1. Introduction

Owing to significant advances in solid state illumination technology [1,2], LEDs have the potential to replace the traditional light sources by which we have seen the world for many decades. Progress in solid state photon generation (e.g. Gallium Nitride LEDs) and package photon extraction, as well as improved efficiency in down-converting phosphors, have resulted in present day devices at 80-100 lm/watt, now meeting or surpassing the luminous efficacy of incandescent, fluorescent, and high intensity discharge (HID) light sources (Fig. 1(a) ). Efficacies of 135 lm/watt for high power LEDs are forecasted within the next several years, with the potential of more than 200 lm/watt [1,2]. Similarly LED brightness has improved from 10³ cd/m² in the early 90’s to forecasted levels of near 10⁸ cd/m² in the near future (Fig. 1(b)), suggesting that LEDs could match the brightness of unfiltered ultra-high performance discharge lamps used in projection applications. Advances in large-scale manufacturing efficiencies and package designs have driven the cost of an LED Lumen down to levels that now compete with fluorescent light sources. Accompanying this affordability trend is a strong relationship between $/LED-Lumen and die size: larger-die, high power LED devices tend to have a much lower cost per lumen than comparable small-die devices [2].

 

Fig. 1 Improvements in (a) luminous efficacy η_L (lm/W) and (b) luminance (cd/m²) for conventional and solid state light sources over time. LED data points are color coded corresponding to emission wavelength. Ranges required for various applications are shown to the right of chart (b). Source: Michael R. Krames, Member, IEEE, Oleg B. Shchekin, Regina Mueller-Mach, Gerd O. Mueller, Ling Zhou, Gerard Harbers, and M. George Craford, Fellow, IEEE; Status and Future of High-Power Light-Emitting Diodes for Solid-State Lighting; IEEE J. Display Technol., Vol. 3, No. 2, June 2007. (c) Area conversion ratio, defined as the ratio of the emissive area of the system to that of the input source, for fluorescent and solid state light sources.

Download Full Size | PDF

In parallel with these developments, numerous trends are driving luminaire and LCD display backlighting technologies. Energy efficiency has taken on increased societal importance [3]. The need to efficiently light homes, offices and outdoor spaces with long-life, low-maintenance lighting systems is placing challenging requirements on materials and light sources. Large-area backlight applications such as found in liquid crystal displays (LCD) TVs, which did not exist until 10 years ago, are extremely challenging, requiring light to be distributed over large areas in an increasingly thin form factor [4]. Advances in fluorescent lamp technology have driven recent increases in luminaire and LCD backlight performance. These fluorescent “low” brightness light sources (~10³ to ~10⁴ cd/m²) can be used in a variety of configurations to achieve efficient, spatially uniform emission of light from a backlight or luminaire output surface. A widely used method of achieving light distribution with fluorescent lamps has been to place an array of lamps behind a diffuse, transmissive/reflective panel in a “direct lit” light mixing architecture (Fig. 1(c)). Here the ratio of the emissive area of the light mixing cavity, to that of the emissive area of the light sources themselves (area conversion ratio), is typically on the order of 10. In contrast, with LED brightness forecast to reach 10⁸ cd/m², the area conversion ratio would be on the order of 30,000, to reach similar emissive area brightness. This highlights a fundamental challenge that LEDs present in light management: how does one distribute light from a source of a few square millimeters to an area of a square meter or more?

An envisioned, high performance LED light distribution system would conserve lumens and deliver highly mixed (for white or red, green, blue (RGB) sources), uniform light output within a relatively small mixing cavity volume, i.e. preferably in a thin or near-planar geometry. Ideally, it would also enable control of the polarization and angular distribution of its output to deliver application-specific “usable” light. Finally it would need to be compatible with high-volume, low-cost manufacturing.

Volume-efficient light distribution from localized, high-brightness sources can be achieved by controlling three design elements which affect system uniformity and efficiency: light injection geometry, light transport and light extraction (Fig. 2 (a) ). Light injection by a lens or external optic acts to shape an LED’s output pattern to match the aspect ratio of the mixing cavity and the light transport capabilities of the optical materials. Light transport is the mechanism by which light propagates away from the source to the remote regions of the cavity, and light extraction acts to perturb the light transport in a way that allows for the light to uniformly exit the mixing cavity.

