High power LED assemblies for solid state lighting – Thermal analysis

Accepted Manuscript Title: High power LED assemblies for solid state lighting thermal analysis Author: Nicolas Kudsieh M. Khizar Bhutta M. Yasin Akhtar Raja PII: DOI: Reference:

S0030-4026(15)00606-3 http://dx.doi.org/doi:10.1016/j.ijleo.2015.07.054 IJLEO 55769

To appear in: Received date: Accepted date:

3-6-2014 7-7-2015

Please cite this article as: N. Kudsieh, M.K. Bhutta, M.Y.A. Raja, High power LED assemblies for solid state lighting - thermal analysis, Optik - International Journal for Light and Electron Optics (2015), http://dx.doi.org/10.1016/j.ijleo.2015.07.054 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

High power LED assemblies for solid state lighting - thermal analysis Nicolas Kudsieh1, M. Khizar Bhutta2, and M. Yasin Akhtar Raja1 1

ip t

Department of Physics and Optical Sciences, Center for Optoelectronics and Optical Communication, University of North

Carolina at Charlotte, 9201 University City Blvd, Charlotte, NC 28223, USA 2

us

cr

Whirlpool Corporation, Benton Harbor, MI 5902257069

We report on the Chip-on-Plate (COP) packaged high power LED assemblies for solid state

an

lighting (SSL) and backlighting display applications from thermal analysis perspective. Three different light modules with different elements (devices) distribution and heatsink design were considered. Thermal resistance circuit (TRC) model and heat diffusion equation were employed

M

in this study of thermal behavior of the assembly elements. Finite element analysis along with 3D modeling and simulations were used for considered lighting modules and assemblies. Also,

d

the effect of thermal uniformity on the emission spectra was evaluated for different operating powers. Comparative study showed that conventional packages with circularly symmetric having

te

non-uniform devices’ distribution per unit area will have temperature differences ~7 oC among the chip-elements operating around 1.25 W power level. Also, it was predicted that high

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

elements temperature difference could result in over 2 nm redshift in the emission spectra accompanied by ~ 4% drop in the output power. However, temperature difference was found to be less than 1 ºC for the uniform rectangular array module distribution, which ensures uniform output intensity for all devices. Such findings are extremely useful for designing larger area LED packaging where thermal non-uniformity effects are expected to be more severe for devices’ longevity and performance.

Key words: light emitting diodes; heatsink; light output; thermal stability; emission spectra

Page 1 of 22

I. Introduction Solid-state light (SSL) sources are playing pivotal role for the design and development of lighting modules for their coherent and non-coherent applications. The low-cost lighting modules

ip t

continue gaining increased interest worldwide owing to their compactness and environmental friendly nature. Nowadays, SSL sources are preferred for most commercial applications because

cr

of their durability, long-lifetimes, low-cost, small size, and their low energy consumption [1],

us

[2]. Despite the major advances highlighted above, some of the basic problems continue to hamper their practical applications. As their market demand increases, it becomes imperative that

an

these problems are addressed efficiently. Thermal stability has been one of the most pressing challenges currently faced by SSL devices [3]. It has been found that all the light sources include

M

LEDs and LDs produce heat when biased using power supplies. Heat is mostly generated by

d

Joule-effect due to series electrical resistance of the device epi-structure and non-ideal quantum

te

efficiency. An increasing device (chip) temperature, further causes their quantum efficiency to drop because heat will excite impurities and defects (electron traps) that will result in less radiative recombinations in the junction region, and thus decrease the intensity of the emitted

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

light [4],[5]. The typical output of an average LED is about 20%, heating presents a further reduction in the output power and the lifetime of SSL modules built using such point sources [6]. In addition to the reduction of the output power; elevated temperatures affect the energy bands structure of the semiconductor materials, and shrink the energy bandgap which gives rise to longer wavelengths in the emission spectrum. This effect leads to a red-shift of the output spectrum [7]. Typically, these point light sources are assembled together in various sizes of arrays depending on their intended application. Some applications use individual devices e.g., short wavelength optical communication. However, solid state lighting, require large assemblies

Page 2 of 22

of LED’s to produce similar or yet higher light intensity than conventional light sources [8][9]. Sometime array of these LEDs are integrated together in different assemblies to achieve recommended levels of optical flux densities.

However, it becomes critical to incorporate

ip t

mandatory thermal management for the optimal performance of the design lighting modules in terms of their color temperature, color purity, and color rendering index (CRI). Thermal

cr

management is a must to ensure the removal of excess heat from the LED chips into the ambient

us

in order to maintain these chips at an acceptable temperature. This results in steady light output, spectral stability and long lifetime. These improvements would ensure efficient, durable, and

an

low-cost light sources, thus help to decrease the transition time needed to complete the conversion from ordinary light sources to next generation SSL technology [10].

