The fundamental challenge in the display of reflective color is to take diffuse incident ambient light, absorb a selected part of the visible spectrum, and efficiently return the rest to the viewer. To display a reasonable subset of the perceivable colors requires control of three independent color channels. Traditionally the spectrum is divided into red, green and blue channels (RGB), roughly corresponding to wavelength bands 600-700nm, 500-600nm and 400-500nm respectively. In the case of emissive color displays, additive color mixing is used with clusters of three subpixels, one for each channel. Light within each channel is emitted and mixes in the eye to give the appropriate color sensation. In a typical liquid crystal display, for example, absorbing color filters in each sub-pixel let only the required color through from a white backlight and then the intensity is modulated by the liquid crystal and the polarizers. As the pixels are side-by-side, only 1/3 of the area is dedicated to each channel. If we display red, for example, 2/3 of the display area will be black. As the light is being emitted, a simple 3x increase in backlight brightness would compensate for this; in practice more than this is necessary to cover other losses in the system.

For reflective operation however, this loss is far too high. On average 2/3 of the available ambient light will be absorbed by the color filters giving a maximum reflectivity of 33%, around half the brightness of white paper. Instead one requires an architecture that can separately modulate the intensity of red, green and blue light at every point on the display surface. Printing solves this problem by using subtractive color and layering the inks – for example printing yellow (blue absorbing) over magenta (green absorbing) to achieve red.

An electronic display can also stack three electro-optical layers, one for each color channel. Two basic types of color channel modulator can be stacked – transparent to reflecting (additive) over a black background, or transparent to absorbing (subtractive) over a white background. Many of the stacked color technologies reported to date have used clear to reflecting layers, implemented either by cholesteric liquid crystal [1] or holographic polymer dispersed liquid crystal [2] modulators. To date, the demonstrated maximum reflectivity of both technologies has been limited, in the first case to <50% of each color channel, as only one polarization is reflected, and in the second by the small refractive index difference, which leads to a reflectance peak much narrower than the ~100nm required for a color channel. Therefore we have chosen to focus on systems that switch between absorbing and clear states [3]. By varying the level of absorbance it is possible to tune the reflectivity, and with the systems that we have used the electro-optical effect does not impose an inherent limit on the reflectivity. However the total reflectivity will be limited by the losses in the layers that make up the display.

In any reflective device, each electro-optical layer will typically have some fixed losses associated with it. Although addressing electrodes can be made quite transparent, there will typically be a few percent absorption. The electro-optical layer will also have limited dynamic range, and the substrate and other materials are unlikely to be 100% transparent. Light must pass the 3 layers twice, on the way in and out, so the total loss in each layer is compounded 6 times. If each of the three layers has just 10% loss, nearly 50% of the incident light will be lost (0.9⁶ = 0.53). A further problem is that of non-ideal absorption spectra for the colorants. Each layer should absorb light strongly within just one of the color channels and should have no absorption outside of that band – corresponding to an ideal cyan, magenta and yellow. In practice however most colorants exhibit absorption overlaps and/or gaps with the other channels, Fig. 1(a) . If the absorption spectra do not overlap to fill the visible range, black will be hard to achieve – some wavelengths will be poorly absorbed. Where spectra do overlap, colors will be darker and less vibrant than desired, because more of the spectrum is being absorbed than necessary [4]. Historically, a lot of effort has gone into engineering stable colorants to mitigate these problems, particularly for printing and art [5]. In a 3 layer display system we have the same issues plus the added complexity of electrically modulating the absorption in those layers, with a high dynamic range. In this paper we describe an architecture that minimizes these losses and which allows a lot more freedom when choosing the color properties of each electro-optical layer.

 

Fig. 1 (a) Absorbance spectra for the yellow, magenta and cyan dye doped liquid crystal mixtures used in this paper. Details are in the main text. (b) Reflectivity spectra for the blue and green ILRs.

Download Full Size | PDF

We adopt the principle of returning incident light as soon as possible. We place spectrally selective interlayer reflectors (ILRs) between each of the (clear to absorbing) electro-optical layers. These ILRs reflect all the light in a particular color channel as soon as it has been modulated by the relevant electro-optical layer, eliminating, for that part of the spectrum, any losses present in lower layers. For example, in Fig. 2 the top electro-optical layer is yellow, to control the intensity of blue light. We place a blue reflector beneath that layer, so that blue light does not have to pass through lower electrodes and electro-optical layers. As a result, since no blue light reaches them, it no longer matters whether those layers absorb (or reflect) blue. The rest of the spectrum continues on to be processed by the lower layers. This simple change in architecture significantly relaxes materials and design constraints.

