1. Introduction

Lighting systems providing high power and high spatial resolution are gaining increasing importance for applications, such as automotive head lamps, projection, and television systems [1–3]. Compared to conventional systems, high-resolution head lamps increase the light quality and safety due to adjustable, glare-free light [4], and adaptive front lighting systems [5]. For automotive lighting applications, high-resolution light emitting diode (LED) arrays and liquid crystal based solutions are state-of-the-art [6, 7]. The spatial resolution of LED arrays, however, is limited by the minimum size of the chip. In 2016, a minimum size of 0.65 mm by 0.35 mm has been reported [8]. An alternative approach is to use a focused light beam scanning rapidly over a phosphor array. In this case, a pixelated phosphor array is needed. The pixels are structured by laser ablation [9] or photolithography and subsequent etching [10]. However, the resolution of such pixelated phosphors is low due to light scattering into neighboring pixels (“optical crosstalk”). The optical separation of the pixels is thus a key parameter for high resolution. For high-power excitation, a thermal management of the phosphor is of great importance. The temperature of the phosphor increases significantly upon intense blue laser irradiation since approximately 20 % of the incident optical power is converted to heat due to the Stokes shift from blue to yellow. This results in a so-called thermal quenching of the phosphor, as observed exemplarily for YAG:Ce [11, 12]. Luo et al. [13] describe the types of heat generation in LEDs and the possibilities of attaching a phosphor onto a LED chip. Literature on phosphor arrays is rare. Katsumata et al. [14] used a 3 × 3 array with millimeter-sized phosphors as luminescence sensor.

In this work, a novel type of porous phosphor microstructures is studied. The structures are fabricated by filling and fixing loose phosphor particles in pre-patterned molds, which have been fabricated on silicon substrates using micromachining techniques [15, 16]. With this technique, phosphor pixels with a size length of less than 50 µm are feasible; the pixels are optically decoupled by the remaining silicon walls, as shown in Fig. 1. Silicon has a relatively high thermal conductivity of 148 W/(m·K) [17], which is one order of magnitude larger than that of the commercial LED phosphor YAG:Ce [18, 19]. The silicon around each phosphor pixel decreases the thermal load in the phosphor caused by the above-described heat generation by the Stokes shift. Furthermore, the silicon structure can be removed selectively, which enables the use in reflectance and transmission geometry as well as an active cooling of the free-standing porous phosphor structure by air or water, as shown in structure E of Fig. 1.

 

Fig. 1 Schematic cross sections of the investigated phosphor structures: The structures A, C, and E represent a single square pattern with a side length of 2 mm. The structures B, D, and F represent pixelated structures. The side length of the square pixels ranges from 250 µm down to 25 µm.

Download Full Size | PDF

A similar concept is tested by Lee and Lee [20]. The LED chip is inserted in pre-patterned silicon molds and a small volume of epoxy is dropped on top of each LED chip. The yellow phosphor powder is then screen printed onto the epoxy and the epoxy is cured with an ultraviolet lamp. Apart from the fact that epoxy or silicone matrices suffer from high temperatures, the phosphor-epoxy composites are not porous and thus do not allow active air or water cooling.

The optical contrast between the individual phosphor pixels, the temperature distribution within the silicon-phosphor-matrix as well as the spectral light intensity distribution are investigated for different pixel geometries upon intense blue laser irradiation. The effect of air cooling on the temperature distribution is analyzed for a free-standing porous phosphor after removal of the surrounding silicon.

2. Experimental details

2.1. Device preparation

All devices are fabricated on 725-µm thick 8-inch silicon wafers. The process starts with the creation of a mold pattern within a silicon substrate by deep reactive ion etching (DRIE) using a common photoresist mask. After removal of the resist, the mold pattern is filled with dry phosphor particles (GAL555-02-13 [21]) using the doctor blade method. Then, atomic layer deposition (ALD) is used to coat the loose particles with a 75-nm thick Al₂O₃ layer for stabilization. At this point, the structures A and B in Fig. 1 are finished. To realize the structures C and D, the silicon at the bottom of the porous phosphor structure is removed by DRIE from the backside of the substrate using another photoresist mask. To allow an air or water flow-through the Al₂O₃ layer at the mold bottom is also removed.

