players. The real headache, though, is the green laser. Because there is no commercial semiconductor laser chip in this color range, the device’s manufacturer had to cobble together a cumbersome contraption known as a frequency doubler. It starts with a laser that emits light at the wavelength of 808 nanometers, in the part of the infrared range just beyond visible red light.

That radiation pumps a crystal that emits infrared light at 1064 nm. This light bounces in a cavity between two crystals, doubling the frequency to 532-nm emission. Voilà: green light.

For every watt of light that comes out of the original, infrared laser, you get about 0.4 watt of green light. What’s even worse—for a TV manufacturer, at any rate—is that the power and space needs of that green-laser kludge add appreciably to the complexity and cost of the control circuits for the TV.

Since the 1960s, academic and industrial research teams around the globe have been running a race to build the first reliable, manufacturable, green-emitting semiconductor laser. After a flurry of research in the late 1960s and early 1970s ended in failure, practically no one in the field saw that the key to victory was an obscure material called gallium nitride.

The story of the green laser actually starts with a different color and a different device: LEDs—blue ones.

To understand how and why, you’ll need to sit through Lasers 101 (skip this and the next four paragraphs if you’ve already taken this class). Lasing happens in a semiconducting chip when a free electron skittering through the semiconductor’s crystalline lattice drops from a higher to a lower energy band, emitting a photon of a very specific wavelength. As the photon bounces back and forth between the reflective ends of the chip, it stimulates the emission of still more photons of the same wavelength and phase. The resulting cascade of photons wrings energy from the system in the form of coherent, monochromatic light. But such a cascade is possible only if a great fraction of the electrons are in the higher energy band to begin with—an oversupply of ’excited’ electrons that’s called a population inversion. Such an inversion can be obtained in gallium nitride.

Gallium nitride is a III-V material—that is, its first element, gallium, is in Group III of the periodic table, and its second, nitrogen, is from Group V. The resulting crystal is a semiconductor. If you add just a trace of silicon to the mix, the sprinkling of additional ’dopant’ atoms will sit on only those sites in the lattice that would normally be occupied by gallium. However, because silicon is from Group IV, each atom contributes an additional, chemically unbound electron to the structure. Such free electrons can move throughout the structure, where together they occupy what’s known as the conduction band. Meanwhile, those electrons that are chemically bound end up in the valence band, which has a lower energy. The ’energy gap’ between the two bands determines how energetic the laser photons will be, and their energy, in turn, determines their wavelength. A high-energy gap means a short wavelength—green, blue, or violet, for example.

You can thus engineer gallium-nitride-based structures so that one region has a bounty of electrons—called n-type—and another an excess of holes that’s called p-type. You can then apply a voltage between those two regions to drive the electrons and holes together so that they recombine to emit light. To turn such an emission into coherent laser light, you must make sure that electrons and holes recombine in a confined space, within a reflective cavity, and that enough of the resulting photons linger inside the cavity. Only then can they start a cascade of stimulated emission that produces photons with identical properties.

To maximize the efficiency of a chip laser, specialists engineer into it semiconductor structures that trap electrons and holes in an extremely thin layer, known as a quantum well. Such a well is essentially a trench in the center of the device. In a nitride device, the trench is sandwiched by a film of aluminium gallium nitride, and the whole structure is surrounded by cladding—that is, layers having a refractive index higher than the other material. As a result, all the photons are confined between the cladding, in what’s called the wave-guiding region, and they can get out only at the ends of the cavity, which reaches to the edges of the laser chip itself. Now you might expect light to pour out at either end, like water from a gutter, but the massive refractive index difference at the semiconductor-air interface ensures that most of the light gets reflected back into the device. There it triggers the release of yet more identical photons. Result: an optical resonator that amplifies light each time it bounces back and forth, until it finally emerges past the interface in all its coherent glory.

In 1993, Shuji Nakamura, a Japanese engineer working alone, came up with improvements in gallium nitride materials science that a year later enabled his employer at the time, the Japanese company Nichia Corp., to release a blue LED with a then-whopping 2.7 percent efficiency. Two years later he fabricated the world’s first gallium nitride laser, a device that emitted ultraviolet radiation. To get green light out of a nitride laser, you need to increase the proportion of indium in the device’s indium gallium nitride quantum well. Nichia has had substantial success with this approach, and to this day it remains the undisputed world leader in gallium nitride lasers, with efficiencies for some wavelengths exceeding 20 percent.

Notoriously secretive, Nichia declined to participate in the reporting for this article. But some of the company’s progress can be gleaned from the papers it presents at academic conferences. At first, Nichia reported rapid success in extending its nitride laser emission to blue and beyond. By 2003 the company had produced a 480-nm laser—still blue but at least leaning toward green. Subsequent progress, however, was painfully slow. By 2008 Nichia’s researchers had hit 488 nm and commercialized that device, which the company describes as ’blue-green.’

If you want to get technical about it, 488 nm falls 12 nm short of the line marking the end of the blue range and the beginning of the blue-green. Then you have to go another 20 nm to reach the magic mark of 520 nm, where green light ’officially’ begins. Although the first 13 years of Nichia’s gallium nitride laser development has produced an extension in emission wavelength by 80 nm, it is far from clear if the company will ever find the additional 40 nm needed to make a truly green laser.

