Color uniformity enhancement for COB WLEDs using a remote phosphor film with two freeform surfaces

The color uniformity (CU) of chip-on-board (COB) white light emitting diodes (WLEDs) has been improved by using remote phosphor films with two freeform surfaces (TFS-RPFs). The finite-difference time-domain (FDTD), Monte Carlo ray-tracing, and colorthickness feedback (CTFB) methods were used to design the TFS-RPFs: the blue light distribution of COB WLEDs is greatly affected by the angular thickness distribution of TFSRPFs, and a high CU can be achieved iteratively. The directional inconsistency of incident and emergent blue light, scattering effect of TFS-RPFs, and illumination characteristics of the COB source were also investigated. COB WLEDs containing optimized TFS-RPFs achieved high CU with a decrease of 26.2% in maximum CCT deviation; thus, TFS-RPFs can improve the CU of COB WLEDs. © 2016 Optical Society of America OCIS codes: (230.3670) Light-emitting diodes; (250.5230) Photoluminescence. References and links 1. P. Pust, P. J. Schmidt, and W. Schnick, “A revolution in lighting,” Nat. Mater. 14(5), 454–458 (2015). 2. K. J. Chen, H. V. Han, H. C. Chen, C. C. Lin, S. H. Chien, C. C. Huang, T. M. Chen, M. H. Shih, and H. C. Kuo, “White light emitting diodes with enhanced CCT uniformity and luminous flux using ZrO2 nanoparticles,” Nanoscale 6(10), 5378–5383 (2014). 3. C. Sommer, J. R. Krenn, P. Hartmann, P. Pachler, M. Schweighart, S. Tasch, and F. Wenzl, “The effect of the phosphor particle sizes on the angular homogeneity of phosphor-converted high-power white LED light sources,” IEEE J. Sel. Top. Quantum Electron. 15(4), 1181–1188 (2009). 4. H. Zheng, S. Liu, and X. Luo, “Enhancing angular color uniformity of phosphor-converted white light-emitting diodes by phosphor dip-transfer coating,” J. Lightwave Technol. 31(12), 1987–1993 (2013). 5. Z.-T. Li, Y. Tang, Z.-Y. Liu, Y.-E. Tan, and B.-M. Zhu, “Detailed study on pulse-sprayed conformal phosphor configurations for LEDs,” J. Disp. Technol. 9(6), 433–440 (2013). 6. C.-H. Chiang, H.-Y. Tsai, T.-S. Zhan, H.-Y. Lin, Y.-C. Fang, and S.-Y. Chu, “Effects of phosphor distribution and step-index remote configuration on the performance of white light-emitting diodes,” Opt. Lett. 40(12), 2830–2833 (2015). 7. H. Luo, J. K. Kim, E. F. Schubert, J. Cho, C. Sone, and Y. Park, “Analysis of high-power packages for phosphor-based white-light-emitting diodes,” Appl. Phys. Lett. 86(24), 243505 (2005). 8. Z. Liu, S. Liu, K. Wang, and X. Luo, “Optical analysis of color distribution in white LEDs with various packaging methods,” IEEE Photonic Technol. Lett. 20(24), 2027–2029 (2008). 9. C.-Y. Liu, K.-J. Chen, D.-W. Lin, C.-Y. Lee, C.-C. Lin, S.-H. Chien, M.-H. Shih, G.-C. Chi, C.-Y. Chang, and H.-C. Kuo, “Improvement of emission uniformity by using micro-cone patterned PDMS film,” Opt. Express 22(4), 4516–4522 (2014). 10. H. Y. Lin, K. J. Chen, S. W. Wang, C. C. Lin, K. Y. Wang, J. R. Li, P. T. Lee, M. H. Shih, X. Li, H. M. Chen, and H. C. Kuo, “Improvement of light quality by DBR structure in white LED,” Opt. Express 23(3), A27–A33 (2015). 11. J. H. Oh, S. J. Yang, and Y. R. Do, “Polarized white light from LEDs using remote-phosphor layer sandwiched between reflective polarizer and light-recycling dichroic filter,” Opt. Express 21(S5 Suppl 5), A765–A773 (2013). Vol. 24, No. 21 | 17 Oct 2016 | OPTICS EXPRESS 23685 #272094 http://dx.doi.org/10.1364/OE.24.023685 Journal © 2016 Received 21 Jul 2016; accepted 25 Sep 2016; published 3 Oct 2016 12. J. S. Lee, P. Arunkumar, S. Kim, I. J. Lee, H. Lee, and W. B. Im, “Smart design to resolve spectral overlapping of phosphor-in-glass for high-powered remote-type white light-emitting devices,” Opt. Lett. 39(4), 762–765 (2014). 