Thermal Management of LEDs: Package to System Mehmet Arik a , Charles Becker b , Stanton Weaver b , and James Petroski c a General Electric Company, Global Research Center, Thermal Systems Laboratory, One Research Circle ES-102, Niskayuna, NY 12309, arik@crd.ge.com b General Electric Company, Global Research Center, Micro and Nano Structures Tech. Lab, One Research Circle Bldg. KW, B1432, Niskayuna, NY 12309 c GELcore, 6180 Halle Dr, Valley View, OH 44125 Abstract Light emitting diodes, LEDs, historically have been used for indicators and produced low amounts of heat. The introduction of high brightness LEDs with white light and monochromatic colors have led to a movement towards general illumination. The increased electrical currents used to drive the LEDs have focused more attention on the thermal paths in the developments of LED power packaging. The luminous efficiency of LEDs is soon expected to reach over 80 lumens/W, this is approximately 6 times the efficiency of a conventional incandescent tungsten bulb. Thermal management for the solid-state lighting applications is a key design parameter for both package and system level. Package and system level thermal management is discussed in separate sections. Effect of chip packages on junction to board thermal resistance was compared for both SiC and Sapphire chips. The higher thermal conductivity of the SiC chip provided about 2 times better thermal performance than the latter, while the under-filled Sapphire chip package can only catch the SiC chip performance. Later, system level thermal management was studied based on established numerical models for a conceptual solid-state lighting system. A conceptual LED illumination system was chosen and CFD models were created to determine the availability and limitations of passive air-cooling. Keywords: Solid-state lighting, LEDs, Thermal management, CFD, Convection. Introduction The introduction of the first practical visible solid state LED occurred in 1962, and was invented by Nick Holonyak of the General Electric Company [Holonyak, 1992]. It was discovered that the wavelength of an infrared GaAs diode could be shifted to the visible spectrum by the introduction of phosphate dopants. The introduction of a compatible large band gap material raises the overall band gap thus shifting emission into the visible spectrum. The wavelength of an emitted photon can be approximated by; bg E hc = λ (1) Where h represents Planck’s constant, c is the velocity of light and E bg is the band gap energy. Thus, raising the band gap of GaAs from 1.4 to 1.9 shifts the emission into the visible red region at approximately 650 nm [Spring et. al, 2003]. These early red devices typically emitted 0.001 lumen per device. Due to their longevity, power requirements, and resistance to shock and vibration the LEDs found uses as indicator and signal applications. Throughout the next two decades advances continued in material development and processing leading to improved efficiency, reliability and additional colors. These advances included changes in elemental proportions, doping and substrate materials resulting in the development of GaAsP LEDs in the orange and yellow portion of the spectrum. During the 1980s refinement of the use of GaAlAs grown on AlGaAs substrates, and multilayer hetero-junction structures in chip fabrication led to an order of magnitude improvement in brightness. Also to appear were AlGaInP on GaAs devices that resulted from the development of organo-metalic vapor phase epitaxy (OMVPE). New materials and techniques enabled high brightness (HB), LEDs from the yellow to red spectrum. During the early 1990s advances were made in OVMPE growth of AlInGaP on GaAs substrates [Maaskant, Third International Conference on Solid State Lighting, edited by Ian T. Ferguson, Nadarajah Narendran, Steven P. DenBaars, John C. Carrano, Proc. of SPIE Vol. 5187 (SPIE, Bellingham, WA, 2004) · 0277-786X/04/$15 · doi: 10.1117/12.512731 64