Research challenges to ultra-efficient ssl

By Jeffery Howard,2014-05-28 14:28
15 views 0
Research challenges to ultra-efficient ssl



     Laser & Photon. Rev. 1, No. 4, 307?C333 (2007) / DOI 10.1002/lpor.200710019


     Abstract Solid-state lighting is a rapidly evolving, emerging technology whose ef?ciency of conversion of electricity to visible white light is likely to approach 50% within the next several years. This ef?ciency is signi?cantly higher than that of traditional lighting technologies, giving solid-state lighting the potential to enable signi?cant reduction in the rate of world energy consumption. Further, there is no fundamental physical reason why ef?ciencies well beyond 50% could not be achieved, which could enable even more signi?cant reduction in world energy usage. In this article, we discuss in some detail: (a) the several approaches to inorganic solid-state lighting that could conceivably achieve ??ultra-high,?? 70% or greater, ef?ciency, and (b) the signi?cant research questions and challenges that would need to be addressed if one or more of these approaches were to be realized.

     c 2007 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

     Research challenges to ultra-ef?cient inorganic solid-state lighting

     Julia M. Phillips 1,* , Michael E. Coltrin 1 , Mary H. Crawford 1 , Arthur J. Fischer 1 , Michael R. Krames 2 , Regina Mueller-Mach 2 , Gerd O. Mueller 2 , Yoshi Ohno 3 , Lauren E. S. Rohwer 1 , Jerry A. Simmons 1 , and Jeffrey Y. Tsao 1

     1 2

     Sandia National Laboratories, P. O. Box 5800, Albuquerque, NM 87185-1427 Philips Lumileds Lighting Company, San Jose, CA 95131 3 National Institute of Standards and Technology, 100 Bureau Drive, Gaithersburg, MD 20899 Received: 22 August 2007, Revised: 12 October 2007, Accepted: 12 October 2007 Published online: 15 November 2007

     Key words: solid-state lighting; light-emitting diodes; lighting; energy ef?ciency; color mixing; semiconductor optoelectronics; phosphors; nanoscience PACS: 85.60.Jb, 78.30.Fs, 78.67.Bf, 78.30.Hv, 78.55.-m

     1. Introduction

     Arti?cial light has long been a signi?cant contributor to the quality and productivity of human life. It expands the productive day into the non-sunlit hours of the evening and night, and even during the day it expands productive spaces into the non-sunlit (windowless) areas of enclosed dwellings, of?ces and buildings [11].

     Corresponding author: e-mail: jmphill@sandia.gov

     Because we value arti?cial light so highly, we also consume huge amounts of energy to produce it. The production of arti?cial light consumed an estimated 8.9% of total global primary energy in 2003 [32], and an estimated 8.3% of total U.S. primary energy in 2001 [64]. These percentages are large and, coupled with increasing concern over energy consumption, have inspired the development of new and more energy-ef?cient lighting

     c 2007 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


     J. M. Phillips, M. E. Coltrin, et al.: Ultra-ef?cient inorganic solid-state lighting

     technologies. In particular, we are currently witnessing a transition from incandescent technology to ?uorescent and high-intensity-discharge (HID) technologies ?C a transition being accelerated in many nations through legislation [85]. In their current forms, however, all of these ??traditional?? technologies have limitations. Filament-based incandescent technology [42], e.g., emits light approximately as a blackbody. Because of the breadth of the blackbody spectrum, however, the majority of the emitted radiation lies outside the human visual response. Even at a hypothetical ?lament temperature of 6,620 K, for which the blackbody spectrum optimally matches the human visual response1 , the luminous ef?cacy of radiation (lm/W, lumens per watt of radiant energy content of the light, a standard measure of the ??visual ef?cacy?? of radiation), is only ?Ö95 lm/W, about 23% that of the optimal multi-component white light source described in Sect. 2.2. And, in practice, ?laments are typically limited to much lower operating temperatures in the 2,700?C3,200 K range, and hence to a luminous ef?cacy of only 16 lm/W [64], about 4% that of an optimal multicomponent white light source. Both