 

Fig. 2 The architecture of the enhanced subtractive display, showing a stack of yellow, magenta and cyan electro-optical layers, interleaved with spectrally selective diffuse reflectors.

Download Full Size | PDF

In Fig. 2 the middle electro-optical layer modulates green light and would normally therefore be magenta. However, using the ILR opens up a wider range of options. The electro-optical layer must still absorb strongly within the green wavelength band and should not absorb at all at wavelengths assigned to the red modulating electro-optical layer beneath it. However, it can now absorb any amount within the wavelengths assigned to the blue layer above, so that, for example, instead of magenta the middle layer could be red. The ILR placed beneath it is designed to reflect green and must be transparent to red light which has to reach the bottom layer, but it does not matter how the ILR affects blue light, as none will reach it.

The only constraint on the color of the bottom electro-optical layer is that it should absorb red light. It can however absorb any amount of green or blue light, as these have already been filtered out. Rather than cyan, the bottom layer could therefore be blue or even black. The reflector beneath it must reflect red light, but its characteristics in the blue and green channels do not matter, so it could be a simple broadband reflector.

Thus instead of the traditional cyan, magenta and yellow colorants used in subtractive systems, the addition of the ILRs allows much greater design freedom over the choice of the colors while also allowing the performance to much more closely approach that of ideal colorants. It will be apparent that, depending on the order of the layers, there are many different combinations of electro-optical layers and ILRs that will yield equivalent colors. In practice, since many organic colorants have absorbance tails and/or sidebands to the high energy (short wavelength) side of the peak it is usually most efficient to stack the EO layers so that the peak absorbance wavelength increases down the stack.

In order to demonstrate the ILR architecture we made wavelength selective reflectors using multilayer dielectric structures. A standard multilayer reflector is made from alternating layers of different refractive index, using materials such as silica and titania. The textbook design has the optical thickness of each layer set to a quarter of the desired central wavelength, with the refractive index difference determining the width of the spectrum and the total number of layers and the refractive indices determining the peak reflectivity [6]. However, this simple design has significant side bands (Fig. 3 ) which are not ideal as they will reflect some of the light that should reach the lower layers in the display. We have therefore optimized the reflectors to minimize the strength of the side bands. We used a matrix optics method [7] to model the reflectivity from the structure and used an iterative random search method to optimize the design. We constrained the intensity of the long wavelength side bands to be as low as possible, but allowed the short wavelength side bands to increase if necessary. In our display architecture the shorter wavelengths will either be out of the visible spectrum or will have been processed by layers higher in the stack. We further constrained each layer thickness to be an integer multiple of a quarter wavelength as this can be helpful when using optical feedback to control the film thickness in the vacuum coating process. In the final display the reflectors will be laminated into the stack using an optical adhesive with a refractive index that matches the substrates, so in the optimization we set the refractive index of the media on either side of the reflector to this value (1.52). The refractive indices of the layers were set to the values used in the coating (1.59 and 1.83). Applying this optimization process gave the design in Eq. (1)

 

Fig. 3 The modeled reflectivity of a standard reflector design (LH)¹⁰ (black) and an optimized design (red). The optimized design has much lower reflectivity in the sidebands between 500 and 650nm.

Download Full Size | PDF

As well as having the right spectral properties the reflectors should be diffuse rather than specular. The diffuse characteristics determine the brightness and viewing angle of the stack. At one extreme an ideal Lambertian surface will take any available light and distribute it uniformly over all angles. If, however, incoming light is scattered over too wide a range of angles, some will get trapped by total internal reflection at the surfaces, reducing brightness. Towards the other extreme the reflector can be designed to scatter light over a much narrower angle which will make the display brighter by a factor known as the gain, but with a compromised viewing angle. In practice the gain of the reflector is chosen to give a balance between brightness and viewing angle.

