Light Emitting Diodes
E. Fred Schubert(1), Jaehee Cho(2), and Jong Kyu Kim(3)
(1)Future Chips Constellation, Department of Electrical, Computer, and Systems Engineering, Rensselaer Polytechnic Institute, 110 Eighth Street, Troy NY 12180, USA
Email: ; Phone: +1-518-253-3762
(2)School of Semiconductor and Chemical Engineering, Semiconductor Physics Research Center, Chonbuk National University, Jeonju 561-756, Republic of Korea
Email:; Phone:+82-63-270-3973
(3)Department of Materials Science and Engineering, Pohang University of Science and Technology, Pohang 790-784, Republic of Korea;
Email:; Phone:+82-54-279-2149
Keywords:Light-emitting diodes; III-V semiconductor device; p-n junction; spontaneous recombination; LED emission spectrum; internal quantum efficiency; light-extraction efficiency; light-escape cone; surface roughening; external quantum efficiency; White LEDs; UV LEDs; phosphor-converted LED; color temperature; solid-state lighting;LED packaging
Abstract
Inorganic semiconductor light-emitting diodes (LEDs)have found widespread use in small-area mobile displays, large-area displays, signaling, signage, and general lighting.The entire visible spectrum can be covered by light-emitting semiconductors: AlGaInP and AlGaInN compound semiconductors are capable of emissionin the red-to-yellow and violet-to-green wavelength range, respectively. For white light sources based on LEDs, the most commonapproachis the combination of a blue LED chip with a yellow phosphor. White LEDs are currentlyused to replaceincandescent and fluorescent sources. In the present review, the properties of inorganic LEDs will bepresented, including emission spectra, electrical characteristics, and current-flow patterns. Structures providing high internal quantum efficiency,namely heterostructuresand multiple quantum wellstructures, will be discussed.Advanced techniquesenhancing the external quantum efficiency will be reviewed,includingchip shaping and surface roughening.Different approaches to white LEDs will be presented and figures-of-meritsuch as the color rendering index andluminous efficacy will be explained.Besides visible LEDs, the technical challenges in newly evolving deep ultraviolet (UV) LEDs will be introduced. Finally, the packaging of low-power and high-power LED chips will be discussed.
Introduction
During the past 100 years, light-emitting diodes (LEDs) have undergone a significant evolution. The first LED emitting in the visible wavelength region was based on SiC compound semiconductor. The device had anexternal quantum efficiencyof much less than1.0%. Today, the external efficiencies of red LEDs based on AlGaInP can exceed 50%; this semiconductoris capable of emitting in theorange, amber, and yellow wavelength range. AlGaInN compounds can emit efficiently in the nearultraviolet (near UV), violet, blue, cyan, and green wavelength range. Thus, all colors of the visible spectrum are covered by semiconductor materials. This has openedup the use of LEDs in areas as diverse as indicator,signage, display, and lighting applications.In particular,as device output powers continuously increase, LEDlampshave reached luminous flux levels of, e.g., 800 lm, comparable tothose of conventional incandescent and fluorescent lamps. Furthermore, LED lifetimesexceeding20000 hours compare favorably with incandescent sources (~500hrs) and fluorescent sources (~5000hrs) thereby contributing to the attractiveness of LEDs.
Inorganic LEDs are generally based on p-n junctions. However, in order to achieve high internal quantum efficiencies, free carriers need to be spatially confined. Furthermore to reduce re-absorption effects, the bandgap energy of the confinement layers should be greater than the bandgap energy of the active region. These requirementsled to the development of heterostructure LEDs that employ different semiconductor materials (i.e. alloy compositions)for the light-emitting active region and the confinement regions. Particularly popular are multiple quantum wells embedded into the active region. The light-extraction efficiency, which measures the fraction of photons leaving the semiconductor chip (and thus become useful, i.e. visible photons), is strongly affected by the device shape and surface structure. Fordevices with high internal-efficiency, the maximization of the light-extraction efficiency is a key challenge.