    glow-discharge-based ?uorescent and HID technology depend on the acceleration of free electrons in a gas discharge, the collisions of those energetic electrons with atoms in the discharge, the resulting excitation of those atoms into excited electronic states, and ?nally the generation of luminescence as those excited electronic states decay. A gas-discharge environment is a complex one, however, and there are many energy-loss channels aside from excitation of atoms into luminescent electronic states (see, e.g., Boeuf [9], for an extended discussion in the context of plasma display panels). As a consequence, less (and often much less) than 50% of the injected electrical energy typically ends up in luminescence at the desired wavelengths. Moreover, for mercury-based ?uorescent technology, there is an additional energy loss associated with the phosphor conversion of ultraviolet luminescence at 254 nm to visible luminescence. The luminous ef?cacies

    of ?uorescent and HID technologies (in an aggregate average over the various lamp types in use in the U.S. in 2001) are about 71 lm/W (lumens per watt of electrical power required to produce the light) and 96 lm/W, respectively [64], about 18% and 24% that of an optimal multi-component white light source. Solid-state lighting (SSL) is an emerging technology with the potential to surpass these luminous ef?cacy limitations, and at the same time to introduce new functionalities and designs in lighting. Based on semiconductor lightemitting diodes, SSL has made remarkable progress in the past decade [47,95,111], to the point where it is now competitive with incandescent technology. There is much research and development worldwide aimed at making SSL competitive with ?uorescent and HID technologies in the coming decade, with ultimate target ef?ciencies in the 25?C


     50% range. These ef?ciencies have the potential to enable signi?cant reduction in the rate of world energy consumption. (A further bene?t is that SSL does not contain toxic materials, whereas the mercury vapor contained in ?uorescent lamps is increasingly causing concern, to the point where used ?uorescent lamps must be treated as hazardous waste in many areas.) There is, however, no fundamental physical reason why ef?ciencies beyond 50% could not be achieved, enabling an even larger reduction in the rate of world energy consumption. As a consequence, it is of great interest to understand and delineate the science and technology challenges that, if overcome, could enable such ef?ciencies. For purposes of discussion we refer to this goal with the term ??ultra-ef?cient SSL,?? indicating an ef?ciency of the order of 70% relative to that of an optimal multicomponent white light source. The remainder of this article is structured as follows. In Sect. 2, we give a general discussion of the ef?ciency of arti?cial lighting, and describe a working de?nition of ??100% ef?ciency?? ?C the highest possible ef?ciency for a multicolor white light source with high color rendering index and reasonable color temperature. Then, in Sect. 3, we describe three distinct approaches to white light creation that could conceivably achieve ultra-ef?ciency ?C an ef?ciency greater than 70% of this theoretical 100% ef?ciency. Each of the three approaches described has signi?cant technology challenges associated it, and, in Sects. 4?C 6 we discuss these challenges. The challenges are divided into three types: those associated with the synthesis and physics of semiconducting materials, discussed in Sect. 4; those associated with photon manipulation in materials and nanostructures, discussed in Sect. 5; and those associated with the synthesis and physics of downconversion materials, discussed in Sect. 6. Throughout, we try to connect these technology challenges with underlying science challenges, many of which have been discussed in more detail in a recent U.S. Department of Energy

    Of?ce of Science workshop report [8]. Conversely, we hope that this article can serve to put the science challenges discussed in that workshop report into the perspective of ultra-ef?cient SSL. Note that the focus of this article is SSL based on inorganic materials. The issues surrounding SSL based on organic materials are treated in [8]. Of the two, SSL based on inorganic materials is considered more mature, though both have differing yet attractive features, and both face differing yet signi?cant scienti?c and technical challenges.