To achieve the required combination of spectral and diffuse properties [8] we applied the multilayer coatings to the textured surface of a commercially available transmissive diffuser film, a 10° light shaping diffuser on 125μm film from Luminit LLC. The angular range and extent of diffusion is controlled by the length scale and amplitude of the surface roughness.

Parallax is a potential concern for any display made with multiple stacked electro-optical layers. If the spacing between layers is large compared with the size of the pixel then illumination and/or viewing at high angles will cause the color layers to appear mis-registered, potentially leading to color artifacts at high contrast edges. Adding the ILRs will increase the interlayer spacing, so it is therefore essential that all the substrate thicknesses are kept as small as possible, ideally smaller than the pixel pitch. Later in this paper we demonstrate a display made using thin plastic substrates, giving a total thickness between adjacent color layers of less than 400μm (two display substrates plus an interlayer reflector on its own substrate and adhesive layers). This is less than the 550μm glass layer reported in [9], which reports virtually no parallax to be observed even for a nominal pixel pitch of 240μm. In the long term the extra substrates for the ILRs could be removed and the reflectors included as coatings on the display substrates, resulting in a thickness increase of just a few microns. We have also observed that the ILR architecture appears to mitigate the parallax problem: as each color channel is immediately reflected, rather than continuing to the rear of the display, residual color artifacts are further reduced.

Multilayer reflector coatings generally exhibit reflection spectra that are sensitive to the viewing angle. In our architecture these effects are reduced by two factors. First, by applying the coating to a textured surface we obtain a surface composed of small domains with layers tilted at a range of angles. As well as diffusing the reflected light, this reduces the angle sensitivity of the reflectivity by convolution over the angle range. It also broadens the reflection band, but this can be compensated for by varying the design of the dielectric reflector. Second, once the ILR is laminated into the display stack, the incoming range of light angles is compressed by refraction to about ±40°. In a final display a number of factors combine to determine the viewing angle. Later in this paper (Fig. 7) we show that these combine to give an acceptable viewing angle.

Figure 1(b) shows the reflection spectra for blue and green reflectors made using alternating layers of metal oxides with refractive indices of 1.59 and 1.83. Layers were laminated together using NOA65 clear optical adhesive (Norland Products Inc.). The modest gain for these reflectors is due to the choice of scattering angle for these diffusers which gives a good compromise between brightness and viewing angle. The measured spectra are very similar to the modeled curve in Fig. 3 and have only very weak side bands to the long wavelength side of the peak.

In Fig. 4 we demonstrate how we can utilize these ILRs to produce bright colors from a non-standard set of primary colors. Usually subtractive color systems employ cyan, magenta and yellow colorants but, as described above, if we use ILRs we have a much wider set of choices. Here we show one of these; yellow, red and black rectangles are printed onto transparencies using a commercially available laser printer. The shapes of the rectangles are chosen so that when we overlay them we generate 8 patches that contain all the possible combinations of the three colorants and thus show the extremes of the color gamut that can be achieved (Fig. 4(a)). In Fig. 4(b) the three layers are stacked with optical adhesive between the layers, and a silver coated diffuse reflector on the rear (200nm layer of silver deposited by sputter coating onto the same Luminit diffuser used for the ILR layers). As would be expected, only yellow, red, black and white can be achieved. In Fig. 4(c) we have introduced blue and green ILRs into the stack and now we see a full gamut of colors which is very similar to that produced if we use CMY prints – we clearly have much greater flexibility in our choice of colorants.

 

Fig. 4 Demonstrating the effect of the ILRs. a) Schematic of the print samples with yellow, red and black (YRK) layers. b) YRK printed transparencies and a silver reflector. c) as (b), but with ILRs between the transparencies. d) Schematic of an LC layer with patterned alignment. The ellipses indicate the alignment of the liquid crystal director.

Download Full Size | PDF

We can now apply this architecture to a color display. In principle we could use any clear to absorbing electro-optical effect – for example in-plane electrophoretic migration of pigments [10], electrowetting to control a dyed fluid layer [11,12], or electrochromic effects [13]. Here we have used the White-Taylor dichroic dyed liquid crystal device [14]. Liquid crystals (LCs) are fluids formed from anisotropic molecules, usually rod shaped, which, in the nematic phase, tend to align in a common direction described by a director. The alignment direction can be changed by applying a voltage. If a suitable rod-shaped dichroic dye is added, the dyes tend to align with the director so that they can be rotated with the LC when a field is applied. The dyes only absorb light that is polarized along their long axis so that a low absorption is achieved when the dyes are aligned vertically, with their long axes orthogonal to the display substrate. Arranging that the molecules lie down horizontally with their long axes parallel to the substrates gives an absorbing state which will be strongly colored. In the White-Taylor configuration, the horizontal state is twisted by doping with a chiral agent, so that through the cell there are dye molecules aligned with the long axes at all directions in the plane in order to absorb all polarizations.