This reviewconcerns inorganic LEDs and introduces the basic concepts of optical emission. Band diagrams of direct-gap and indirect-gap semiconductors and the spectral shape of spontaneous emission will be discussed along with electrical properties and current flow.Subsequently, strategies to improve light extraction out of the LED chip are presented including surface roughening and chip shaping. Due to total internal reflection at the surfaces of an LED chip, the light-extraction efficiency in standard devices is well below 100%. A particular challenge in achieving efficient emissionisthe minimization ofoptical absorption processes inside the semiconductor. This can be achievedby, for example, covering absorbing regions, such as absorbing substrates, with highly reflective mirrors. The development of white LEDs, including phosphor-based approaches and multiple-LED approaches, is presented and their properties, such as color rendering and luminous efficacy, are discussed. A short section will address the accomplishments and challenges in deep ultraviolet (deep UV) LEDs. Finally, the current state of the art inLED packaging including packages with low thermal resistance, will be discussed.
Recombination in direct-gap and indirect-gap semiconductors
The probability that electrons and holes recombineradiatively is proportional to the product of electronand hole concentration, that is, Rnp. The recombinationrate per unit time per unit volume can bewritten as
(1)
whereB is a proportionality constant, the bimolecularrecombination coefficient, with a typical value of10–10cm3 s–1 for direct-gap III–V semiconductors.Fundamental types of band structures and associatedrecombination processes are shown in Figure 1.
Figure 1: Semiconductor band structure for an (a) direct-gap semiconductor, (b) indirect-gap semiconductor, and (c) indirect-gap semiconductors with a deep isoelectronic impurity level. Indirect radiative transitions have a much smaller probability due to the requirement of a phonon.
During the recombination process, the electron momentum (p = (2m*E)1/2) cannot change significantly because momentum must be conserved and the photon momentum (p = h/) is negligibly small. Thus optical transitions must be ‘‘vertical’’ in nature, that is, electrons recombine only with holes that have the same k value as shown in Figure 1. Efficient recombination occurs in direct-gap semiconductors shown in Figure 1a. However, the recombination probability is much lower in indirect-gap semiconductors because a phonon is required to satisfy momentum conservation, as shown in Figure 1b. The radiative efficiency of indirect-gap semiconductors can be increased by isoelectronic impurities, for example, N in GaP. Isoelectronic impurities can form an optically active deep level that is localized in real space (small dx) but, according to the uncertainty relation, delocalized in k space (large dk), so that the impurity can satisfy the momentum conservation, as indicated in Figure 1c. During nonradiative recombination, the electron energy is converted to vibrational energy of lattice atoms, that is, phonons. There are several physical mechanisms by which nonradiative recombination can occur with the most common ones being nonradiative recombination at point defects (impurities, vacancies, interstitials, anti-site defects, and impurity complexes) and spatially extended defects (screw and edge dislocations, cluster defects). It is quite common for such defects to form one or several energy levels within the forbidden gap. The defects act as efficient recombination centers (Shockley–Read–Hall or SRH recombination centers), particularly if the center’s energy level is close to the middle of the gap.
Optical emission spectrum
The spectral line-width of spontaneous emission canbe calculated to be 1.8kT (see, for example, Schubert, 2006). The theoretical emission spectrum is shown in Figure 2. The line-widthof 1.8kTis caused by the thermalenergy of carriers and is therefore referred to as the “thermal broadening”. It indicates that the line-width broadens with increasing temperature. Italso indicates that the line-width becomes narrower at cryogenic temperatures. For example, at room-temperature, the theoretical line-width of a GaAs LED emitting at 870nm is E=46meV or =28nm.
Figure 2: Theoretical emission spectrum of an LED. The full-width at half-maximum (FWHM) of the emission line is 1.8kT.