M

When array type LED assemblies are designed, it is essential that all chips produce nearly

d

similar light output and have the same thermal behavior. We introduce thermal uniformity

te

among various elements of the assembly as a method to evaluate their thermal stability. Generally, thermal uniformity is realized by mapping the overall temperature difference over the entire assembly-elements. This temperature difference would ensure if there was any issue linked

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

with the poor thermal uniformity; where a small difference implies high thermal uniformity. In this paper, high power LED chips based assemblies have been designed and simulated for their thermal management. Moreover, finite element analysis has also been performed for comparative study of their thermal uniformity in order to enhance the thermal stability that uses specific geometrical distribution of the assembly elements. All assemblies considered in this study use Chip-on-plate (COP) packaged high power AlGaN/GaN based LEDs. COP packaging is one of the most recent technique being exploited for the thermal management of photonicdevices, where high thermal conductivity metal plates are used as submounts. Figure 1 shows a

Page 3 of 22

typical COP package diagram used for each assembly element; a more detailed analysis of the COP package and performance is described elsewhere [11].

ip t

LED chip 10µm Sapphire 90µm Ag solder 40µm

cr

Al plate (submount) 300µm

us

Heatsink

an

FIG. 1. Cross-sectional view of COP LED package (not to the scale) and typical thicknesses of layers are denoted.

II. LED assembly’s thermal uniformity analysis

M

As mentioned, multiple assemblies of GaN/AlGaN high-power LEDs chips were modeled.

d

The aim is to study the thermal behavior of these LED chips using Chip-on-Plate technique. In

te

brief, each LED chip is packaged using COP design where each chip is biased individually. Thermal grease is used to fix the COP chips on the heatsink. Standard plate-fin copper heatsinks of 1-cm in length and 1mm thickness were used. Thermal behavior of individually packaged

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

chips was studied using thermal resistance circuit model (TRC) [12]. As a part of operating power dissipates through LED chips, heat is generated. Heat spreads in the downward direction from the junction layer and dissipates through the heat-sink. Thermal paths of each LED chip are shown in figure 2. It is evident that the main path is vertical from chip to heatsink; however, it branches out horizontally at plane of heatsink base. For simplicity, it is feasible to consider that each thermal path has four horizontal channels branching at the base of the heatsink. Thermal resistance varies on each path depending on the material layers it includes, thus different thermal paths will introduce different thermal resistance values to the TRC. Each thermal path introduces a new thermal resistance to be added to the TRC. Considering the assembly presented above it is

Page 4 of 22

clear that horizontal thermal paths for different assembly elements are location dependant. Inner elements would have longer more uniform horizontal thermal paths rather than outer elements which are close to the edge of the heatsink. This will give inner chips more thermal mass

ip t

underneath them and increase their heat dissipation capabilities. It results in different elements having different total thermal resistances, considering identical devices and operating power

us

cr

values, it is expected to observe different chip temperatures for different elements.

Al plate

d

M

Heat-sink

an

Chip

te

FIG. 2. Thermal paths for one LED assembly element [13]

Thermal behavior of the LED assembly cannot be realized depending on the individual

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

behavior of the elements. Mutual effects among elements play essential part in the final heat distribution and thermal stability of the assembly. These mutual effects can be understood by employing heat diffusion equation to the assembly and solving it at the plane on the heatsink base. The heat diffusion equation (1) can be expressed as:

(1)

Page 5 of 22

Where T(r,t) is the temperature distribution over position and time, K is thermal conductivity, σ is specific heat, and ρ is density of material. The solution of such equation is obtained through separation of variables method and involves Fourier series and selected boundary conditions.

ip t

Boundary conditions include the distance that heat has to travel in the heatsink base plane, which shows the impact of the element location on the thermal behavior, and initial temperature values;

cr

which will vary depending on the effects of the neighboring elements. Having high initial

us

temperature in the surrounding perimeter of an assembly/module, this will decrease the amount of heat dissipated from the element to the heatsink. Therefore, elements arrangement and

an

positions directly affect the final temperature distribution and thermal uniformity of the assembly. Non-uniform thermal behavior, if increased, will create instability in output power and

M

emission spectrum among the assembly elements. This might results in different color rendering,

d

temperatures and lifetimes of the elements. It is essential to take these effects in consideration,

te

designing high power LED assemblies for SSL. We found that such undesired effects can be suppressed by increasing the symmetry and uniformity of the assembly. Uniform arrangement of the elements/device on the heatsink surface creates relatively equal horizontal thermal paths.

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Constant element density per unit area will unify the effects of neighboring elements in the assembly to almost a constant value. This will create homogenous boundary conditions and initial temperature for the diffusion equation, thus allows similar heat diffusion from each element. All these effects will result in similar chip temperature for all assembly elements and secure more stable output and longer lifetime of the assembly as a whole.