We made liquid crystal cells using ITO coated glass treated with alignment layers patterned by lift-off lithography. 10 micron bead spacers were dispersed in a UV curing adhesive (NOA73 from Norland Products Inc.) and used to space two substrates to form a cell. The cells were filled with a negative dielectric anisotropy liquid crystal (zli2806 from Merck KGaA) doped to be chiral (1% by weight of zli811, from Merck), and dyed with yellow (G232 from Hayashibara Biochemical Laboratories, Inc.), magenta (G471 also from Hayashibara) and cyan (AC1, Nematel GmbH & Co. KG) dyes, typically 1-2% by weight.

We built two LC stacks where in each layer we patterned the liquid crystal alignment so that half the cell was in a planar, colored state and half in a vertical and therefore light state, Fig. 4(d). The layout of the colored rectangles was the same as the print sample, Fig. 4(a), with a yellow dye in the top layer but instead with a magenta dye in the middle layer and a cyan dye in the bottom layer. In one stack we included ILRs and in the other we did not. We took photos of each of the 8 colored patches in the two stacks which are shown as insets in Fig. 5 . The left image was taken on the stack with ILRs and the right on the stack without. The stack with the ILRs is brighter, and many of the colors are more vivid as the ILRs mitigate the effects of the non-ideal absorption spectra of the dyes (Fig. 1(a)). In particular the magenta dye has strong absorption in the blue which adversely affects the magenta and blue patches in the stack without ILRs.

 

Fig. 5 Measurements of the reflection spectrum of each colored patch in the LC stacks. The solid lines were measured on the device with ILRs and the dashed lines were measured on the device without ILRs. The color insets in each graph are the photos of the corresponding patches, with the left image taken with the ILRs and the right without.

Download Full Size | PDF

The color properties were measured by illuminating the samples with diffuse white light incident normal to the sample with a cone angle of approximately 18°, and measuring the light reflected normal to the sample with a collection angle of 2° and normalized with respect to a diffuse white reference (WS-1-SL, Labsphere Inc.). For each color patch we measured the reflectivity spectrum. The CIELAB system [15] was then used to combine the reflectivity spectra to give the gamut volume. The aim of the CIELAB system is to give values that take account of the non-linearities of the human vision system and so give values that more closely match the perceived lightness and color. Table 1 summarises the results and shows a 2x increase in the gamut volume, an increase of 15% in the lightness of the white state and an 8% increase in the contrast.

Table 1. Measured Lightness, Contrast and Gamut Volumes for the LC Stacks with and Without ILRs

View Table

Figure 5 shows the measured reflectivity spectra for each patch in the two devices, with (solid lines) and without (dashed lines) the ILRs. The gamut volume, lightness and contrast values are calculated from these curves. The spectrum for the white patch with ILRs shows two clear peaks corresponding to the reflectivity maxima of the ILRs (Fig. 1(b)). The blue component of the reflectivity spectrum peaks at 0.7 and shows the most enhancement as the blue light passes through the smallest number of layers and only encounters the yellow electro-optical layer. Additionally the slight gain of the reflectors further enhances the reflectivity. The green component shows slightly less enhancement as green light has to pass through double the number of layers. The red component is not enhanced by the ILRs as red light has to pass through all the layers, with or without ILRs. The spectra for the other colored patches show the same effect. For example the blue patch shows a sharp peak with the ILRs which is absent in the conventional device as the light is absorbed by the overlapping spectrum of the magenta dye (Fig. 1(a)). Comparison of the spectra and the inset photos in Fig. 5 shows that one needs to achieve a large improvement to the spectrum to have a significant effect on the perceived color.