To investigate different aspects of the potential of these structures, different patterns are fabricated. Structures A and C comprise of a single square pixel with a side length of 2 mm (labelled A2000 and C2000). Structures B and D consist of a phosphor matrix of several pixels with a side length ranging from 250 µm (labelled B250 and D250) down to 25 µm (labelled B25); the remaining silicon walls have a thickness of 20 µm, i.e., the pixel pitch amounts to 270 µm and 45 µm, respectively. A photograph of a reticle containing these different structures is shown in Fig. 2. It contains 42 structures in total, the structures investigated are indicated by a red square. The dimensions of the pixels are summarized in Table 1. Scanning electron microscopy (SEM) images of the four different pixel sizes investigated are shown in Fig. 3.

 

Fig. 2 Photographic image of the structures. The investigated structures are marked with a red square.

Download Full Size | PDF

Table 1. Structures investigated with corresponding pixel size, pixel pitch, and pixel depth.

View Table

 

Fig. 3 Secondary electron images of the phosphor-filled silicon structures with pixel sizes of (a) 25 µm, (b) 125 µm, (c) 250 µm, and (d) 2000 µm.

Download Full Size | PDF

It is typical for DRIE that the etching depth of a mold decreases with its size. For structures A, C and E in Fig. 1 (2 mm × 2 mm pixel size) a depth of 490 µm is measured. For structures comprising 250 µm × 250 µm pixels, the etching depth is about 455 µm. For structures with 25 µm × 25 µm, it decreases to about 210 µm. Note, that the standard deviation of the etching depth amounts typically to approximately 2 % over the entire wafer.

Though the filling of the molds is carried out manually, the fabrication procedure is reproducible. For example, the characterization of micro-magnets agglomerated by ALD from NdFeB powder (mass median diameter (D50) of 4.5 µm) revealed for the remanence repeatedly a standard deviation of about 8 % over an 8-inch wafer (40 samples per wafer, micro-magnets of same size as structure A in Fig. 1). For the phosphor structures such investigations are not yet available. However, it is believed that the results are comparable.

To obtain structures E and F of Fig. 1 the substrate is diced into chips. Then, on chip level, the entire silicon surrounding is removed selectively in XeF₂ gas phase. The SEM image of the phosphor array with a pixel size of 125 µm after selective removal of the surrounding silicon is shown in Fig. 4. The filling height of the remaining, phosphor-filled Al₂O₃ containers is uniform, there are no large voids. However, the filling level is slightly lower than the wafer’s surface. For flow-through experiments, these phosphor structures are glued to glass capillaries (structure E in Fig. 1). For additional details, the reader is referred to the literature [15, 16].

 

Fig. 4 Secondary electron image of a phosphor array with 125 µm pixel size after selective removal of the surrounding silicon. Note that the Al₂O₃ walls of the phosphor-filled containers are 75 nm in thickness.

Download Full Size | PDF

Since the phosphor particles are relatively large (D50 of 14 µm) the smallest pixels (25 µm) are not properly filled (Fig. 3 (a)). The fill factor of the phosphor structures is in the range of 35 % to 40 %. For silicon structures filled with NdFeB micro-magnets (D50 of 4.5 µm), a fill factor of even 45 % has been found. Note that the above-described micromachining technology is still at its beginning. Using a sophisticated filling procedure and dedicated equipment significant improvements are feasible.

2.2. Methods

Secondary electron images are recorded using scanning electron microscopes (JEOL JSM-6510 and Zeiss CrossBeam 1540EsB). Absolute photoluminescence quantum efficiency measurements are performed with a commercial quantum yield measurement system (Hamamatsu C9920-02G) coupled to a 3.3-inch integrating sphere. A xenon lamp (150 W) is used for excitation, a photonic multichannel analyzer (PMA 12) for detection.

Angularly resolved emission spectra are recorded with a robot goniophotometer (opsira GmbH robogonio mrg-6), equipped with a spectrometer (opsira spec’3-v-f). The excitation is carried out with a mid-power (30 mW) 450-nm laser diode (Laser Components Flexpoint Laser Diode Module FP-D-450-40-C-F) with a spot size of approximately 300 µm (Full Width at Half Maximum, FWHM). The incident angle is 60° to the surface normal. The CIE (Commission Internationale de l’Èclairage) chromaticity coordinates are calculated on the basis of the CIE 1931 system with 2° observer.