 

Fig. 2 (a) Hollow mixing cavity consisting of a collimated LED light source, a front transflector of hemispheric reflectance R_Hemi,F and a back reflector of hemispheric reflectance R_Hemi,B that together provide light transport and light extraction. In this illustration, the front transflector transmits only the chosen polarization (blue arrows) within a desired angular cone while the orthogonal polarization (red arrows) is recycled. (b) Cavity efficiency computed as a function of R_Hemi,F × R_Hemi,B for various levels of material absorption per bounce (effects of light source absorption not included). The area in green represents the preferred material property combination for an efficient and uniform hollow cavity.

Download Full Size | PDF

One practical application of the above principles for creating efficient, planar light mixing cavities uses transparent solid media to transport and mix high brightness LED light sources: controlled light injection into low-loss material can efficiently transport light by total internal reflection (TIR), and subsequent extraction features can “meter out” the light to give uniform emission. This approach, however, is not well suited to the use of the smallest number of LEDs, as the low spatial spreading of light in solid media inhibits light mixing, and the weight and cost of solid media mixing cavities has limited widespread adoption for large-area systems. In addition, optically coupling the light source to the TIR medium severely constrains light injection angles (that enable TIR), while uncoupled light sources tend to suffer injection loss and have narrowed light mixing cones within the medium.

Attempts to use hollow light transport systems date back to the 17th century using reflective materials such as brass and silver, to transport candle light [5]. Much later, emissive hollow distribution systems were explored, relying on silver films with holes to allow light extraction along the length, and incorporating arc lamps as the source. Present day commercialized systems employ circular hollow tubes of prismatic polymer films, which transport light by TIR and typically use metal halide or high pressure sodium sources [6]. However, the application of hollow, high-bounce light distribution systems has been limited by light-loss associated with multiple reflections, combined with restrictive light-injection requirements for TIR transport.

We have found that systems comprised of low-loss transflectors, high hemispheric-reflection (R_hemi) back reflectors, and a semi-specular characteristic over each light-transport reflection cycle, act to favorably control the ratio of light transport and extraction characteristics, with relatively high efficiency. The semi-specular reflection characteristic diffusely propagates light primarily in a chosen direction, attenuating the multiple images (hall of mirrors) associated with specular reflectors. The large number of bounces associated with this mixing approach magnify material losses, making highly reflective materials paramount to achieve required luminaire or backlight efficiency (Fig. 2(b)). In the next two sections we introduce novel optical materials that, as a front and back reflector combination, provide a high-efficiency uniformity solution for low numbers of high output LEDs.

Semi-Specular interference reflectors

Multilayer interference reflectors can exhibit very high reflectivity independent of incidence angle, a property called omni-directional reflectivity. Weber et al. [7] showed that a large class of reflectivity performances can be achieved in birefringent polymer multilayer interference reflectors by carefully controlling the birefringence of their individual layers. These specular reflectors have proven to have very high hemispheric reflectivity (R_hemi>99.5%) and have been used for low-loss light transport in commercial applications.

High-reflectivity diffuse reflectors, with near-Lambertian reflection properties, are widely commercialized (MCPET) [8]. Light rays incident from any angle on these diffuse reflectors are scattered (reflected) into the entire reflection hemisphere. As a result, they tend to localize light upon multiple reflections, providing little light transport. However, they can provide a very efficient mixing mechanism since each bounce strongly randomizes direction and polarization, and non-specular light redirection can aid in extraction from a light mixing cavity.

We have found that a low loss reflector with both specular and diffuse properties (semi-specular) is a key component in a hollow mixing cavity that can both transport and mix light from localized high-output light sources. We propose a simple approach to combine the light transport benefits of high-performance specular reflectivity and the angle mixing properties of diffuse reflection, by using a uniform weakly diffusing layer at the surface of a specular multilayer birefringent polymer reflector.