The spectral line-width of LED emission is important in several respects. Firstly, the line-width of an LED emitting in the visible range is relatively narrow compared with the range of the entire visible spectrum. Furthermore, the LED emission is narrower than the spectral width of a single color as perceived by the human eye. For example, red colors range from 625 to 730nm (=105nm), which is much wider than the typical emission spectrum of a red LED (25nm). Therefore, LED emission is perceived (by the human eye) as monochromatic. Secondly, optical fibers are dispersive, which leads to a range of propagation velocities for a light pulse comprising a range of wavelengths. The material dispersion in optical fibers limits the “bit rate distanceproduct” achievable with LEDs. The spontaneous lifetime of carriers in LEDs in direct-gap semiconductors typically is on the order of 1–100ns depending on the active region doping concentration (or carrier concentrations) and the material quality. Thus, modulation speeds up to 1Gbit/s are attainable with LEDs.
In practice, the emission line-width is usually broader than the theoretical value of 1.8 kT. Aspectral width of 1.8 kT is expected for the thermally broadened emission. However, due to other broadening mechanisms, the line-width can exceed the theoretical value. A prominent broadening mechanism among ternary and quaternary III–V alloy semiconductors isalloy broadening, i.e. the statistical fluctuation of the active region’s alloy composition. Alloy broadening is independent of temperature and can broaden the line-width to 2 to 5 times of its thermally broadened value. For example, at room temperature,GaInNblue and green LEDs can have a line-width of 3 kT – 10kT.
Figure 3: Emission spectra, at room temperature, of AlGaInP/GaAs red, GaInN/GaN green, GaInN/GaN blue, GaInN/GaN UV, and AlGaN/AlGaN deep UV LEDs.
Figure 3 shows typical LED emission spectra at room temperature. The emission spectra include an AlGaInP/GaAs red, GaInN/GaN green, GaInN/GaN blue, GaInN/GaN UV, and an AlGaN/AlGaN deep-UV LED. The active region of visible-spectrum LEDs is typically comprised of a ternary or quaternary alloy, for example, Ga1–xInxN or (AlxGa1–x)0.5In0.5P. An idealized alloy-broadened emission spectrum has a gaussian lineshape. It may be noted that deep UV LEDs are not yet as mature as visible-spectrum devices. For example, the emission spectrum of the 290nm device shown in Figure 3 has a low-energy shoulder that is due to an undesired parasitic emission, most likely a defect-related emission.
Homostructures and heterostructures
In the vicinity of the unbiased p-n junction plane, electrons originating from donors on the n-type side diffuse to the p-type side where they recombine, as shown in the band diagram of Figure 4. A corresponding process occurs with holes. As a result, a region near the p-n junction, the depletion region, is depleted of free carriers. In the absence of free carriers, the types of charges in the depletion region are ionized donor and acceptor charges. The charges in the depletion region produce a potential, the built-in potential, Vbi, given by
(2)
where NA and ND is the acceptor and donor concentration, respectively, and ni is the intrinsic carrier concentration of the semiconductor. The built-in voltage represents a barrier that free carriers must overcome in order to reach the neutral region of opposite conductivity type.
Figure 4: Free carrier distribution in (a) homostructure under equilibrium conditions, (b) forward-biased homostructure, and (c) forward-biased double heterostructure (DH). Ln/p and n/p are the minority carrier diffusion lengths and lifetimes, respectively. WD is the width of the depletion region. In the forward-biased homostructure, carriers are distributed over the diffusion length (Ln + Lp); in a DH, they are distributed over the thickness of the DH (WDH). The carrier confinement provided by the DH reduces SRH recombination.
The width of the depletion region is given by
(3)
where = r0 is the dielectric permittivity of the semiconductor and V is the diode bias voltage.
An external bias applied to a p-n junction will drop across the depletion region, which is resistive due to the lack of free carriers. A forward bias decreases the p-n junction barrier causing electrons and holes to be injected into the neutral regions with opposite conductivity type. As the current increases, carriers diffuse into the regions of opposite conductivity type and recombine, thereby emitting photons.
The theory of current transport in p-n junctions was first developed by William Shockley in the early 1950s and the equation describing the I–V characteristic is known as the Shockley equation
(4)
where A is the junction area and Dn,p and n,p are the electron and hole diffusion constants and minority carrier lifetimes, respectively.