These results show that the addition of the ILRs has significantly enhanced the color performance of the display, particularly where the blue and green channels are involved. Earlier published work on the optimization of a three-layer dichroic display [16] demonstrated a gamut volume of about 40000, in line with our results without ILRs. For stacked cholesteric displays, Coates [17] claims a reflectivity of 0.30-0.32 (62-63L*), well below the results shown here for the device with ILRs.

The different degrees of enhancement for the three colors need to be taken into account when designing the display so that a balanced color gamut is produced. One can vary the absorbance of each electro-optical layer by adjusting the dye concentration and cell gap, and can optimize the spectral characteristics of the reflectors, by adjusting the peak reflectivity, and width and shape of the peak.

These results show the potential improvement that comes from including the ILRs in the display architecture. In this case, the majority of that improvement comes from mitigating the losses due to the non-ideal absorption spectra, but the ILRs also reduce the broadband loss from the ITO electrodes lower in the stack which improves the overall brightness.

The dyes were chosen to have the best compromise of absorption spectra and electro-optical properties that we could find. In other experiments we were able to use dyes with much more overlap between their absorption spectra which would mean that they are not normally of practical interest. The interlayer reflectors mitigate these losses to give very similar results to Fig. 5. This gives us a much greater design freedom and allows us to consider using materials with sub-optimal absorption spectra, but which might have other beneficial properties such as higher dichroic ratios or lower driving voltages.

Figure 6 shows a segmented display built using these principles. The plastic display was made using 100μm thick ZF16-100 film (Zeon Chemicals) with lithographically patterned gold addressing electrodes coated with PEDOT:PSS (Heraeus GmBH) on one side of the LC layer. As the counter substrate 120μm thick ITO coated PES (Sumitomo Bakelite) was photopatterned with a sparse array of SU8 polymer structures to space the two substrates accurately apart. A typical arrangement of spacer structures consists of cylinders with a diameter of 23 microns and a height of 10 microns, arranged in a hexagonal lattice with a lattice constant of 250 microns. Both surfaces were coated with homeotropic alignment layers (SE4811 polyimide from Nissan Chemical Industries Limited).

 

Fig. 6 Prototype segmented display photographed in office lighting on top of a monochrome electrophoretic ebook display (iRex Digital Reader 1000).

Download Full Size | PDF

The 5cm diagonal device is made from plastic substrates, has a total thickness of less than 1mm and uses the same liquid crystal and dyes as described above. The blue and green ILRs are also the same as those above, but in this case we used a dielectric reflector in place of the broadband reflector beneath the cyan layer. This dielectric reflector was made to the same design as described above, but with a central wavelength of 615nm. The display area is broken up into a number of segments each of which can be individually addressed. The device is pictured on a monochrome electrophoretic display and the photo is taken in normal office lighting. Our display has similar black and white performance but can also show a wide range of colors. Although the color display is segmented, the electrode structures were designed to give similar optical loss to a fully pixellated display at 100pixels per inch; in further work we hope to demonstrate full image capability.

Figure 7 demonstrates the viewing angle of the display. As the angle increases the display darkens slightly but there is only a modest color shift. The darkening is due partly to the inherent properties of the dichroic guest host system. The light state is achieved when the LC and the dyes are oriented vertically so that for viewing angles away from normal the absorbance starts to increase. The gain of the reflectors will also limit the viewing angle, although in this case the illumination was diffuse which reduces that effect. The reflection spectra for the reflectors will also have some angle dependence, although as described above, the textured surface and lamination mitigate that shift. Figure 7 shows that the combination of these effects gives a very acceptable viewing angle, and that the use of thin plastic substrates reduces the effect of parallax so that color artifacts at the segment edges are not significant under normal viewing conditions.

 

Fig. 7 Prototype segmented display photographed in diffuse window light, taken at (a) 0°, (b) 27° and (c) 48° from the display normal.

Download Full Size | PDF

In this paper we have shown that one can significantly improve the color performance of a 3 layer subtractive color system by successively filtering the color channels out of the spectrum at each layer. By integrating wavelength selective diffuse reflectors into a stacked architecture we reduce the critical optical losses while at the same time allowing much greater design freedom over the properties of the electro-optical layers. We have demonstrated these enhancements in a segmented device and have shown a 2x increase in gamut volume, 15% enhancement in perceived lightness and an 8% enhancement in perceived contrast.