Do high-power LEDs generate IR heat in the forward direction like a filament lamp?
If one could review the all the lighting industry’s literature of the last 10 years about high power LED lamps and luminaires, would you question this headline? That is, if you read that a high power LED emits no UV or IR and that any heat is created only in the PN junction and then transferred to a heat sink. Okay on the UV part; but the IR part? Read on.
What you will further seem to know for sure is that the PN junction is the hottest point in an LED (or any other power semiconductor for that matter) with the LED mounting substrate and or heat sink being somewhat cooler. How much cooling is needed can be rather accurately calculated if you know a) the heat sink temperature (easy to determine) and b) the junction-to-case thermal resistance (easy to determine for the data sheet).
You likely agree with what you have just read. Having been involved over several decades with silicon power semiconductor and power LEDs in one way or another I also would have agreed… until I did not.
It all started to unravel during some simple experiments relating to a high-power LED array used to illuminate a phosphor-coated sheet (remote phosphor application) to create white of certain properties. The application related to a need for high CRI (over 90) /high CCT (5600K) lighting for TV and motion picture studio lighting.
Operating the array at only 25 watts (about 30% of maximum, and a fan-cooled heat sink and LED substrate both under 35ºC (meaning the junction was, with 100% certainty, under 40-45ºC), I observed, as I positioned the small remote-phosphor sheet over it, that my hand immediately got too hot. In the past I had never paid much attention to this kind of thing. I put my hand virtually on top of LEDs and almost got burned. How could this be, I asked, if the LED substrate is less than 40ºC.
I positioned a thermocouple wire across the LEDs and measured over 125ºC!! Again, how could this be? I measured again with non-contact IR meter and got same temperature. I wondered if this involved some peculiarity associated with these specific royal blue LED arrays. Over a period of time I proceeded to operate a whole range of LED arrays rated from 10 to 100 watts— white, royal blue, green– COB arrays, multi-SMD arrays and large-single-chip types (See Figure 1: A- Bridgelux white RS multi die; B – Philips royal blue LXK multi-SMD ; C – Epistar royal blue multi-die; D – Luminus CBT90 green (single large die) E – Citizen CLL330 white multi-die F – Cree CXA2520 white multi-die).
Figure 1 A variety of LEDs tested with consistent results.
Virtually identical observations. Operating at less than full power, with substrate below 50°C, the top surfaces of the array were measured at well over 100°C and as high as 150°C. Instrumentation error? Really only 35-40°C? My imagination? A drop of water placed on top of any one of the devices boiled off in seconds. The last I knew, water boiled at 100°C.
A sliver of paraffin, specified for a 70°C melting point, placed on any one of those surfaces, melts ”immediately”. During a break in that comparative process, I placed a small square of phosphor coated PET (high temp plastic) over one of the royal blue arrays, while monitoring color temperature (i.e. CCT). Immediately at LED turn-on, the CCT was about 5600K but in less than 30 seconds it rose to 7,000, 10,000 and finally above 12,000K and the light became very “bluish.
I then examined the tiny phosphor sheet and found that the phosphor coating, with an organic, somewhat temperature-sensitive binder, in a couple of places, had actually melted away, allowing blue light passage.
To further determine if this perceived high temperature was a “phantom”, I proceeded to place over all devices at one time or another a small (1” by 1” by .032) piece of aluminum. Without exception, each became a hot plate, elevating the aluminum temperature (as measured by a thermocouple on the top illumination-immune side away from the light). Without exception, the temperature got to between 125-150ºC.
To make it more interesting I put a 3.5-inch tall “Lena” model collimating reflector from Ledil over the blue LED array, and the thin aluminum plate on top of it with a tiny water-filled aluminum cup on top of that plate (Figure 2).
Figure 2 The experiment – A 3.5-inch tall “Lena” model collimating reflector from Ledil over the blue LED array, with a thin aluminum plate on top of it and a tiny water-filled aluminum cup on top of the plate.
I could, with the royal blue light at 75 watts, turn that little piece of aluminum into a hot plate and, even from 3.5 inches away, bring the water to 56°C while the LED substrate was only at 40°C. With a 100 watt Bridgelux RS array also at 75 watts and substrate at about 40°C, the temperature of the water rose virtually the same amount, suggesting that any Stokes-effect heat generation in the phosphor was minimal. Painting the light-receiving side of the aluminum plate black resulted in a 50% increase in the temperature of the plate and of the water.
As further investigation, formal thermographic images were taken at Advanced Thermal Systems with a 100 watt royal blue COB array. For such a thermographic camera image to be taken, the surface point being looked at must have high emissivity. It was therefore necessary to paint a small dot onto the coating using a special high-emissivity material intended for this purpose. Those experienced with taking non-contact IR temperature measurements of shiny surfaces know you have to affix some black tape or other non-reflective and non-transparent material.
This seemingly proper “official” measurement technique resulted in still another anomaly which seemed to be leading to the smoking gun. The black, high-emissivity dot, affixed to a surface which, per all industry conventional wisdom, could not possibly be over 50°C, quickly measured over 200C per the computerized IR image.
