What is epitaxy?

At the heart of every LED is a stack of multiple semiconductor layers, each with a different composition and doping concentration. The semiconductors used in LEDs differ from those in integrated circuits both in terms of chemistry (LEDs use III-V compounds like GaAs and GaN instead of Si) and the manufacturing process. The layers of an LED are deposited using a technology called organometallic vapor-phase epitaxy (OMVPE), also known as metalorganic chemical vapor deposition (MOCVD), which has clear advantages over competing techniques when considering factors such as the control of growth rates, uniformity and throughput.

The term “epitaxy” comes from the Greek epi (upon) and taxis (arrangement), and refers to growth on a substrate uniquely determining the orientation of the grown material. Epitaxy ensures that all locations on the wafer have the same atomic orientation and avoids defects that reduce efficiency. The term “epi” is short for epitaxial/epitaxy, and is commonly used in the field. In an OMVPE reactor, group III organometallic and group V hydride precursor molecules are transported to the surfaces of heated substrate wafers by a carrier gas flow, typically hydrogen or nitrogen. A complex series of reactions and diffusion processes take place near the wafer surface, beginning with decomposition of precursor molecules and ending with the incorporation of new atoms at regular, specific locations on the surface defined by the atomic structure of the substrate.

The various competing physical and chemical processes in OMVPE result in a highly complex system, and growing any given layer of an LED with uniformity and reproducibility requires simultaneous optimization of a large number of input parameters (pressure, temperature, multiple gas flow magnitudes and ratios). The optimum growth conditions for each layer must be developed individually, and the LED performance has a high sensitivity to not only the thickness, composition, doping (including residual impurities), and quality of interfaces for each specific layer, but equally to the specific attributes and process conditions used for growing both the preceding and subsequent epi layers in the structure. 

Lumileds has a long history in OMVPE technology going back to its roots in Hewlett-Packard, and is now the world leader in epitaxy manufacturing. This is exemplified by Lumileds industry-leading performance and strong track record of consistent improvements, world class on-wafer color uniformity and yields, and advanced process repeatability on 6” wafers, where Lumileds led industry adoption. Lumileds built this leading position by developing extensive in-house expertise in OMVPE reactor hardware (including heat and mass flow simulations, and in-situ process metrology), III-V crystal growth and materials science, LED device physics, applied statistics, and advanced materials characterization. A broad understanding of the correlations between in-situ metrology data, measurable materials properties, sophisticated device simulations, and final device performance, combined with the confidence that layer structures can be well reproduced across experiments, gives us a competitive advantage in research and development.

III-Nitrides

The vast majority of white LEDs produced today involve a blue LED that pumps a phosphor coating on the LED that down converts some of the blue light into longer wavelengths, the combination of which is observed as white light. The blue LED is made of III-Nitride epitaxial layers, which can be tuned to emit in a wide color range from ultraviolet to green by changing the InN, AlN and GaN contents. III-Nitride LEDs have specific challenges including the limited thickness to which the light-emitting quantum wells (QWs) can be grown without causing defects due to a lattice mismatch between InGaN and GaN, the tendency for the QW material to break down under certain growth conditions due to a miscibility gap between GaN and InN, and the intrinsic polarization-induced electric fields in the QWs that separate the electrons and holes, which decreases the chances of them recombining to form a photon.  These challenges limit the structures that can be grown before material quality or radiative efficiency is hurt. Another major challenge of the III-Nitride materials system is the strong Auger non-radiative recombination that lowers LED efficiency, especially in high-power LED devices. Auger recombination was first proposed by Lumileds as the main mechanism contributing to high-current efficiency droop in commercial III-Nitride LEDs (the reduction in LED efficiency as the drive current is increased). 

FIGURE 1: RELATIVE WALL PLUG EFFICIENCY IMPROVEMENTS AND CORRESPONDING GENERAL TREND LINE OVER TIME. ALL VALUES ARE NORMALIZED TO THE WPE NUMBER AT THE BEGINNING OF 2006.

Despite these challenges, Lumileds has demonstrated significant improvements in epitaxy performance by using its extensive in-house expertise in materials science, epitaxial growth and hardware, LED device physics, and device simulation, to continuously engineer improved solutions to each constraint. For example, through active region design, Lumileds has been able to further delay the onset of efficiency droop within each QW with each new epi generation.  Consequently, the performance of Lumileds’ III-Nitride epitaxy has improved year-over-year, making Lumileds a leader in terms of epi performance within the LED industry (Figure 1).

III-Phosphides

High-brightness visible LEDs operating at long wavelengths, amber (590nm) to deep red (660nm), are all based on the AlInGaP material system. The development of AlInGaP LEDs started in the early 1990s at Hewlett Packard, which pioneered the use of a GaP transparent window layer and produced the first high-brightness LED in the industry. All current AlInGaP LED companies employ this layer in their AlInGaP LED portfolio to provide competitive brightness. The AlInGaP material system is conveniently lattice matched to GaAs at compositions around 50% In, which remains fixed while the Al/Ga ratio is varied to change the band gap and wavelength of emission without affecting the lattice constant. Lattice matching is critical for most light-emitting semiconductor devices, the notable exception being III-Nitrides. If an AlInGaP layer is grown with a composition that is not lattice matched to GaAs, defects form in the crystal, which reduce the radiative efficiency of the device. Using lattice matched compositions near 50% In, one has complete flexibility to design the thickness and band gap of the epitaxial layers to achieve optimal device performance.

While the LED emission color may be tuned by varying the %Al of the light emitting layer, there are practical limits defined by the electronic properties of the material system. AlInGaP alloys undergo what is referred to as a transition from direct band gap to indirect band gap at Al content greater than 53%, which results in a dramatic drop in radiative efficiency. Indirect band gap materials, such as Si, are extremely inefficient light-emitters due to the requirement of a phonon to assist the light-emitting transition in the semiconductor. Thus, this direct to indirect band gap transition defines the lower wavelength limit of AlInGaP materials.

The main development challenge for AlInGaP LEDs is that the radiative efficiency varies significantly with wavelength and current. Optimizing an epitaxial structure for one particular application therefore creates trade-offs in another. For example, it is difficult to confine charge carriers to the light-emitting region of the device at shorter wavelengths, as the direct to indirect band gap transition prevents one from increasing the energy barrier that confines carriers to a given light-emitting layer. Furthermore, the light emitting layer itself starts to have a band gap similar to the barrier layer at shorter wavelengths, further reducing carrier confinement. These issues are not present at longer wavelengths. Thus the design that is ideal for a longer wavelength may be inadequate for short wavelength LEDs.

Lumileds has developed epitaxy processes and device designs to circumvent many of these challenges, such as strain engineering to push the limits of the direct to indirect band gap transition, as well as epitaxial layer designs that are tailored for particular colors or current ranges. Lumileds has also pioneered the use of 6” substrates for the growth of AlInGaP LEDs, utilizing our knowledge of reactor hardware design and epitaxy process to achieve exceptional on-wafer uniformity.