The optical challenges of fabricating a high-performance semi-specular reflector are illustrated below. A visible band multilayer polymer reflector with R_hemi >99% is illustrated in Fig. 3(a) . A surface diffusing layer can be optically connected to the polymer reflector by lamination, co-extrusion, coating, micro-replication or other techniques. It can be a bulk diffuser where diffusion is determined by scattering parameters such as mean free path and asymmetry parameter of the scatterers, or a surface diffuser where diffusion is determined by the surface-normal distribution, or some combination. Either way, this diffusing layer can induce some rays from the incident hemisphere to propagate into the underlying reflector past the critical angle for resonant reflection, effectively acting to immerse the underlying reflector into a medium of refractive index n>1. This has several consequences, each of which contributes to reduction in the hemispheric reflectivity (Fig. 3(b)).

 

Fig. 3 Reflectance for s- and p-polarized light averaged across 450-650 nm as a function of incident polar angle for (a) a specular multilayer interference reflector on a strongly absorbing substrate; (b) the same reflector as in (a) with a surface diffuser of index 1.5; (c) the same reflector as in (b) with a reflection band extended into the near infrared; (d) the same reflector as in (c) with a surface diffuser of index 1.35. Calculated and experimental total hemispheric reflectivity values are given for each example.

Download Full Size | PDF

Each of these mechanisms can be mitigated in turn through appropriate design choices:

Figure 3 illustrates the use of these techniques to achieve a high hemispheric reflectivity (R_hemi > 98.2%) with a semi-specular reflection characteristic. The specular reflector in Fig. 3(a) is available from 3M Company as Vikuiti^TM ESR.

Angle and polarization selective transflectors

The front emitting surface plays a critical role in efficiently generating a uniform light output from a volume-efficient light mixing cavity. Its hemispheric reflectivity must be designed to ensure sufficient light transport in the cavity and efficient light extraction while minimizing absorptive losses. Either or both of the bottom and top reflectors must have a semi-specular character appropriate to the requirement of uniform spatial mixing within the cavity. In addition, for edge-lit mixing cavities, angle-dependent front surface reflectivity can provide an engineering handle for high reflectivity at solid angles where most of the light transport occurs, and low to medium reflectivity for solid angles from which light is preferably extracted. Simultaneously, the front emitting surface should be designed to extract light into solid angles and polarization states appropriate for a given application and recycle all other light.

Recent advances in birefringent polymer resin and manufacturing technology have enabled a new family of angle and polarization selective multilayer reflectors that can provide a tuned emission surface for a variety of efficient, thin mixing cavities. Four distinct angular behaviors are presented in Fig. 4 , in combination with polarized and unpolarized examples. These transflectors consist of 275 to 550 alternating quarter-wave layers of high and low index materials. Generally, the low index materials are amorphous and high index materials are birefringent. The transflector in Fig. 4(a) is available from 3M Company; those in Fig. 4(b)-(d) were produced as part of internal development.

 

Fig. 4 Angle and polarization selective transflectors. The principle-axis refractive index values of the birefringent materials enabling these transflectors are given on the right. Those corresponding to the conoscopic plots of transmittance on the left are highlighted in black.

Download Full Size | PDF

Polarized transmission in Fig. 4 is defined as being substantially s-polarized in one azimuthal direction and substantially p-polarized in the orthogonal direction so that the electric field always vibrates in a chosen plane. Unpolarized output is defined as not being substantially polarized in any azimuthal direction.

A unique attribute of these transflective (partially reflective, partially transmitting, low absorption) optical films is that unlike TIR-enabled prismatic films, they can combine high selectivity in polar angle with relatively low selectivity in azimuthal angle. Achieving a weak azimuthal sensitivity is a key to providing a broad acceptance angle of injected light and a large spatial zone of influence for each source (Fig. 5 ). As a result, injected light need be collimated only in the thickness direction. This enables planar high-aspect-ratio mixing cavities with a low number of widely spaced light sources, and substantially improves the uniformity robustness to point failure of one or multiple sources, as well as system sensitivity to LED-to-LED variation of color and brightness [9]. This can reduce the need for costly color binning, or enable much wider color or brightness ranges to be used. For this class of transflectors, the hemispheric reflectivity can be tuned independently of the angular selectivity by adjusting the total number of layers in the interference stack.

