Because LEDs are so often used as indicator lamps on electronic equipment and in low-power lighting applications such as flashlights, Christmas decorations or rear lamps on bicycles, you might think that they are not hazardous to the eyes. However, high-power LEDs have been developed for uses such as streetlights and car headlights, and these certainly can cause retinal damage if used inappropriately.The output of an LED intended for lighting is usually quoted in lumens. Modern high-power devices, such as the Luxeon K2 line or the Cree Xlamp 7090 XR-E series, can produce more than 100 lm of output from a single LED chip. A typical 100-W bulb produces an optical output of about 1750 lm, however, so why are LEDs more hazardous?The answer lies in the extremely small size of the emitting region of the LED and, to some extent, the different spectrum of the emitted light. Because the emitting region is small (3 mm2 in the case of the Luxeon K2, for example), the eye can focus the light to form a small image on the retina with a high power density capable of causing localized damage (Figure 1). This damage can be either photothermal or photochemical.Figure 1. The small emitting area of an LED chip can be focused to an even smaller area on the retina. Photothermal damage, caused by local heating, is associated with the red end of the spectrum. If the temperature of the retina is raised by more than about 10 °C, damage occurs. Photochemical damage, a photon-induced molecular change, is associated with the blue end of the spectrum. The absorption of blue photons interferes with the visual process and ultimately leads to the death of the photoreceptor cells.Some white light LEDs are blue diodes with a phosphor added to convert some of the blue light to other wavelengths. There is, however, a pronounced spike in the spectrum in the blue light hazard region. The phosphor is deposited on the LED chip itself in the case of the Luxeon K2 but is dispersed throughout the covering lens of the Cree Xlamp, producing a somewhat larger emitting region.Currently, white LEDs present little risk of causing retinal damage in normal use, but one should not stare directly at them for prolonged periods — say, tens of seconds. The problem is more acute with blue LEDs, which emit essentially all of their output in the spectral region associated with the retinal blue light hazard. This is formally defined as the 300- to 700-nm range, but the 400- to 500-nm region is the most dangerous. The Luxeon K2 royal blue LED peaks at around 450 nm, with a width at half height of about 20 nm, and can emit as much as 750 mW of radiant power from a chip less than 3 mm2. Despite being a powerful emitter, it is not perceived as such because of the reduced sensitivity of the eye at these wavelengths (Figure 2). This and similar products from other manufacturers certainly present a risk of retinal damage if used carelessly. Figure 2. A white light LED spectrum shows the pronounced spike in the retinal blue light hazard region and the low sensitivity of the eye to blue light. The real problem is probably not with LEDs in general use. The bigger problem occurs in development labs where researchers or technicians may handle LEDs while testing possible applications but are unaware of the possible hazards. LEDs can be overdriven to emit much more than their rated output for quite some time before failing. It is vital that safety hazards are identified, that good practices are established and that information is widely publicized now, while hazards are minimal and usage of LEDs is relatively low.The current LED safety standards are unsatisfactory in that they are not mandatory in the US, and European standards are not really designed for LEDs. The 60825-1:2001 laser standard of the International Electrochemical Commission (IEC) says that, wherever the word “laser” occurs, “LED” can be substituted. In 2001, the FDA accepted the IEC classification and labeling for lasers, but LEDs were still excluded.LEDs are not lasers, however. Those with white light, in particular, cover a wide spectral range, from short wavelengths with predominantly photochemical hazards to long wavelengths with potential thermal hazards. The IEC 60825 standard does not really address such broad spectrum sources except by treating them as extreme examples of multiwavelength sources.Alternatively, LEDs can be classified under standard S 009/E:2002, titled “Photobiological Safety of Lamps and Lamp Systems,” issued by the International Commission on Illumination (CIE). The scope of this document includes LEDs but excludes lasers. It is written for the characterization of physically extended broadband sources and, hence, is perhaps more in line with typical LED characteristics. However, the standard uses radiance measurements (power per unit area per unit solid angle) for the specification of retinal hazards, rather than the simpler concept of irradiance (power per unit area) as used in the laser standard. It is, therefore, sometimes perceived as a more difficult document by those unfamiliar with radiance. In practice, however, current LED chips are so small that many can be treated as “small sources” for the purposes of evaluation of the blue light retinal hazard, and the calculation can be performed in terms of irradiance rather than radiance.The CIE standard is much narrower in scope than the IEC 60825 standard. In particular, having classified your product in a given group, you receive no recommendations as to how it should be labeled or how it should be handled. Manufacturers using LEDs in consumer products will probably lean toward this standard to avoid putting yellow warning labels on consumer products that contain LEDs, as required if classified by the IEC laser standard. The measurement of the effective radiance of a product such as a flashlight containing a number of LEDs and beam-shaping optics by CIE S 009/E:2002 protocols is a nontrivial task, however, and few manufacturers have done such measurements as yet.New standardsStandards under discussion by the IEC will propose that LEDs used for fiber optics, free-space communications and medical applications be classified according to the 60825 laser standard, while those used for general lighting, signaling or indicating, intense pulsed light sources, surveillance, IR illumination and UV lamps should be classified according to the S 009/E:2002 lamp standard. Explicit advice