13. E. Kim, S. Unithrattil, I. S. Sohn, S. J. Kim, W. J. Chung, and W. B. Im, “Facile one-step fabrication of 2layered and 4-quadrant type phosphor-in-glass plates for white LEDs: an insight into angle dependent luminescence,” Opt. Mater. Express 6(3), 804–814 (2016). 14. H. C. Kuo, C. W. Hung, H. C. Chen, K. J. Chen, C. H. Wang, C. W. Sher, C. C. Yeh, C. C. Lin, C. H. Chen, and Y. J. Cheng, “Patterned structure of remote phosphor for phosphor-converted white LEDs,” Opt. Express 19(S4 Suppl 4), A930–A936 (2011). 15. S.-P. Ying and H.-Y. Chien, “Effect of reassembled remote phosphor geometry on the luminous efficiency and spectra of white light-emitting diodes with excellent color rendering property,” IEEE Trans. Electron Dev. 63(3), 1117–1121 (2016). 16. C.-H. Kuo, W.-H. Wu, W.-T. Hsiao, and K.-C. Huang, “The multi-shadow analysis of LED secondary optics,” Phys. Procedia 19, 233–238 (2011). 17. X. Yu, B. Xie, B. Shang, Q. Chen, and X. Luo, “A cylindrical tuber encapsulant geometry for enhancing optical performance of chip-on-board packaging light-emitting diodes,” IEEE Photon. J. 8, 1600709 (2016). 18. H. Xiao, T.-M. Shih, Z.-Q. Guo, Y. Lin, Y.-J. Lu, and Z. Chen, “3-D Microlens phosphor with curvatures manufactured by imprinting for chip-on-board light-emitting diodes,” IEEE Trans. Electron. Dev. 63(3), 1128– 1133 (2016). 19. X. Luo, X. Fu, F. Chen, and H. Zheng, “Phosphor self-heating in phosphor converted light emitting diode packaging,” Int. J. Heat Mass Tranf. 58(1-2), 276–281 (2013). 20. S. Hsu, Y. Chen, Z. Tu, H. Han, S. Lin, T. Chen, H. Kuo, and C. Lin, “Highly stable and efficient hybrid quantum dot light-emitting diodes,” IEEE Photon. J. 7, 1–10 (2015). 21. K. Wang, D. Wu, Z. Qin, F. Chen, X. Luo, and S. Liu, “New reversing design method for LED uniform illumination,” Opt. Express 19(S4 Suppl 4), A830–A840 (2011). 22. H. Zheng, X. Guo, X. Lei, C. Li, J. Chu, and S. Liu, “Fabrication of Lens with Large View Angle through Droplet Evaporation for High-Power Light-Emitting Diodes,” IEEE Photonic Technol. Lett. 28, 1589–1592 (2016). 23. H.-T. Huang, Y.-P. Huang, and C.-C. Tsai, “Planar lighting system using array of blue LEDs to excite yellow remote phosphor film,” J. Disp. Technol. 7(1), 44–51 (2011). 24. H.-T. Huang, C.-C. Tsai, and Y.-P. Huang, “A direct-view backlight with UV excited trichromatic phosphor conversion film,” J. Disp. Technol. 6(4), 128–134 (2010). 25. X. Ding, J. Li, Q. Chen, Y. Tang, Z. Li, and B. Yu, “Improving LED CCT uniformity using micropatterned films optimized by combining ray tracing and FDTD methods,” Opt. Express 23(3), A180–A191 (2015). 26. H. Xiao, Y.-J. Lu, Z.-Q. Guo, Y. Lin, T.-M. Shih, and Z. Chen, “Improvements on optical and chromatic uniformities of light-emitting diodes with microscale-roughness-controlled surfaces,” IEEE Photonics J. 7, 1–8 (2015). 27. S. Yu, Z. Li, G. Liang, Y. Tang, B. Yu, and K. Chen, “Angular color uniformity enhancement of white lightemitting diodes by remote micro-patterned phosphor film,” Photon. Res. 4(4), 140–145 (2016). 28. Z. Cheng, Y. He, F. Cui, B. Du, Z. Zheng, and Y. Xu, “Comparative and sensitive analysis for parabolic trough solar collectors with a detailed Monte Carlo ray-tracing optical model,” Appl. Energy 115, 559–572 (2014). 