A piece of paper coated on both sides with that same high emissivity material, began to smoke and catch fire in less than a few seconds. Since when does 50°C start a fire?
Where does this lead us? Contrary to conventional wisdom or, better said, industry awareness, high power LED’s DO generate IR and the heating resulting from such.
The small percentage of 450 nanometer blue light, generated in the PN junction, is absorbed by the various semi-transparent chip and coating materials before exiting the device. That absorption creates—surprise— some heat.
All materials on earth exhibit various degrees of radiant-energy-absorption properties (witness the effects of sunlight) but it is not necessarily a simple matter to precisely quantify any given situation. When the radiant energy is absorbed, it can be manifested as re-radiated IR. Is that heat really there before I put my hand over it? My hand absorbs some of that re-radiated IR and I perceive heat. Same for the evaporated water drop. But that is “real“ heat. The revelation here that this heat has absolutely nothing to do with the PN-generated heat… heat that is typically transferred to a heat sink via conduction.
What is interesting is that the manner of trying to measure the hot surface can distort the measurement. When we put that high-emissivity paint dot on the surface to get an accurate thermographic image, that dot is unfortunately “too” good. Having the very highest emissivity meant it was most absorbing of the blue light and/or re-radiated IR and as a result skyrocketed in temperature – over 250°C, But make no mistake about it…the LED surface, blue or phosphor coated, for a power level over 10 watts, gets VERY hot.
Finally, in spite of difficulties in establishing exactly how hot is VERY hot ( is it 125° C or is it 175°C?), even when the junction is only 40-50°C, the next step was to quantify reasonably close how much heat was coming from the phosphor-coated or uncoated surface.
With the LED operating at less than half power and the LED substrate below 35°C and junction below 45°C, a closed cardboard cylinder ( an empty Quaker Oats oatmeal container) was placed over the LED array to create an “oven”. If, as the world believes, the only heat generator is the PN junction, reflected in the heat sink temperature, that would mean the air inside the cylinder could not possibly rise above the heat sink temperature.
However, the inside air rose 12 degrees C. proving there was a source of heat other than the junction. Turning off the LED and suspending a power resistor inside the cylinder resulted in the same heat rise when the resistor was powered to a level corresponding to about 8-9% of the power applied o the LED.
What does this mean? It means that if a high power LED arrays is powered to X watts, most of that power, (80-90%) will be manifested as heat in the heat sink. The balance will show up as blue light. But a portion of that will in turn be transformed into heat on the light-emission side, just like a standard old incandescent lamp
Many papers have been written about the Stokes effect, wherein the process of converting blue light to white, via blue light absorption in a phosphor, causes some heat. But the heating described here has nothing to do with Stokes effect since it occurs whether or not a phosphor is present. The effects described here are not easy to detect in 1-2 watt LEDs since the total top-side heat energy, being only 8-9% of that 1-2 watts is small. In a 5mm, that heat is so little and irrelevant to ever be noticed.
Now that we know about it, does it even matter? Well it means that when a phosphor coating is deposited right on top of an LED array such as a 100 watt COB, the phosphor must have high temperature inorganic binders rather than low temp epoxy-type organic binders. Organic binders can indeed degrade with high temperature and cause color shift toward the blue as the phosphor ‘falls apart.’
It also means that if the top surface of a high power LED array is too close to a temperature sensitive object (such as chocolate candy in a display case?) bad things could happen. Also, if one were to put a 100 watt white COB in a very compact, airtight enclosure with an acrylic lens cover very close to the LED emitter, one could find that the acrylic discolored or melted even though the heat sink never really got to a temperature unfavorable to acrylic.
An additional comment— Much is written about the danger of running phosphor too hot and how that might result in lumen degradation. It seems apparent that nearly ”everybody” producing high wattage luminaires, with COB arrays is operating with phosphor surfaced hotter than they are aware of. This fact seems to have escaped acknowledgement or scientific documentation, for a variety of reasons. Yet devices are surviving as long as the phosphors have inorganic binders. It may be time for LED makers and academic researchers to their update information so that engineering assumptions are not based on 2003 one-watt LED phosphor practices.
In conclusion I should note that thermographic measurements were made with a Thermovision A20 camera from FLIR Systems and thermocouple measurements were made with Fluke Hydra 1625A and type J thermocouple.
Special thanks for their analyses and thermal measurement assistance goes to Dr. Kaveh Azar, Dr. Bahman Tavassoli, and Dr. Ning Lei of Advanced Thermal Solutions, to Dr. Steve Johnson, CTO of Solais Lighting Inc for his analysis of various phosphor and chip-related radiation-absorption mechanisms and to Dr Hisham Menkara, CTO of PhosphorTech for his insights into remote phosphor technologies.
For a more comprehensive summary of the physics of heat from radiated light sources, one might refer to Wikipedia under “Infrared” and “Thermal Radiation”.
About the author
Ed Rodriguez is President of OptoThermal Technologies Inc., Wakefield MA. He was founder/CEO of Theta-J Corp and is a veteran of the power semiconductor, power supply and lighting technology industries. He holds 22 patents in those areas, with 8 pending.