From Indicator Lamps to Illumination- Quality Devices
The shift from conventional tungsten bulbs to LED lighting has been a long time coming. Faiz details the essentials of LED technology that every engineer should know.
It has been nearly 50 years since semiconductor diodes with the ability to emit light were first demonstrated. During this half century, light-emitting diodes (LEDs) have evolved from tiny devices emitting only a glimmer of light to veritable light engines capable of putting out several watts of visible light. The first LEDs were made from semiconductors such as gallium phosphide (GaP) and aluminum gallium arsenide phosphide (AlGaAsP). Those devices emitted red, yellow, and green light and were only used in electronic instruments. Only when blue-emitting LEDs based on the semiconductor gallium nitride (GaN) were developed in the 1990s, LEDs began to be seriously considered for space-lighting applications. That key enabling development won three Japanese scientists—Shuji Nakamura, Hiroshi Amano and Isamu Akasaki—the 2014 Nobel Prize in physics.
At first, red, green and blue (RGB) LEDs were used for producing white light, but later phosphor-based white LEDs were also developed. Generating white light by combining LEDs of three (or more) different colors is difficult because of the need to properly mix together light from different emitters. Those who have tried this know the difficulty involved in proper color mixing. A single LED solution for this purpose is, therefore, much more preferable. White LEDs consist of a blue LED chip coated with a suitable luminescent material. This material, called a phosphor, converts some of the blue light from the LED into red, orange, and yellow light. These colors, in combination with the residual blue light from the LED, give the impression of white light to human eyes. A typical phosphor consists of one or more rare earth materials (such as europium or terbium) doped into suitable host crystals such as silicates, aluminates or tungstates. Blue light from the “pump” LED energizes the phosphor coating which then down-converts the light to a longer wavelength. Many different types of LED phosphors and phosphor-converted LEDs are now widely available (see Photo 1).
While white LEDs are made by coating appropriate phosphors on blue LEDs, solid-state light bulbs can employ either white LEDs or blue LEDs that project their light on a phosphor-coated envelope. This later approach is termed remote phosphor and can result in longer bulb lifetimes because phosphor heating is greatly reduced in such an arrangement. Phosphors are extremely robust materials but continued heating over long periods of time lead to reduction in their wavelength conversion efficiency. For this reason, LED bulbs progressively get dimmer over thousands of hours of usage rather than failing suddenly, as is the case with incandescent lamps.
LED DEVICE TECHNOLOGY
Once blue and phosphor-based white LEDs were developed, attention turned toward making them brighter so that these devices can be used for lighting applications. This required many innovations ranging from new materials and phosphors to novel device packaging schemes. In order to appreciate some of these developments, it is worthwhile to take a look at the structure of a typical modern LED chip, as shown in Figure 1.
Blue LEDs are now almost universally made from GaN and indium gallium nitride (InGaN) deposited on either sapphire or silicon carbide wafers. As you can see in Figure 1, a typical LED wafer consists of a number of very thin layers that are deposited on a 2″ diameter sapphire substrate through a process called Metal Organic Chemical Vapor Deposition (MOCVD). The very first layer consists of aluminum nitride, which makes it possible to deposit subsequent layers. A buffer layer of gallium nitride is then followed by an n-type GaN layer. Next comes a series (usually five) of so-called “quantum wells,” each consisting of a thin layer of InGaN sandwiched between GaN layers. Finally, a layer of p-type GaN is deposited at the top. To make an LED device, some of the material is removed from the top down to the n-type GaN layer. Electrical contacts are then made to the n- and p-type GaN, which, together with the quantum wells, form the diode structure. If this device is connected to a power supply, then electrons and holes (electron deficiencies) are injected into the diode. These come together in the quantum wells where they merge with each other and in that process release light.
A particularly severe problem with all LEDs is that due to the high refractive index of the semiconductor material only a small amount of light generated inside LED chips is able to escape as visible radiation. Most of the light remains trapped inside the confines of the chip due to total internal reflection at the chip’s surfaces. It is estimated that if nothing is done about this problem, then as much as 90% of the light can remain confined inside an LED chip. This phenomenon causes undesirable heating of LED die and also lowers the efficacy (lumens output per Watt electrical energy input) of LEDs as light sources. Techniques have been developed to extract much of the trapped light from inside LEDs—increasing their apparent brightness and efficacy. One widely used method is to simply roughen the top surface of LED chips through a short etch in warm potassium hydroxide solution. This process causes uneven etching, resulting in a rough surface with random topography. This roughness aids light to escape by making the chip geometry less favorable for total internal reflection to take place. Almost all of the LEDs commercially sold today have roughened emitting surfaces.
Another technique relies on producing an ordered arrangement of tiny blind holes or depressions on LEDs’ top surface (see Photo 2). This structure is called a photonic crystal. LEDs with photonic crystals are not only brighter than other types of LEDs but also emit light in a narrower, more collimated beam. This ordered structure is much harder to produce than a simple rough surface so photonic crystal LEDs are much more expensive compared to ordinary LEDs. Such LEDs are used in some projectors and other display devices.