29. S. T. Flock, M. S. Patterson, B. C. Wilson, and D. R. Wyman, “Monte Carlo modeling of light propagation in highly scattering tissue--I: Model predictions and comparison with diffusion theory,” IEEE Trans. Biomed. Eng. 36(12), 1162–1168 (1989). 30. Z. Liu, S. Liu, K. Wang, and X. Luo, “Measurement and numerical studies of optical properties of YAG:Ce phosphor for white light-emitting diode packaging,” Appl. Opt. 49(2), 247–257 (2010). 31. J. Li, J. Chen, L. Lin, Z. Li, Y. Tang, B. Yu, et al., “A detailed study on phosphor-converted light-emitting diodes with multi-phosphor configuration using the finite-difference time-domain and ray-tracing methods,” IEEE J. Quantum Electron. 51, 1–10 (2015). 32. A. Jones, “Light scattering for particle characterization,” Prog. Ener. Combust. Sci. 25(1), 1–53 (1999). 33. L. Reynolds and N. McCormick, “Approximate two-parameter phase function for light scattering,” J. Opt. Soc. Am. 70(10), 1206–1212 (1980). 34. J.-W. Kim, J.-H. Jang, M.-C. Oh, J.-W. Shin, D.-H. Cho, J.-H. Moon, and J. I. Lee, “FDTD analysis of the light extraction efficiency of OLEDs with a random scattering layer,” Opt. Express 22(1), 498–507 (2014). 35. J. Li, Y. Chen, Z. Li, S. Yu, Y. Tang, X. Ding, and W. Yuan, “ACU Optimization of pcLEDs by combining the pulsed spray and feedback method,” J. Disp. Technol. (to be published).</other>


Introduction
One of the most important techniques to generate white light in white light-emitting diodes (WLEDs) is using blue LED chips to excite phosphor. As the luminous flux (LF) of WLEDs has already far exceeded traditional lighting sources and has satisfied the demands of most applications [1], the color uniformity (CU) [2] has now become a more outstanding concern, especially for agricultural and medical applications. Phosphor layers (PLs) are critical for the CU of WLEDs because of the inelastic collision occurring between photons and phosphor, which can be adjusted to balance spatial distributions of blue and converted light by the amount of down-conversion events and scattering effect [3]. Compared with dispensing PLs (DPLs) [4] and conformal PLs (CPLs) [5], remote PLs (RPLs) [6] are able to achieve a high light output efficiency (LE) because of the reduction in the absorption loss of back-scattered light [7]. However, it is challenging to balance the color distributions for WLEDs with RPL structures owing to insufficient blue light scattering [8]. Combining RPLs with special optical elements, such as micro-cone patterned films [9], distributed Bragg reflectors [10], and lightrecycling dichroic filters [11], can improve the CU by utilizing their scattering or reflecting effect to fully mix different colors of light. However, this will introduce inevitable increases of interface loss and absorption loss. In addition, RPL structures [12][13][14][15] have also been designed to achieve high CU; the geometry, area, and location are adjusted to balance the color distributions by deflecting the propagation direction of the blue light to modify the number of blue photons in a particular direction. This is a low-cost and effective approach to improve CU. However, all of these studies focusing on RPLs are based on discrete components constructed with a single LED chip. The components required to output a high LF are combined by a complex surface-mounting process, which leads to multiple shadow patterns [16] in the illumination area. This has limited the further use of RPL structures in lighting applications. Chip-on-board (COB) WLEDs can solve such problems by using only one source produced by an LED chip array, but have not been widely considered previously.