The problem of stretching the emission wavelength toward green comes down to a single challenge: increasing the indium content in the quantum well while maintaining the material’s quality. These lasers are grown by metal-organic chemical-vapor deposition, which involves heating the substrate and using nozzles to painstakingly inject exact quantities of organic molecules in precise ratios, which then dissociate to release atoms that form the device’s crystalline layers. At low temperatures, you get enough indium at the cost of a lot of defects. At higher temperatures, you minimize the defects but get less indium. And the indium atoms that do get in tend to skitter around on the surface, finally clumping together in a way that hampers light emission.

And as if navigating your way through this system of trade-offs weren’t bad enough, there’s an even tougher problem—known, imposingly enough, as the quantum-confined Stark effect. It stems from the interplay of the electrostatic forces and the internal stresses due to the slightly different spacings of the lattices of gallium nitride and indium gallium nitride. This interplay gives rise to internal electric fields as high as 100 volts per micrometer, enough to cause piles of static charges to form where the quantum well touches the surrounding material. The charges pull electrons to one side of the quantum well and push holes to the other. That’s bad, because it reduces the chances that those charge carriers will recombine to emit light.

Now this separation of charges does have an advantage: It pushes the emission greenward, to longer wavelengths. But this is of little benefit to lasers. Charges at the interfaces make it hard to inject more electrons and holes into the well, and if you try to force those charge carriers through by ramping up the voltage, you’ll negate the benefit of the push toward the green. True, those extra charge carriers will set up a field of their own that ’screens’ the interface charges, making it easier to inject additional carriers into the well. But this screening effect gets you nowhere because it also causes the energy gap between electrons and holes to revert to its original form—and good-bye, longer wavelengths.

This whole problem of internal electric fields can be sidestepped by switching from a crystal grown according to the standard plan to another kind altogether. Any crystal can be sectioned along a given plane, as if you’d passed a thin sheet of paper right through the lattice. Vary the angle of the section and you get different planar arrangements of atomic charges. Once you’ve cut, sliced, or milled a crystal to expose such a plane—which is like the facet of a cut diamond—you may then grow new crystalline material from that base. Whatever you grow will take on the pattern of that base.

The standard pattern, based on what’s known as the polar plane, is plagued by strong electric fields, but there are a number of alternative, nonpolar planes [see the diagram below, 'Anatomy of a Nitride Crystal']. Their intricacies began to be unraveled only in the 1990s, when David Vanderbilt, a physicist from Rutgers University, in New Jersey, published work on the topic.

It was one thing to grasp the theory of nonpolar electronics, but quite another to find a suitable substrate on which to grow nonpolar crystals. The first breakthrough came in 2000, when Klaus Ploog’s group at the Paul Drude Institute, in Berlin, produced nonpolar gallium nitride films. Though Ploog’s films depended on exotic materials and therefore couldn’t be commercialized, his work encouraged many researchers to start working on other nonpolar films. This included James Speck, a materials specialist at the University of California, Santa Barbara, who teamed up with his colleague Stephen DenBaars. And it wasn’t long before Nakamura, who had moved from Nichia to UCSB in 1999, joined them.

Their central challenge was coping with the fundamental stumbling block of all work on gallium nitride semiconductors: the lack of a good platform on which to grow the devices. Like all semiconductor devices, laser diodes are made by taking a substrate—a thin, circular slice of a crystalline material—and piling on a stack of different crystalline layers. You have to make sure that the materials in the substrate and on the layers above have compatible crystalline structures. If they don’t, one lattice will stretch or compress the other, and you’ll get strains that can fill a growing crystal with defects. Such defects can propagate like a spreading crack, causing the laser to fail after just days or weeks, rather than years.

Of course, you’d have a perfect match if you laid gallium nitride down on a gallium nitride substrate. But it’s extremely difficult to make a defect-free gallium nitride substrate (recently, however, a good deal of progress has been made). So for LEDs and lasers researchers have used sapphire as their substrate and found esoteric ways of growing reasonably good nitride crystals on top, despite the mismatch in atomic spacing.

Speck wanted to find a better substrate. He stumbled on the solution in a chance meeting with a representative from NGK Insulators, a Japanese manufacturer of materials for electrical power distribution. The visitor, who was at UCSB primarily to see Nakamura, told Speck that NGK could grow nonpolar aluminum nitride on a plane of sapphire known as the r-plane. ’I was unaware of this relationship, and I just about came off my chair,’ says Speck. Armed with this knowledge, his team started using r-plane sapphire to develop gallium nitride films. The LEDs that followed were a massive disappointment. Defects plagued the devices, and as a result they had pitifully low efficiencies.

But Speck and his coworkers could not be deterred, thanks to an unshakable belief that the winning orientation for nitrides was the nonpolar one. And success finally came in 2006, thanks to access to very high-quality nonpolar gallium nitride substrates from the Japanese firm Mitsubishi Chemical Corp. These substrates, no bigger than a thumbnail, are the key driver behind a transformation from incredibly dim emitters to those that could compete with the best conventional nitride devices.