Manufacturers of “white” LEDs make their devices by applying a slurry of a phosphor material on to blue-emitting LED chips, as seen in Photo 3. Single white LEDs are packaged in small surface-mount plastic packages. LEDs with power dissipations as high as 5 W are now available as discrete single-chip packaged devices.
Efficient removal of heat remains the most serious issue facing designers of high-power LEDs. Devices made by Cree with silicon carbide instead of sapphire as the substrate have an advantage here because of the former’s superior thermal conductivity. The ultimate design in this respect is one where the LED chip is inverted before being packaged. In this “flip-chip” configuration, the top portion of the LED chip where the light (and heat) is generated is coated with a layer of metal to act both as reflector and electrode and this side is then bonded to a metal pad which acts as a heatsink. The bottom part of the LED die is now exposed at the top from where light can escape through the transparent sapphire. As the heat no longer needs to escape through sapphire, the chip can dissipate heat much more easily. This design prolongs LED life and enables these devices to run at higher currents, producing more light. Flip chip configuration is now standard for all moderate and high-power LEDs.
In order to reduce the manufacturing cost of blue LEDs, manufacturers have developed a new technology that makes use of silicon instead of sapphire as the substrate for the LED structure shown in Figure 1. This approach offers several advantages. Silicon is a much better thermal conductor compared to sapphire so heat can be removed much more easily with this so-called GaN-on-Si scheme. Silicon can also be easily separated from the LED structure formed on top of it. This makes it possible to fabricate thin-film substrate-less LEDs with much superior characteristics when compared with ordinary LEDs. Perhaps the most important argument in favor of GaN-on-Si LEDs is that silicon is available economically in wafer sizes as large as 12″ in diameter whereas sapphire wafers are mostly available in only 2″ diameter size—making possible great economies in scale. It costs about the same to process a 2″ diameter wafer as a 12″ diameter one so using large silicon wafers as LED substrates becomes very attractive. There are also other incentives with this approach, such as the availability of established silicon IC processing tools for processing GaN-on-Si LED wafers. This device technology is now well-developed and is gradually being commercialized. Companies, such as Plessey Semiconductors of the UK, are currently manufacturing thin film LEDs called Magic LEDs, with this technology. Packaged in a very small and thin form factor, these LEDs are targeted at mobile and wearable gadgets.
Over the past several years, much progress has been made in the quality of light emitted by LEDs. The very first white LEDs were made using a cerium-doped yttrium aluminum garnet (Ce:YAG) phosphor and emitted a harsh blue-white light. Consumers prefer a rich golden-yellow color tone to white light—close to that produced by conventional tungsten incandescent lamps. Cool white light produced by early white LEDs was considered of poor quality, incapable of bringing out the natural color of illuminated objects. This so-called color rendering can be improved by the use of better phosphors that generate a more balanced spectrum. Modern white LEDs utilize phosphors that are actually mixtures of two or three phosphors for optimizing the spectral distribution of emitted light. Thus, for instance, warm white LEDs are made from phosphors that have a small amount of europium-containing red-emitting phosphor mixed with more traditional yellow-emitting phosphor.
By properly compounding phosphor mixtures it is possible to generate very high-quality, full-spectrum (broadband) white light from LED sources. Researchers are now working on developing phosphors where a single host matrix contains multiple rare-earth ions for generating several colors through the same phosphor. Such a phosphor may contain, for instance, a mixture of europium, terbium, and cerium for producing red, green, and blue lights, respectively. This approach promises an even better lighting quality than has been achieved so far. Figure 2a and Figure 2b show the spectrum from a broadband white LED and that from an LED designed to mimic light from tungsten halogen lamps. It is remarkable that present-day LEDs can generate light that is completely indistinguishable from that emitted by thermal sources.
Spectral engineering has also been extended to applications such as lighting for indoor plant cultivation. Horticultural LED lamps are now available that can generate spectra closely matched to what is best absorbed by vegetation. Most plants appear green because they do not absorb green light; instead, they reflect it back to our eyes. Photosynthesis in such plants requires mostly red and deep blue light. Figure 2c shows the spectrum from a broadband red LED that is used as a plant grow light. Notice that the spectrum contains all shades of red light and, in addition, also contains some blue light. Specialized LED lighting, targeted toward horticultural applications, is now commonplace in the indoor cultivation industry. Their use is enabling the cultivation of unseasonable fruits and vegetables as well as farming in inclement weather. With further advances on the horizon, it is not inconceivable that we’ll have adjustable spectrum LEDs in the not too distant future.
Going to a supermarket, looking for LED light bulbs, you will notice that while 40- and 60-W bulbs are readily available, it is difficult or impossible to get 100-W LED bulbs. A major reason for this is the difficulty in removing excess heat generated by LEDs. Although LEDs are very efficient when converting electrical energy to light, they still generate very significant amounts of heat, especially when running at high current levels. Tungsten bulbs generate even more heat, but the heat they generate is mostly radiated away as near- to mid-infrared radiation, radiatively cooling the bulb in that process. As a result, incandescent bulbs radiate both light and heat.