Presently, optimization in DPLs structures is still an major approach to improve the CU of COB WLEDs [17,18]. However, the thermal reliability of PLs inside COB lead-frames is significantly worse owing to the great amount of heat generated by hundreds of chips flowing to the PLs, which can even cause phosphor quenching and silicone carbonization [19]. The remote structure can also effectively overcome this issue because an isolated layer with low thermal conductive is inserted between the RPLs and chips [20]. However, it is more difficult to design and manufacture RPLs for COB WLEDs with a high CU, compared with discrete components with a single point source. The extended source in COB WLEDs shows more complicated lighting distributions in the near-field [21] where blue photons interact with RPLs, causing great difficulties in balancing color distributions by the structure design of the RPLs. On the other hand, the depth-width ratio of COB lead-frames is much smaller than discrete components; thus, it is difficult to manufacture RPLs with special structures such as curved shapes [22]. Hence, effective design and simple manufacturability is expected for remote phosphor architectures on COB WLEDs with excellent CU. Phosphor films (PFs) have advantages of easy manufacture, low cost, and flexible integration with COB WLEDs, which is a promising approach for WLEDs applications [23,24]. However, the traditional PFs with two regular planar surfaces [25][26][27] have not realized their potential advantages on the CU improvement for COB WLEDs; complicated micro-structures and additional processes are still necessary when optimizing the CU for COB WLEDs.
In this paper, we have proposed a method for modeling remote phosphor films with two freeform surfaces (TFS-RPFs) and investigated the effect of TFS-RPFs on the CU of COB WLEDs. Finally, optimized TFS-RPFs with thicknesses of 0.12 -0.98 mm were fabricated and assembled into COB WLEDs, and the CCT deviation was measured and discussed.

Optical properties of TFS-RPFs
The Monte Carlo ray-tracing method [28] has been widely used when studying light propagation into PFs where the down-conversion and scattering events occur owing to the great number of phosphor particles. The absorption coefficient abs μ , scattering coefficient sca μ , and scattering phase function p are used to characterize the optical properties of TFS-RPFs in the bulk scattering model [29], which has the forms [30,31] where ( ) n D is the phosphor particle size distribution function, c is the phosphor concentration (g/cm 3 ), P λ θ φ is the far-field scattering power distribution. Here, we introduced the finitedifference time-domain (FDTD) method [34] to solve the scattered electromagnetic field of phosphor particles in order to obtain these three parameters, and the detailed simulation setup is shown in the following.

Geometry calculation of TFS-RPFs
We proposed a color-thickness feedback (CTFB) method to calculate the geometry of TFS-RPFs in order to obtain a high CU for COB WLEDs. The principle of the CTFB method is shown in Fig. 1. The aim is to balance the blue light output power, PB , from the LEDs, with the yellow light output power, PY , from the RPFs, at various viewing angles θ and obtain the target color. Increasing the thickness of the RPF unit could lead to a color with less proportion of blue light, and thus become yellowish; contrarily, it could lead to a color with more proportion of blue light, and thus become bluish. Therefore, the geometry of the TFS-RPFs can be determined by its angular thickness ( ) t θ distribution, which is in turn determined by the output color of COB WLEDs. The critical point of the CTFB method is understanding how to use the color deviation (the deviation of PY / PB ) to adjust the thickness of the TFS-RPFs. Firstly, we define n ( ) CDC θ as the color deviation coefficient at the n th feedback iteration process, as per [35] n n 1 where TC is the target color that is a constant; n 1 ( )  On one hand, the amount of blue rays emitted with different α is not the same owing to the illumination characteristic of LED chips, which is commonly treated as the Lambert distribution. On the other hand, the power of blue rays that propagate into the TFS-RPF inside surface with different β is also not the same because of the Fresnel loss. In other words, the change in the thickness of the TFS-RPF unit, which the blue rays propagate into with larger α or β , has less influence on the ( ) PB θ , as the power of these blue rays is relatively lower.