LEDs were the first to benefit, with device efficiency rocketing from below half of 1 percent to over 40 percent. A 404-nm laser followed in early 2007; soon after, the university issued a press release to tell the world about the first nonpolar laser.

Three days later the UCSB team had the shock of their lives, when the Japanese electronics company Rohm Co. put out a press release stating that it had also produced a 404-nm nonpolar laser. What’s more, Rohm claimed that, unlike UCSB’s device, its laser could perform continuously at least as well as conventional commercial equivalents. Details of the UCSB and Rohm devices appeared in the academic press a few weeks later.

No one who read Rohm’s press release could be left in any doubt about its plans: It wanted to build a 532-nm laser for color displays, and it hoped to do it by the end of that year. Although it hasn’t gotten there yet, the company remains the leading proponent in the push toward nonpolar lasers that emit ever longer wavelengths.

By the end of 2007 Rohm reported a 459-nm nonpolar laser that featured an indium gallium nitride layer that both guided light (a waveguide, that is) and shored up the structure just at the point where it was prone to cracking. At the beginning of 2008, another Japanese laser manufacturer, Sharp Corp., entered the fray with a 463-nm nonpolar laser, but a few months later Rohm was back in front with a 481-nm device. And in February 2009 it snatched Nichia’s crown for the longest wavelength nitride laser with a device operating at 499.8 nm.

Just nine days after Rohm reported its latest result in Applied Physics Letters, the German electronics company Osram Opto Semiconductors announced the first 500-nm nitride laser in the same journal. This laser, however, employed polar crystals, following Nichia’s approach, using the conventional plane. To hit this wavelength, Osram’s researchers devoted a substantial effort to optimizing the device structure and the growth processes. ’It’s not one magic step—it’s a lot of small fantastic steps,’ explains Uwe Strauss, Osram’s director of semiconductor laser technology.

Nichia then surpassed Osram in spring 2009 with a series of 510- to 515-nm lasers, also polar. The cause of the success is still secret; Nichia says only that it has improved the material quality of its structure, with particular attention to quantum wells.

The winner of the race for the first truly green nitride laser emerged on 16 July 2009, when the Japanese tech giant Sumitomo Electric Industries described a device that emits at 531 nm. Green at last! Sumitomo employed neither a polar laser nor a nonpolar one, instead using a halfway house—a semipolar device, built on yet another plane cut through the crystal. Although semipolar lasers do suffer from internal electric fields, the fields are far weaker than those in nonpolar lasers; moreover, semipolar lasers lend themselves more readily to the production of indium-rich quantum wells, which hold the key to reaching longer wavelengths.

So we’re in the green, but not yet in the money. To make it into laser television sets, green nitride lasers will need output in the ’tens of milliwatts,’ says David Naranjo, director of product development at Mitsubishi. Today the most powerful long-wavelength laser is a nonpolar device from Rohm that produces a constant output of 92 milliwatts, but its emission is at only 493 nm, a merely bluish green. In comparison, the best results for continuous output from a polar laser come from Nichia’s 5-mW device, which emits at 510 to 515 nm; the best results using a semipolar design are produced by Sumitomo’s 2.5-mW device, which emits 520 nm. In February, Kaai, a UCSB spin-off founded by Nakamura, Speck, and DenBaars, announced a 525-nm laser; however, it didn’t say whether the laser was semipolar or nonpolar.

These devices will also have to demonstrate lifetimes of at least 100 000 hours. So far, companies and academic institutions are closely guarding all information on lifetimes, but Nichia has revealed that its 515-nm, 5-mW laser runs for over 5000 hours.

Manufacturing costs are another serious obstacle. Today it is astronomically expensive to produce nonpolar and semipolar devices, because their gallium nitride substrates are prohibitively costly and typically no bigger than a square centimeter. But Mitsubishi has started trial production of 50-millimeter (2-inch) nonpolar substrates, and it plans a commercial launch in 2010. Should nonpolar substrate costs fall to a level comparable to that of the polar gallium nitride substrates now used for laser manufacture, then nonpolar devices could have the upper hand.

Of course, a key element in unit price is the manufacturing yield. When UCSB’s Speck wanted to know how the yield of polar and nonpolar lasers compared, he went to his colleague Matthew Schmidt, a graduate student, who explained that while polar lasers work only occasionally, nonpolar lasers always do. ’His answer just converted me for life,’ says Speck. He adds that nonpolar lasers also give you the possibility of building structures that are far quicker and cheaper to grow. Because the quantum wells can be grown far thicker, they don’t require aluminium-based cladding layers, which means that growth times can be slashed from 6 to 8 hours to just 1.

While the details of the winning formula are clearly a big deal for those involved in this research, most of us just want a reliable green nitride laser to see the light of day. When it comes, laser TV prices should tumble, and the technology should find its way into the smaller displays that are common outside the United States. Ultimately this will allow all the readers of IEEE Spectrum, wherever they may be, to kick back in their living rooms and watch really colorful TVs.