LEDs are different in that they produce non-radiative heat which heats up the device and cannot be easily removed. Good thermal engineering and use of substantial heatsinks are needed to safely dissipate heat from LEDs and this makes the design of LED light bulbs complicated. Generally speaking, proper thermal management is the most challenging aspect of designing any high brightness LED-based illumination system. Power LEDs (devices with power ratings of 1 W or higher) come packaged in special power packages where the LED chip is mounted directly on top of a metal slug. Such packages are mounted directly on metal heat sinks such that heat can flow through a low thermal impedance path from the chip to the heat sink. The metal slug is highly polished, so use of a thermal interface compound is usually not required and in some cases may impede the flow of heat rather than aiding it.
In many designs multiple power LEDs are first mounted on a metal core PCB (MCPCB) made of electrically-insulated aluminum sheet. The mount serves as a heat spreader and is attached to an aluminum or copper heatsink for convective cooling. This method of mounting power LEDs is now widespread and many PCB makers offer custom MCPCB manufacturing services. Electrically insulating but thermally conductive materials used in commercial technologies, such as SinkPAD MCPCBs, allow thermal conductivities as high as 385 W/m∙K. For more demanding applications—for example, in LED searchlights—water-cooled mounting substrates—such as that used with high-power dissipation microprocessors—can be used.
Driving LEDs seems straightforward enough, but care is needed in the design of long-life lighting systems. Being diodes, LEDs are best driven using constant current sources. Both bipolar transistors and MOSFETS can be used for this purpose. Variable voltage at the base or gate of a transistor can then be used to adjust the brightness of LEDs. Like all semiconductor devices, LEDs too become electrically more conducting as their temperature increases. This can lead to failure through thermal runaway, if their temperature is allowed to rise unchecked. However, thermistors can be used to sense LED temperature and adjust their drive current to maintain a constant current and, therefore, constant brightness.
Simple AC mains-driven LED lighting systems can be designed with a number of different techniques. Seoul Semiconductor produces LED modules that can be driven directly from 120- or 240-V AC mains. These modules contain a chain of series connected LEDs together with a rectifier diode. Although somewhat pricier than simple DC-driven LED modules of the same brightness, Seoul Semiconductor LEDs make it a snap to develop mains-powered solid-state lighting systems. Yet another approach for driving LEDs from mains power is to implement a switch-mode buck converter with constant current drive. ICs, such as SSL5031BTS from NXP Semiconductors, allow designers to develop a simple mains-driven LED lighting system of up to 30-W power with the addition of a few external components.
If a trio of sets of red, green, and blue LEDs is used to generate light of different colors, including white, then a pulse width modulation (PWM) scheme is usually implemented with a microcontroller to control the brightnesses of the red, green, and blue channels. Normally, the PWM output is filtered with a simple one-pole RC filter before it is used to drive LEDs, either directly or with a power transistor.
A better alternative has now become possible with the availability of Linear Technology’s LTC2645. This IC is a PWM-to-DC level converting chip, which Linear Technology calls a PWM-to-VOUT DAC. This device takes in up to four channels of PWM waveforms and produces corresponding DC voltage levels. Unlike an RC filter, this is accomplished with no latency and very high accuracy. Linear Technology is also making available a reference board for this device which plugs directly on to Arduino UNO form factor boards, like a shield, to make PWM-based multichannel lighting systems very simple to implement (see Photo 4).
Solid-state lighting is making rapid strides with developments taking place on many fronts. With ongoing advances, LED-based lighting sources are becoming increasingly common with applications such as automobile headlights being the latest making the transition from incandescent to solid-state lighting.
Other prominent ongoing developments include the use of laser diodes instead of light-emitting diodes as pump sources for color-converting phosphors and the use of quantum wire structures for making LEDs of almost any desired color. For many years now, work has also gone into exploring alternative materials for making LEDs. Zinc oxide (ZnO) is a very promising material for making blue LEDs, but so far no one has managed to make good p-type ZnO—essential for making pn-junction LEDs. The shift from conventional tungsten bulbs to LED lighting is a great success story of our times and continuing developments will ensure that incandescent bulbs will become a relic of the past much as vacuum tubes were relegated to museums with the advent of solid-state electronics.
PUBLISHED IN CIRCUIT CELLAR MAGAZINE • FEBRUARY 2016 #307 – Get a PDF of the issueSponsor this Article
About the author
Faiz Rahman, PhD, is a materials and device engineer, working on novel materials and device structures for both electronics and photonics. He develops new LEDs and solid-state lighting concepts at Electrospell and is also a visiting professor in Electrical Engineering at Ohio University. Faiz is a senior member of the Optical Society of America and of IEEE. His current research interests include the development of nano LEDs, plasmonic structures, and advanced nanofabrication technologies.