Simulation setup
The commercial software packages FDTD Solutions and Tracepro have been utilized in this study. Figure 3(a) shows the particle size distributions of YAG: Ce phosphor (density 4.5 g/cm 3 , refractive index 1.78). The average particle size and standard deviation are 13.7 μm and 5.0 μm, respectively. The particles are assumed to be spherical in the simulation. Figure  3(b) shows the FDTD model of the YAG:Ce phosphor particle. A 3D FDTD simulation region was used with a boundary condition of a perfectly matched layer (PML). A total-field scattered-field source (TSS) with a center wavelength of 455 nm was set to surround the phosphor particle. The background is silicone with refractive index of 1.54. The transmission detector box (TDB) was used to preserve the electromagnetic field information in order to calculate the sca ( ) P λ , abs ( ) P λ , and far sca ( , , ) P λ θ φ [25]. The grid size was 10 nm and the shutoff level was set to 5 10 − to ensure accurate results. In the Monte Carlo ray-tracing model, two typical wavelengths of 455 nm and 565 nm were used for the blue LED chips and TFS-RPFs, respectively [3]. The TFS-RPL was mixed with the silicone and the YAG:Ce phosphor. The COB was packaged by the silicone and had a circular light-emitting surface which was produced by horizontal blue LED chips (size 0.76 × 0.56 mm) distributed in a 6 × 7 array. The spherical detector was 500 mm from the center of the COB WLEDs.

Results and discussion
The method proposed above was performed to improve the CU of COB WLEDs with TFS-RPFs. The r was set to be 7.75 mm and the initial ( )    Fig. 5(a). With increasing N , the ( ) t θ at central angles become larger, indicating that the increase of ( ) t θ helps to decrease ( ) PB θ . Contrarily, the decrease of ( ) t θ helps to increase ( ) PB θ . One reasonable explanation is that the increasing (decreasing) ( ) t θ contributes to the decrease (increase) in the power of blue light that propagates through the TFS-RPF units located at θ owing to the higher probability of down-conversion events. It is evident that these results are satisfied with the principle of the proposed CTFB method. In addition, it should be noted that the RPFs for 0 N > are all TFS-RPFs as the ( ) t θ is no longer constant. is larger than 1, especially when 5 N < . This means that the output color is mixed with a higher proportion of blue light at central angles, which is also the reason that the ( ) t θ at central angles increases with increasing N . On the contrary, the CDC at 60~90°° is smaller than 1, indicating that the output color is mixed with less proportion of blue light at edge angles, which causes the ( ) t θ to decrease at edge angles with increasing N . Furthermore, the CDC at different θ tends towards 1 as N increases, meaning that the ( ) YBR θ is more and more closer to the TC , leading to a uniform color distribution. An interesting result is that further increase or decrease of the CDC towards 1 becomes more difficult; after 7 N > , the increase of the (75~90 ) CDC°° and the decrease of the (0~45 ) CDC°° become obviously slower. This means that control of the ( ) YBR θ by ( ) t θ is more and more difficult, and thus it is very difficult for the ( ) YBR θ to be the TC when N is too large. One explanation is that TFS-RPFs have an inside and outside freeform surface instead of the regular spherical surface; when the same blue ray propagates into and escapes from the TFS-RPFs, there is a small difference between the incident angle and the emergent angle owing to the optical refraction at the two interfaces. The inconsistency of the incident and emergent direction may disturb the relationship between the ( ) t θ and the ( ) PB θ ; this inconsistency becomes especially significant when the freeform shapes greatly deviate from the spherical shape, causing t to have no influence on the PB in the same θ and the ( ) CDC θ only showing a slight change. Moreover, a CDC with small θ is more likely to be 1 while a CDC with large θ is likely to be far from 1. Thus, the YBR with smaller θ is more easily controlled by ( ) t θ . We think that the scattering effect of the TFS-RPF and the illumination characteristic of the COB source are responsible for this issue. We calculated the geometry of TFS-RPFs by only using the blue rays emitted directly from the COB source, i.e., not scattered, to adjust the ( ) PB θ through several feedback iteration processes; however, in reality there was a contribution to the ( ) PB θ by the blue rays scattered by phosphor. As for the PB at large θ , its value mainly depends on the scattered blue rays owing to the Lambert illumination distribution of the COB source with lower power at larger θ . On the contrary, the PB at small θ depends mainly on the blue rays that have not been scattered.