To determine the optical contrast between adjacent pixels, the luminance of the samples is measured with a luminance camera (opsira GmbH luca4c-2-fw) in combination with a neutral density filter (Schott NG9). Here, the sample is exited with an attenuated focused He-Cd laser (Kimon Koha IK5751I-G) at a wavelength of 442 nm. The spot size is approximately 16 µm (FWHM) in diameter. Again, the incident angle is 60° to the surface normal. The optical power at the sample’s surface amounts to 2 µW. The detection is in direction of the surface normal.

Measurements of the temperature distribution at high-power operation are performed with an infrared (IR) camera (InfraTec GmbH ImageIR 8380S), which uses an indium antimonide (InSb) focal plane array (FPA) snapshot detector with a geometric resolution of 640 px by 512 px. The spectral range for detection is between 2.0 µm and 5.7 µm. In this experiment, a high-power 450-nm laser diode (Thorlabs GmbH L450P1600MM) with a spot size of approximately 25 µm (FWHM) is used for excitation. The maximum optical power at the sample’s surface amounts to 880 mW. The incident angle is 60° to the surface normal, the detection is in direction of the surface normal. Prior to the measurements, the sample is brought to its thermal equilibrium by continuous laser irradiation for approximately 20 min.

The laser spot sizes of the He-Cd laser and the two laser diodes are determined with the knife-edge method [22, 23]. The optical output power is measured with a compact power and energy meter console (Thorlabs GmbH PM100D) in combination with a slim photodiode sensor (Thorlabs GmbH PM130D) for powers lower than 500 mW. For powers higher than 500 mW, an integrating sphere photodiode power sensor (Thorlabs GmbH S142C) is used.

To investigate the possibilities of active air cooling, a phosphor membrane is prepared for which all silicon has been removed. The phosphor membrane is glued to the tip of a glass capillary (structure E) with an outer diameter of 1.5 mm and a length of 127 mm. For active cooling, air is blown through the capillary and the membrane by a diaphragm pump. The thermography setup is the same as described above. The optical excitation power amounts to 600 mW.

3. Results and discussion

3.1. Optical properties

To check for the efficiency of the used phosphor powder, absolute photoluminescence quantum efficiency measurements are performed (not shown). The experiments are carried out for Al₂O₃-coated phosphor agglomerates (structure F, Fig. 4). The silicon has been selectively etched away to obtain reliable data without any influence of the mold material. The quantum efficiency of the phosphor amounts to approximately 95 % in the investigated blue spectral range. This value is similar to that of comparable commercial phosphors [24].

The scattering and emission properties of the phosphor structure are analyzed by photometric far-field studies on structure A2000. For this, emission spectra are recorded in the C0, C90, C180, and C270 planes under an incident angle of 60°. Fig. 5 shows exemplarily the spectrum detected in direction of the surface normal, i.e., for γ = 0°. The spectrum comprises the backscattered light from the 450-nm laser diode and the broad emission of the phosphor in the spectral range from 465 nm to 800 nm. The calculated CIE chromaticity coordinates are x = 0.38 and y = 0.46, which is in the yellow-white color range, as shown in the inset of Fig. 5.

 

Fig. 5 Backscatter and emission spectrum of structure A2000 excited under an incident angle of 60° to the surface normal with a mid-power 450-nm laser diode. The detection is in direction of the surface normal. The backscattered laser light is centered at the laser emission wavelength (blue shaded area), while the phosphor emission ranges from 465 nm to 800 nm (orange shaded area). The inset shows the CIE chromaticity diagram with the point of equal energy, E, and the coordinates for the spectrum of structure A2000.

Download Full Size | PDF

Subsequently, the measured spectra are integrated for the backscattered laser emission in the blue spectral range and the phosphor emission in the yellow-orange spectral range. The results are presented in Fig. 6: The investigated structure A2000 shows an almost perfect Lambertian behavior in the C90 and C270 planes. However, in the C0 plane at an angle of γ = 60°, an intense peak is observed due to specular reflection of the incident laser beam. In the C180 plane, the angular range from γ = 45° to 90° is blocked by the experimental setup (laser diode mount).