will be given in regard to labeling, even for products classified as lamps.This is a reasonably sensible compromise but would still leave us with two standards with different classifications of sources. It might be better to create a single set of classification levels that treats four classes according to whether the source is broadband or narrowband and whether it is a small source or an extended source.One useful suggestion under discussion is that products using a small source such as an LED be classified according to their radiance rather than the properties of the final product. This could be a great boon to users of LEDs. If LEDs were classified by the manufacturers according to such a standard, the final assembler would be freed from much arduous measurement. Note, however, that LED arrays still would have to be tested by the final assembler, and this is not a simple task. Because most applications use arrays rather than single LEDs, this concession will be of limited value.The standard is unlikely to be published until well into the year and will not be in general use until much later. In the meantime, users of high-power LEDs should press manufacturers to make full safety information and advice readily available. Information about the apparent source size and position also should be provided, together with spectrally resolved radiance data to help users perform their own safety analyses according to CIE S 009/E:2002 or IEC 60825-1.There should be a concerted attempt to alert users to the possible hazards associated with inappropriate use of LEDs. Companies working with the devices must take the responsibility of warning their employees of the possible dangers. At the very least, everyone should be made aware that it is unadvisable to stare directly at a high-power LED source.Meet the authorGeoff R. Davies is a retired professor of physics and a senior consultant at Lucid Optical Services Ltd. in Garsdale, UK; e-mail: geoff@lucidos.co.uk.Radiometry and Photometry PrimerWhen dealing with optical safety, we usually are concerned with how much radiated energy is absorbed at some surface. The risk is usually related to the amount of incident power per unit area of surface. For example, the heating of your skin when exposed to strong sunlight depends upon the radiant power per unit area of skin. The power per unit area is known as the irradiance and is measured in watts per square meter. Much of the heating effect, however, comes from infrared radiation, which is invisible to the eye. If the infrared radiation were absorbed before reaching us, the sun would appear just as bright, but we would not get as warm. Thus, we need to make a distinction between radiometry and photometry.Radiometry is the measurement of the energy content of that part of the electromagnetic spectrum within the frequency range 3 × 1011 to 3 × 1016 Hz, or 1 mm to 10 nm in wavelength. This covers the infrared, optical and ultraviolet regions of the spectrum.Photometry relates to that part of the energy perceived by the human eye as light. It is thus limited to the range of ~360 to 830 nm. Photometric measurements are weighted by the relative sensitivity of the eye to different wavelengths to express the perceived effect of the radiation, rather than the absolute energy content. The output of devices intended for lighting is specified in lumens. This is the radiated power corrected for the spectral response of the eye.The sensitivity-weighted power per unit area is known as the illuminance rather than the irradiance. Illuminance is thus measured in lumens per square meter, or lux. The level of lighting in an office or on a highway, for example, is specified in lux. A 100-W lightbulb at 2 m produces about 500 lux.Unfortunately, safety standards always have to deal in “real” power, measured in watts, but product specifications are given in terms of the optically effective power, measured in lumens. Given the optical output of an LED in lumens, we can make the conversion to watts only if we know the spectral distribution of the source.By definition, green light at 555 nm, where the eye is most sensitive, has 683 lm/W or 1.46 mW/lm. For high-power white LEDs, we get about 300 lm/W or 3.3 mW/lm, reflecting the lower sensitivity of the eye to the red and blue ends of the spectrum. The radiant power of a white LED emitting 80 lm is thus about 260 mW. Note that this is the electromagnetically radiated power, not the electrical power consumed by the LED, which will be many times greater — typically a few watts.The number of lumens a device is rated as having does not actually tell us how bright it appears to be. Somewhat confusingly, neither does the intensity of a source, which usually means the power per unit of solid angle. The luminous intensity of a source (that is, corrected for the eye’s response) is measured in candela, which is equivalent to lumens per steradian. This is a useful measure for small sources because it remains constant with distance when the illuminance is falling as 1/r2.How, then, do we measure brightness? Obviously, a small source will appear brighter than a large source emitting the same power. What is not quite so obvious is that something emitting its power over a narrow range of angles will look brighter than something that emits the same power over a wide range of angles, because only that proportion of the light that enters the eye is seen.The brightness of a source is thus measured by the power per unit source area, per unit solid angle of emission. This is known as the radiance or, if corrected for the spectral response of the eye, the luminance of the source. Radiance is measured in watts per square meter per steradian; luminance is measured in lumens per square meter per steradian, candela per square meter or lux per steradian. All of these combinations of units are equivalent to the unit called the nit (from the Latin niteo, meaning “I shine”). The luminance of a tungsten filament source is about 20 × 106 nits, or 20 Mnt. Modern white LEDs achieve a similar level of brightness.Radiance is important in safety because it is conserved in ideal optical systems. We cannot deduce the irradiance at the retina from a knowledge of the irradiance at the surface of the eye. It can, however, be related to the radiance at the eye’s surface and the subtense angle of the source. Hence, safety limits are set in terms of radiance and source subtense angle.