     2. Ef?ciency of arti?cial lighting

     The question of lighting ef?ciency arises often, but is inherently ambiguous. In this section, we discuss two kinds of lighting ef?ciencies: ?rst, the ef?ciency of an overall lighting system; and second, the ef?ciency of the lightgeneration component itself.

     Y. Ohno, unpublished.

     c 2007 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


     Laser & Photon. Rev. 1, No. 4 (2007)


     2.1. Component vs. system ef?ciency

     We start by noting that lighting is an energy service delivered by a complex system composed of a fuel delivery portion, a fuel-to-light conversion portion, and a light utilization portion. Fuel delivery At the ??front end?? of the system, the fuel delivery portion includes all the steps, beginning with a raw fuel (like natural gas), to its transformation into a preferred fuel (like electricity), to the transport of that preferred fuel from point-of-transformation to point-of-use. There are, of course, many different fuels, each having undergone different degrees of re?nement, and each preferred for particular applications. In general, the less the degree of re?nement the more raw energy remains in the fuel. To take the example of natural-gas-to-electricity conversion, only about 1/3 of the raw energy initially present in natural gas remains by the time the electricity arrives at point-of-use. As a consequence, a lighting technology that could convert natural gas into light could in principle have only 1/3 the ef?ciency of a technology that converts electricity into light, yet the lighting systems would have the same net energy ef?ciency. Because of the convenience and ubiquity of electricity in the developed nations of the world, in this review we do not consider further the fuel-delivery portion of the ef?ciency of arti?cial lighting. We simply point out here that in large portions of the less-developed nations of the world electricity is not ubiquitous, while chemical fuels are. Because the conversion of chemical fuel to light is at present extremely inef?cient (less than 1 lumen of light per watt of chemical energy consumed), new approaches to chemicalfuel-based lighting could be very

    important for these nations [32]. Electricity-to-light conversion At the heart of the lighting systems of interest in this article is electricity-to-light conversion. It is this portion that we have referred to in Sect. 1 when targeting ??ultra-ef?ciency?? of greater than 70%. Even this portion, though, can be broken down into two sub-components: one that conditions the electricity from 120?C220 V ac to the 3?C5 V dc typically necessary for the semiconductor devices at the heart of SSL; and one that produces light from 3?C5 V dc electricity. We do not discuss the ?rst component in this article, except to note that it is itself a signi?cant challenge. Today??s state-of-the-art drivers are 65?C90% ef?cient, and rated at temperatures well below those desirable (150 ? C) for driving high-power SSL.2

     Note, however, that ultra-ef?cient SSL has the potential to run ??cooler?? than is the case with current state-of-the-art SSL, since much more energy would be converted to light than heat.


     Instead, it is the second component that is the focus of this article, and that we assume will ultimately be the main determinant of the overall electricity-to-light conversion ef?ciency. In other words, we assume that the ?rst component will ultimately achieve near-100% ef?ciency. If it does not, then the ef?ciency of the second component would need to be increased to achieve a 70% overall electricity-to-light conversion ef?ciency. This is a nontrivial point, since quoted ef?ciencies for traditional lighting technologies normally aggregate both components. Light utilization At the ??back end?? of the system, the light usage portion includes all the steps, beginning from when light emerges from the electricity-to-light conversion device, until it strikes a human eye. This is by far the most inef?cient part of the system. In typical use, the fraction of photons leaving a lamp that ?nally strike the retina of a human eye is probably less than a millionth, even in an enclosed space such as an of?ce. The loss factors include: the less-than-100% re?ectance of non-white objects being illuminated; the small entrance aperture and ?eld of view of the human eye relative to the area and solid angle of the illuminated space; and the less-than-100% percentage of the time a room is occupied while it is illuminated. It is not possible to eliminate all of these loss factors. Through aggressive and sophisticated application of sensor-based real-time controls, however, they could possibly be reduced by as much as an order of magnitude. SSL, in particular, is characterized by fast switching speeds and the potential to focus and direct light while tuning the absolute and relative ?uxes of its component colors. With this capability, one might imagine a sensor-based control system that tailors in real-time the placement, quality and luminous ef?cacy of

    light to its use by humans. Clearly, there is signi?cant future opportunity in this area (e.g., sensors, micro-opto-electro-mechanical systems (MOEMS), human visual perception, lighting architectures and controls), and one can anticipate more attention will be paid to it in the coming decades. Beyond this short discussion, however, we will not consider further the lightutilization portion of the ef?ciency of arti?cial lighting.