 

Fig. 6 Normalized intensity plot of the backscattered blue laser light and the yellow-orange phosphor emission of structure A2000 in the (a) C0-C180 and (b) C90-C270 planes. To obtain the corresponding intensity values, the measured spectra, as shown in Fig. 5, are integrated. The intense peak, observed in the C0 plane at an angle of γ = 60°, is caused by specular reflection of the laser beam. In the C180 plane, the angular range from γ = 45° to 90° is blocked by the experimental setup (laser diode mount).

Download Full Size | PDF

As shown in Fig. 7 in the CIE 1976 (u′, v′) uniform chromaticity diagram, the averaged color coordinates are ${\overline{u}}^{\prime }=0.197$ and ${\overline{v}}^{\prime }=0.539$, calculated for all angles between 0° and 80° (angular resolution of 1°) in the C90-C270 plane. To further characterize the color over angle (CoA) variation, a ten-step u′v′ circle having the averaged color coordinates as center and radius of r = 10 · 0.0011 [25]:

 

Fig. 7 CIE 1976 (u′, v′) uniform chromaticity diagram with coordinates for all angles between 0° and 80° (angular resolution of 1°) in the C90-C270 plane, indicated by crosses. The ten-step u′v′ circle has the averaged color coordinates ${\overline{u}}^{\prime }$ and ${\overline{v}}^{\prime }$ as center (red dot).

Download Full Size | PDF

The luminance imaging measurement, as shown in Fig. 8, are performed on structure B25 under low-power laser excitation at a wavelength of 442 nm and an incident angle of 60°. An analogue measurement on structure A2000 is shown for comparison. The pixelated structure shows a significant drop of the luminance between the laser-excited pixel and the adjacent pixels by two orders of magnitude (−20 dB), whereas the reference structure shows a reduction by a factor of 6.5 (−8.1 dB) at a similar length scale.

 

Fig. 8 Luminance of structure B25 under low-power excitation at 442 nm. The incident angle of the laser excitation is 60° to the surface normal, the detection is in direction to the surface normal. (b) Luminance profile of structure B25 along the line indicated in (a). The vertical dashed lines indicate the position of the silicon walls between the pixels. The luminance profile of structure A2000 is added for comparison.

Download Full Size | PDF

3.2. Thermal properties

The maximum temperature of the four investigated structures A-D upon 450-nm excitation is depicted in Fig. 9. As expected, the temperature increases with increasing excitation power. Prolonged laser excitation of single pixels (structures A2000 and C2000), i.e., without silicon walls, results in phosphor temperatures above 250 °C for excitation powers of 200 mW, while pixelated samples (structures B125 and D125) withstand much higher excitation powers. Here, only for an excitation power of 800 mW and higher the phosphor temperature rises to above 250 °C. It is evident that the silicon walls enable an efficient passive cooling of the structures. In the case of the pixelated phosphor the cooling effect is still significant even without backside silicon (structure D125) due to the silicon wall structure remaining between the pixels.

 

Fig. 9 Maximum temperature versus optical excitation power for the four different structures investigated. The dashed lines represent exponential fits to the experimental data.

Download Full Size | PDF

Results of the experiments with active air cooling are shown in Fig. 10. It is obvious that active air cooling results in a significant temperature decrease: The maximum temperature of the uncooled phosphor amounts to 300 °C, while it is 260 °C for the cooled sample.

 

Fig. 10 Temperature profiles of an air-cooled (left, blue curve) and an uncooled (right, orange curve) phosphor sample of structure E. The insets show the as-recorded thermal images for a phosphor area of 700 µm by 700 µm, each. The laser excitation power is 600 mW.

Download Full Size | PDF

4. Conclusion

Porous square phosphor pixels are suitable for high-resolution laser-based lighting systems. For pixels with a size length of 25 µm and a pixel pitch of 45 µm, luminance imaging yields that the optical contrast between the illuminated and the non-illuminated adjacent pixel amounts to two orders of magnitude. In addition, the silicon walls of the pixelated structures help to decrease the temperature of the structures. To investigate the improvement of the cooling further, preliminary experiments on an actively cooled phosphor membrane are performed. Even a slow air flow results in a phosphor temperature decrease at the center of the laser excitation by 40 °C. The most promising candidate for high spatial resolution and low phosphor temperatures is probably a structure which is passively cooled by the silicon walls and in addition, actively-cooled by air (or even water if possible). For active cooling, however, the backside silicon has to be removed.