 

Fig. 5 Computational comparison of transported light rays in (a) solid light-guide, (b) hollow light-guide using TIR prismatic structures and (c) hollow light-guide using new angle-selective transflectors. In each case the light source is an un-coupled Lambertian LED. Only the rays that propagate to the distal end are shown. Perfectly reflecting specular walls are assumed for the simulation. One of the key advantages of angle-selective transflectors is their low dependence on azimuthal angle, which allows for near-Lambertian injection in the plane of the cavity and enhanced mixing.

Download Full Size | PDF

In some cases, the reflected angular distribution required for optimum light transport, mixing and extraction may not yield the angular distribution of emitted light desired for a particular target application. In such instances, a prism-like structure can be added to the outer surface of the front transflector to adjust the amount of output collimation without significantly affecting the angle selectivity inside the cavity. However, prism structures tend to depolarize light and reduce overall polarized system efficiency.

Examples

The structural simplicity of the edge-lit hollow mixing cavity (Fig. 2(a)) allows for a high degree of design flexibility. In particular, the LED distribution (i.e. edge spacing or layout) is very much unconstrained and can be chosen for optimum heat management and LED performance. Light from LEDs, typically exhibiting a forward emitting Lambertian output pattern, can be injected in the hollow cavity using a simple collimating optic such as a reflective wedge. Because our approach includes a transflector and back reflector with little variation in their optical properties over broad ranges of azimuthal angle, no in-plane angular control of injected light is required (Fig. 5(c)). This is one of the key advantages of the edge-lit hollow design in delivering superior uniformity and robustness.

A first example consists of a 17cm x 6.6cm x 1.8cm hollow mixing cavity lined with a 99.13% R_hemi reflector (see Fig. 3(a)) and lit with a single red LED (Fig. 6(a) photographic image), resulting in an area conversion ratio of approximately 10,000. A simple specular wedge reflector was used to collimate the Lambertian LED light in this case. Figure 6(b) shows an image of the output distribution of the cavity when the specular angle-selective transflector of Fig. 4(b) is used as the front emission surface. Because all surfaces in the cavity are specular, multiple images of the LED source are generated inducing a “hall-of-mirrors” effect. This effect is eliminated by adding an inward facing diffuse bead coating to the angle selective transflector to create semi-specular reflections. The result (Fig. 6(c)) is highly uniform luminance across the output surface that retains essentially no information of the original source location within the cavity. Uniformity was measured at 84% using a 13-point uniformity metric. Efficiency was measured at 80% using a 1-m integrating sphere. Introducing a Lambertian reflector (Fig. 6(d)) produced significant non-uniformity resulting from rapid light extraction near the source.

 

Fig. 6 Photographic images of the output light distribution for (a) an open hollow cavity lined with a 99.13% R_Hemi specular reflector and containing a single red LED; (b) the same hollow cavity with a specular angle-selective transflector as the front output surface demonstrating the “hall-of-mirrors” effect; (c) the same hollow cavity with a semi-specular transflector as the front output surface demonstrating high spatial uniformity; (d) the same hollow cavity as in (c) with a 98.5% R_Hemi Lambertian reflector (MCPET) on the back surface. (e) Side-view and (f) front-view of a 32” LCD backlight prototype consisting of a 16mm-deep single-edge illuminated hollow cavity with 72 high power white LEDs. The measured output angular distribution of the backlight combined with a 1080p LCD panel from a Sharp LC-32D62U TV is shown as an inset in (f).