     2.2. Auxiliary characteristics of white light

     In this and the next sections, we consider the theoretical maximum ef?ciency with which electricity can be converted to white light. The central ambiguity in this question is the fact that there are many ways to make white light. For example, on the Commission Internationale de

     This relaxes the requirements on operating temperature for all aspects of electricity-to-light conversion, including power conditioning.


     c 2007 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


     J. M. Phillips, M. E. Coltrin, et al.: Ultra-ef?cient inorganic solid-state lighting

     l??Eclairage (CIE) chromaticity diagram [73], the Planckian locus3 which de?nes white is a line, and any point on that line can be created by superposing any one of an in?nite number of combinations of two or more saturated colors at the perimeter of the diagram [73]. Because the responsivity of the eye and the characteristics of semiconductor materials both vary with wavelength, each combination will achieve a different ef?ciency. Hence, out of the many ways of making white light, it is necessary to specify auxiliary characteristics required of the light before one can de?ne a maximum ef?ciency. Correlated color temperature A ?rst characteristic is the chromaticity position along the Planckian locus ?C the correlated color temperature (CCT) of the white light. For traditional lighting technologies, CCT spans a wide range, from 2,700 K to 6,500 K or so. For incandescent lamps, the luminous ef?cacy of radiation (LER)4 is higher for higher CCT (up to 6,620 K, as mentioned in Sect. 1). However, for white light sources based on discrete component colors (as obtained from light-emitting diodes (LEDs)), the luminous ef?cacy of radiation increases slightly as the CCT decreases [93].5 Hence, there is a slight luminous ef?cacy advantage to lower CCT??s. In our analysis, therefore, we choose a common and relatively low (3,000 K) default CCT. This CCT is known as ??Warm White?? for ?uorescent lamps and is in the most common range of lamp color used in residential lighting in the USA.6 Color rendering index A second characteristic is the ability of the white light to render the colors of objects in the environment to the huThe locus of points on the

    chromaticity diagram representing Planckian (or blackbody) radiation at various temperatures. 4 The luminous ef?cacy of radiation (LER, lm/W) is the ratio between the luminous ?ux (in lumens) and the optical power of radiation (in watts, W). This metric differs from a related but distinct metric, the luminous ef?cacy of a source (LES, lm/W), which is the ratio between the luminous ?ux (in lumens) and the input electrical power (in watts, W) used to produce that luminous ?ux. The LES is essentially the product of the LER and the radiant ef?ciency (ratio of optical power to input electrical power) of the light source. 5 The physical reason is that the human visual system is much less sensitive to the blue component than to the red component at the wavelengths necessary to produce a white chromaticity. As CCT decreases, the blue contribution decreases (and the red contribution increases), so the overall luminous ef?cacy increases. 6 Note that the chromaticity position transverse to the Planckian locus (referred to as ??uv or Duv [3] is also important. If the chromaticity position is above the Planckian locus, e.g., luminous ef?cacy increases but the light source will appear unacceptably greenish or yellowish. Only very small shifts from the Planckian locus are acceptable for general lighting, so in this article we have assumed Duv = 0.000 (exactly on the Planckian locus).