Under typical forward bias conditions, the diode voltage is VkT/e, and thus [exp(eV/kT)–1] exp(eV/kT). For such forward bias conditions, the Shockley equation can be rewritten as
.(5)
The exponent in the equation illustrates that the current strongly increases as the diode voltage approaches a threshold, which is about equal to the built-in voltage, i.e. VthVbi.
In an ideal diode, every electron injected into the active region will generate a photon. Thus conservation of energy requires that the electron energy eV equals the photon energy h, i.e.
eV=h .(6)
Beyond turn-on, the diode becomes highly conductive and the diode voltage V is about the same as the threshold voltage Vth. The energy of photons emitted from a semiconductor with energy gap Eg is given by
.(7)
Thus equation (6) can be rewritten as
.(8)
For example, a GaAs IR LED emitting at 870 nm has a threshold voltage of about 1.4 V; similarly, a GaInN blue LED has a threshold voltage of about 2.8 V.
Homojunction LEDs have significant drawbacks and hence no longer used. At the present time, virtually all commercial devices are heterojunction LEDs. Heterojunction active regions have several advantages that will be discussed below.
The effect of a double heterostructure (DH) on the carrier concentration is illustrated in Figure4. Under forward bias, carriers diffuse across the p-n junction. In the case of a homostructure under forward bias, minority carriers are distributed over the electron and hole diffusion lengths as illustrated in Figure 4b. In III–V semiconductors, diffusion lengths can be 10µm or longer. The wide distribution of carriers and the correspondingly low carrier concentration (particularly towards the end of the diffusion tail) can be avoided by the employment of double heterostructures. Carriers are confined to the active region of width WDH, as shown in Figure4c. To attain good carrier confinement, the barrier heights should be much greater than the thermal energy kT.
An important advantage of the DH design is illustrated in Figure5. The DH structure has a much smaller number of defects within the recombination region so that non-radiative SRH recombination via deep levels is much less significant in a DH and in quantumwell (QW) active regions. As a result, DH or QW LEDs have a much higher efficiency than homostructure LEDs.
The term “double heterostructures” is frequently used for active layers with thickness of 100 to 500 nm. The term “quantum well” active region is used for active layers with thickness of 2 to 100 nm. Single quantum well (SQW) and multi-quantum well (MQW) active regions provide additional carrier confinement, which can further improve the internal quantum efficiency. If a MQW active region is used, the barriers between the wells may impede the flow of carriers between adjacent wells. Thus the barriers in a MQW active region need to be sufficiently “transparent” (low and/or thin barriers) in order to allow for efficient carrier transport between the wells and to avoid the inhomogeneous distribution of carriers within the active region.
Figure 5: In a double heterostructure (or QW structure), the recombination region, and thus the number of defects contained in this region, is much smaller than in a homostructure. Therefore, non-radiative SRH recombination will play a much smaller role in DH active regions (as well as QW active regions).
Another advantage of DH and QW structures is the optical transparency of the barrier layers. That is, the two cladding layers (cladding the active region) have a wider bandgap than the active region. That is, light generated in the active region cannot be absorbed in the cladding layers because the bandgap energy is greater than the photon energy (Egh).
Light-extraction in light-emitting diodes
Owing to the high refractive index of semiconductors, light incident on a planar semiconductor–air interface is totally internally reflected, if the angle of incidence is sufficiently large. Snell’s law gives the critical angle of total internal reflection. As a result of total internal reflection, light can be “trapped” inside the semiconductor chip. Light trapped in the chipwill eventually be reabsorbed, e.g. by the substrate, active region, cladding layer, or by a metallic contact.
If the light is absorbed by the substrate, the electron–hole pair will most likely recombine non-radiatively due to the inherently low efficiency of substrates. If the light is absorbed by the active region, the electron–hole pair may re-emit a photon or recombine non-radiatively. For active regions with internal quantum efficiencies of less than 100%, any reabsorption event will reduce the efficiency of the LED. For this reason, light should be coupled out of the LED chip with the shorted possible path length inside the chip.