Download Full Size | PDF

A second example consists of a large-scale 32” diagonal LCD backlight prototype that illustrates the advantages of the edge-lit hollow light mixing concept in facilitating the adoption of high brightness LEDs by the display industry. Figure 6(e)-(f) show side-view and front-view images of the 16mm-deep hollow cavity system. Seventy-two cool white Philips Lumileds’ Luxeon Rebel LEDs were arranged on metal core printed circuit boards (MCPCB) in a linear array along a single edge with a pitch of 9.8 mm. The specular back reflector had a R_hemi value of 99.13%. Shaping the backplane was used as an optional engineering handle for brightness profile adjustment. The front emission surface consisted of an angle and polarization selective transflector transmitting linearly polarized light with a viewing-angle appropriate for an LCD TV application. System uniformity measurements using a Prometric camera demonstrates a 13-pt. uniformity of 85% and an average polarized backlight luminance of 6000 nits for a power consumption of 135W. The viewing angle of the backlight with a 1080p LCD panel from a Sharp LC-32D62U TV was measured using an Eldim conoscope and the result is shown as an inset in Fig. 6(f). The center luminance through the LCD panel in white-state, was 430 nits. For all these measurements, an optional prism film oriented horizontally with 123° apex was used for viewing angle tuning. Efficiency measurements were conducted using an integrating sphere with the diameter of 2m. Backlight efficiency and the polarized efficiency were measured to be 74% and 57%, respectively.

Optical characterization

We defined the hemispheric reflectivity as R_hemi = R/I, where I is the total radiation energy incident from all solid angles of the hemisphere on a surface A, and R is the total energy reflected and scattered from surface A into all solid angles of the same hemisphere. Unless specified otherwise, we averaged R_hemi values over the visible spectrum (450-650 nm). R_hemi can be computed by 1st principles (interference stack morphology and refactive index tensors) using the 4x4 Matrix Method proposed by Berreman (ref. 10). Effects of a diffuse structure overlaying the interference structure are computed by immersion of the interference structure in a half-space medium with a refractive index value of the diffuse medium, and appropriate solid angle weighting of reflected and transmitted light. R_hemi values are measured experimentally using a 2-meter diameter Optronic Laboratories Inc. integration sphere (model IS-7600) and a set of known standard reflectors.

The degree of semi-specularity of a reflector or transflector can be quantified through a metric called the transport ratio defined for a given angle of incidence as the difference between the energy of the scattered rays in the forward direction and the energy of the scattered rays in the backward direction, normalized by the total reflected energy: TR(θ) = (EF-EB)/(EF + EB). A Lambertian reflector has a TR of 0 while a specular reflector has a TR of 1. Transport Ratio values were measured using an Autronic-Melchers ConoScope. The TR value of the semi-specular transflector used shown in Fig. 6(c)-(f) was equal to 0.65 at 45deg incident angle.

Backlight uniformity for examples 1 and 2 was measured using a Prometric camera (Radiant Imaging PM Series Imaging Colorimeter PM-9913E-1). The backlight systems were mounted vertically at a distance of 1.5m from the camera. A Nikon 50mm lens at f/11 was used along with the internal ND1 filter to collect the images.

Efficiency measurements were conducted using a 1-meter diameter Optronic Laboratories Inc. integration sphere (model IS-7600) for example 1 and a 2-meter diameter Optronic Laboratories Inc. integration sphere (model IS-3900) for example 2. For both examples, efficiency was calculated as the ratio of the total LED output and the total system output.

To measure polarized efficiency (important for LCD displays), an absorptive polarizer (Sanritz 5618) was placed above the front plate during the measurement, and the pass axis aligned with the multilayer film pass axis. The polarized output over the total LED output gave the polarized efficiency.

Summary

In summary, we have designed, manufactured and characterized new multi-functional optical materials based on birefringent interference multilayer films that advance hollow mixing cavities to new performance levels. Emissive hollow light guides as a class of light management are now practical for a wide range of mainstream optical systems and applications. Their ability to efficiently distribute and mix light from bright point sources addresses several central issues key to the broad adoption of LEDs in luminaires, display backlights, and other solid state illumination systems. The properties of these hollow cavities are useful for mixing light from white or separate R, G, and B LED sources, or to average out binning variations inherent in the LED manufacturing process. Also, these systems have proven to be highly robust to localized failure of one or a large number of LEDs. Systems with very few, widely spaced sources are possible. Even further increases in system efficiency will be possible through future reductions of both bulk polymer absorption and high angle internal scattering. As the performance of LEDs further increases, these new materials could provide a simple and practical light management solution to address important industry challenges and take advantage of the new design freedoms that LEDs present.