     man visual perception system in an accurate and pleasing manner. Here, we use the internationally agreed upon metric the Color Rendering Index (CRI) [17]. The values of CRI reported in this article are those of the general color rendering index (Ra ).7 In general, the higher the CRI, the lower the possible luminous ef?cacy. This is because the human visual response is not ?at, but peaks at 555 nm. Light composed of just that wavelength would have a very high luminous ef?cacy of radiation (683 lm/W) but would render well only objects that re?ect at that wavelength. To render well objects that re?ect at other wavelengths, a light source must include light that spans a wider wavelength spectrum, but that the human eye is less sensitive to. Hence, the wider the included spectrum, the better the color rendering, but also the lower the luminous ef?cacy. For our analysis, we use a CRI of 90 as the default requirement. Such a CRI is considered excellent and would satisfy virtually all white-light applications. Number of component colors A third characteristic is the number of component color sources used to create the white light. Solid-state lighting is ultimately based on relatively narrowband light emission from inorganic or organic semiconductor materials; the more materials used, the greater the number of component colors that can be mixed and used to create the white light. Furthermore, the more component colors used, the higher the obtainable CRI, as the human eye will be able to

    discriminate between the colors of a wider range of objects in the surrounding environment. In practice, the spectral re?ectances of most objects in the environment vary relatively broadly with wavelength. This is exempli?ed by the Munsell samples, illustrated in Fig. 1, upon which Ra in the CRI scale is based [17]. Hence, it is not necessary to use a great many component colors to produce good color rendering: ?ve can provide CRIs that reach 99 and four can provide CRIs that reach 97, though three can only provide CRIs of 85 or less [110]. Since on the one hand, for a CRI of 90, three are insuf?cient, and on the other hand there is no advantage in going to ?ve, we use four component colors in our analysis of all-semiconductor white light: R (red), Y (yellow), G (green) and B (blue). Linewidths of component colors A fourth characteristic is the linewidths of the component color sources. As mentioned above and also illustrated in Fig. 1, the re?ectances of the Munsell samples used to calculate Ra vary smoothly with wavelength. Thus, for component colors whose linewidths are fairly narrow (e.g., 20 nm, or less), the re?ectances they sample (and the resulting CRI of the overall white light source) do not depend much on their linewidths.

     This method is based upon color-rendering tests using eight Munsell samples with re?ectance spectra that are relatively nonsaturated and evenly distributed over the range of hues.


     c 2007 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


     Laser & Photon. Rev. 1, No. 4 (2007)



     Relative Power


     530 nm

     1.0 0.5 463 nm 0

     573 nm 614 nm

     Reflectance Factor


     0.6 0.4 0.2 0 400 500 600 700

     8 1 7

     red color has a vanishingly narrow linewidth (for a CRI of 90). In contrast, for the yellow and green components, for which the human eye response is strong and relatively broad, narrower linewidths are not so critical for ef?cacy. Likewise, for the blue component, whose lumen contribution to the white light is low, narrower linewidths are also not critical. Hence, because there is no CRI penalty for narrower linewidths, but there is a slight luminous ef?cacy advantage, in our

    analysis we chose essentially delta-function linewidths of 1 nm as the default. Such 1-nm linewidths are much narrower than the 20-nm or so linewidths typical of spontaneous-emission devices (kT ?? 6 to 10 nm) such as light-emitting diodes (LEDs), and approach those typical of stimulated emission devices such as (multi-mode) laser diodes.