Those nonscattered blue rays are adjusted by the ( ) t θ to obtain a certain ( ) PB θ ; thus, it is more effective to control the PB at smaller θ . To explain the CU of COB WLEDs with TFS-RPFs more conveniently, we defined the CU as the maximum deviation of the ( ) YBR θ . It should be noticed that the unbalance of the power distribution is the primary reason for color variation; therefore, we would rather use the power distribution instead of the CIE coordinates to explain the CU. Figure 6 shows the CU and LE of COB WLEDs at different N . It is obvious that the CU increases with increasing N , resulting from the balance between ( ) PB θ and ( ) PY θ distributions as shown in Fig. 4; this further indicates that adjusting ( ) t θ can be used to control the ( ) PB θ and improve the CU. However, the increase becomes less obvious with increasing N , owing to the inconsistency of the propagation direction of the blue rays, phosphor scattering effect, and illumination characteristics of the COB source as discussed above. After 19 feedback iteration processes, optimized TFS-RPFs were achieved with thickness from 0.12 mm to 0.98 mm, as shown in Fig. 5 (a). The optimized COB WLED (when 19 N = , with TFS-RPFs) can achieve a CU of 90.1%, which is an increase of 37.6% compared with the referenced COB WLED (when 0 N = , with TSS-RPFs). Contrarily, the ( ) t θ of COB WLEDs has a slight reduction as N increases, which is mainly owing to the difference in the amount of downconversion events occurring in RPFs. However, the reduction in the LE of the optimized COB WLED is just less than 1% compared with the referenced COB WLED, and can be neglected in practice. According to the theoretical results, the optimized TFS-RPFs and reference TSS-RPF, consisting of silicone and YAG:Ce phosphor (phosphor concentration: 20 wt%), were fabricated by the compression molding method as shown in Fig. 7(a). Then, they were assembled to the COB source with a circular lighting surface generated by 42 square-shaped horizontal blue chips. The COB WLEDs were mounted to a heat sink and spectrally measured with an integrating sphere from Instrument Systems. The drive current was provided by an adjustable DC source from Keithley. The angular CCT distributions were measured by our lab-developed angular CCT tester (T950) at a typical drive current of 350 mA, as shown in Fig. 7(b). The optimized COB WLED with a TFS-RPF has a very uniform CCT distribution almost entirely concentrated on ~4100 K, while the referenced COB WLED with a TSS-RPF has a dispersed CCT distribution ranging from 4000 to 4600 K. The maximum CCT deviation max CCT Δ of the optimized WLED is 141.7 K, which is just 26.2% of that of the reference WLED ( max 540.8 K CCT Δ = ). This demonstrates that the TFS-RPFs can effectively improve the CU of COB WLEDs through reasonable optimization by our proposed method.

Conclusion
In this study, we have improved the CU for COB WLEDs by using TFS-RPFs obtained by the proposed method. The FDTD, Monte Carlo ray-tracing, and CTFB methods were combined to determine the optical properties and geometry of the TFS-RPFs. In order to optimize the CU, we utilized the angular color deviation to modify n ( ) t θ of TFS-RPFs by several CTFB iteration processes, which is realized by the inverse relationship between the amount of blue photons escaping from RPFs and the PRFs thickness. The extended source was also used for calculations, considering the integration effect of blue photons reach the same location of the detector and the RPFs. The results show that increasing N is totally opposite to that of ( ) t θ owing to the CTFB effect. This means that the color balance can be achieved simply by controlling the ( ) PB θ through proper design of the ( ) t θ of the TFS-RPFs. Interestingly, the results also indicate that the inconsistency of the propagation direction of the blue rays, scattering effect of TFS-RPFs, and illumination characteristics of the COB source can potentially affect the impact of the TFS-RPF on the CU of COB WLEDs, which may diminish its ability to control the which is 37.6% more than the reference COB WLED with a TSS-RPF, and the reduction in LE is less than 1% and can be generally neglected. The optimized TFS-RPFs (thickness from 0.12 mm to 0.98 mm) and TSS-RPFs (thickness is 0.5 mm) were fabricated according to the theoretical results and assembled into COB WLEDs. Experimental results show that the max CCT Δ of the optimized COB WLED with a TFS-RPF is 141.7 K, which is just 26.2% of that of the reference COB WLED with a TSS-RPFs ( max 540.8 K CCT Δ = ) at an average CCT of about 4200 K. This demonstrates that the TFS-RPFs can effectively improve the CU of COB WLEDs through reasonable optimization by our proposed method.
In future, much work is still required to study this phosphor architecture, such as with phosphor materials with different scattering abilities and sources with different sizes and illumination characteristics.