     6 3 4 5

     2.3. Luminous ef?cacy and color rendering

     With these four characteristics speci?ed, we used a sophisticated white-light simulator (described in slightly more detail in Sect. 3.1) to vary the four center wavelengths of an RYGB white light source so as to maximize luminous ef?cacy of radiation for a given CRI value, while constraining the chromaticity to lie on the Planckian locus at a CCT of 3,000 K. The results are shown graphically in Fig. 2. The results are also shown numerically and with greater completeness in Table 1: center wavelengths, luminous ef?cacies, and fractional contributions to the total lumen output and total power (Watt) consumption for all four RYGB components. Fig. 2(a) illustrates how, as anticipated, the maximum obtainable luminous ef?cacy decreases as CRI increases. For a CRI of 90, the maximum luminous ef?cacy of radiation (LER) is 408 lm/W (dashed horizontal line in both halves of Fig. 2). This luminous ef?cacy, then, is the value we take as our working de?nition of ??100%?? ef?ciency, and is in reasonable agreement with previous Monte Carlo simulations [12, 110]. If a light source were constructed with 100% wall-plug ef?ciency, the maximum obtainable luminous ef?cacy of the source (LES) would thus be 408 lm/W (that is, per watt of electrical power in). Our working de?nition of ??ultra-ef?cient?? SSL, in turn, is 70% of this luminous ef?cacy, or 286 lm/W (LES). For the remainder of this article, we will refer to 286 lm/W as being the threshold luminous ef?cacy that de?nes ??ultraef?cient?? SSL. Fig. 2(b) displays the center wavelengths9 that maximize luminous ef?cacy of radiation for the luminous ef?cacy and CRI combinations illustrated in Fig. 2(a). The center wavelengths that enable our de?ned ??100%ef?cient?? 408 lm/W (LER) at a CRI of 90 are: R (red)

     Note that white-light luminous ef?cacy of radiation maxima are relatively ?at for variations of a few nm around these center wavelengths, but fall off rapidly for larger variations.


     Wavelength (nm)

     Figure 1 Top panel shows the relative power spectrum of the ideal RYGB SSL white light source discussed in Sects. 2.2 and 2.3 with a CRI of 90, a luminous ef?cacy of 408 lm/W (LER), a CCT of 3000 K, and component-color linewidths of 1 nm. Bottom panel shows the relative re?ectance spectra of the eight Munsell samples used in calculating

    the general color rendering index (Ra). Even though the SSL source is far from continuous, it samples the re?ectances of the Munsell samples at enough different wavelengths to render colors with high CRI.

     This is a key point that we reiterate for emphasis. Within the existing framework for calculating CRI, there is no penalty for narrow (even ?Ä-function-like) linewidths.8 Instead, as discussed above, CRI is much more sensitive to the number and wavelength spread of the component colors. Luminous ef?cacy, however, does depend on linewidth even at 20 nm or less, particularly with respect to the red component. A narrower linewidth for the red component minimizes the spillover of light to longer wavelengths, where the human eye response falls off rapidly. The percentage decrease in overall ef?ciency as the red linewidth increases is roughly 0.15%/nm, so an RYGB white light source whose red component has a 20 nm linewidth will be roughly 3% less ef?cient than a source in which the

     8 Note, however, that CRI is known to have some de?ciencies when used for narrowband sources, and alternative metrics beyond CRI are being actively developed (see, e.g., [20]). Use of these alternate metrics could alter some of the speci?c conclusions of this article.


     c 2007 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim


     J. M. Phillips, M. E. Coltrin, et al.: Ultra-ef?cient inorganic solid-state lighting

     Luminous Efficacy of Radiation (lm/W)



     408 lm/W



     300 0 20 40 60 80 100 400 500 600 700

     Color Rendering Index

     Wavelength (nm)

     Figure 2 Maximum luminous ef?cacies of radiation of CCT = 3,000 K white light composed of 100%-ef?cient 1-nm-linewidth (essentially delta-function) RYGB component colors: (a) for various CRIs, and (b) versus the RYGB center wavelengths associated with those maximum luminous ef?cacies. As discussed in the text, high luminous ef?cacy of radiation is achieved at the expense of high CRI, and vice-versa. For a CRI of 90, considered excellent, the maximum luminous ef?cacy of radiation is 408 lm/W, and is the working de?nition of ????100%???? ef?ciency used in this article.

     Table 1 Center wavelengths, lumen fractions, luminous ef?cacies of radiation, and Watt fractions corresponding to the various components

Report this document

For any questions or suggestions please email