Page 6 of 22

III. Thermal modeling and simulation In this study multiple high power AlGaN/GaN LEDs of die size 0.5×0.5 mm2 are used for

analysis by using

ip t

numerical modeling and analysis to create different assemblies/lighting modules for their thermal ANSYS 13TM finite element code. Three different lighting-modules were

cr

modeled and analyzed in order to study the effect of thermal uniformity of the generated heat

us

[14] [15]. A comparative study was also carried out to investigate the effects of lighting-modules /assembly element arrangement for corresponding chips temperature, emission spectra and

an

output. Three type lighting-modules/assemblies were modeled in this study. The first module design (A) consists of 12 LED chips having radial symmetry, were attached to a 1cm radius

M

cylindrical plate-fin heatsink of 1cm fin length and 1mm thickness. The second design (B) is an extension of the first design with total of 24 LED chips attached to a bigger cylindrical plate-fin

d

heat-sink of similar fin dimensions to design (A). For this design, only half of the structure was

te

modeled (based on symmetry) in order to minimize the processing time and to overcome those limitations linked with the use of the FEM analyzer.

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

All necessary measures were taken to ensure that half domain modeling yields the similar results to the whole structure; this was accomplished by setting the boundary thermal loads to zero interaction at the plane of division. The third module design (C) is a 12-element rectangulardistribution assembly with constant element per-unit area density attached to a rectangular platefin heat-sink. In order to keep similar boundary conditions for the dissipation of heat, it was ensured that all assemblies have similar fin dimensions. All three assemblies were ‘modeled and meshed’ carefully with special attention to the chip-heatsink contact regions for consistency. Modeled assemblies/lighting modules and their meshed structures are presented in figures 3 and figure 4, respectively.

Page 7 of 22

ip t cr us

an

FIG. 3. 3D models of different LED assemblies/modules. (i) Lighting-module (A), (ii) Lighting-module (B), and (iii) Lighting-module (C)

M

All lighting modules/assemblies were simulated using similar set of operating power values ranging from 0.25 W to 1.25 W. Conduction and convection loads were carefully applied to each

d

surface and interface to create real-time thermal behavior. Results of steady-state thermal

te

analysis were used to determine the maximum temperature difference between the elements of each module/assembly at different operating powers. Comparative study was carried out to show

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

which of the structure achieved better thermal uniformity. In addition to the temperature contrast calculations, emission spectrum and output intensity comparative study was also carried out using Rsoft LaserMODTM numerical simulator to compare the output power difference between different assembly elements.

Page 8 of 22

ip t

FIG. 4. 3D meshed models of different LED assemblies. (i) Module/Assembly (A), (ii) Module/Assembly (B), and

cr

(iii) Module/Assembly (C)

us

IV. Results and Discussion

an

Steady-state thermal analysis was carried out for the three separately designed lighting modules assemblies accordingly. Numerically simulated results were analyzed systematically in

M

order to determine the temperature differences ranges for each chip mounted on the respective module/assemblies for different power values. It is evident that chips temperature contrast is

d

related to the elements layout and arrangements. It was found that for the case of more uniform

te

and constant element-per-area assemblies relatively lower temperature difference between the chips was observed. For 1.25W, assembly (A) showed the highest temperature contrast of 7 oC

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

whereas for the case of assembly (C) which has almost constant element-per-area density showed lowest temperature contrast value of 0.7 oC. Whereas, assembly (B) which has relatively more uniform element distribution compared to (A) showed less temperature difference between its chips with a value of 4.5 oC at 1.25W. It is essential to recall that the temperature difference mentioned above is different from the chip temperature; here we just calculated ΔT the temperature change from one device to the other. Chip temperature is a function of several factors, most importantly are the heatsink design and its dimensions. Figure 5 shows the temperature contrast for assemblies (A), (B), and (C) as a function of input power. Heat distribution contours were also used to investigate heat dissipation capabilities of each assembly.

Page 9 of 22

Figure 6 shows temperature distribution in each heatsink at nominal operating power of 1.25W. It is evident that assembly (C) has the most uniform temperature contours, which indicates that

M

an

us

cr

ip t

its elements have relatively similar heat dissipation capabilities and thus similar temperature.

te

d

FIG. 5. Chip temperature difference ΔT of assemblies (A), (B), and (C) at different operating powers

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

FIG. 6. Temperature contours in heatsink at 1.25 W power (i) assembly (A), (ii) assembly (B), (iii) assembly (C)

The temperature difference has an impact on the emission spectrum and output power of the assembly elements. In order to appreciate these effects, Rsoft LaserMODTM was utilized to investigate the changes in the emission spectrum and output intensity among assembly elements;

Page 10 of 22

all the three designs presented in this paper were considered. Chip temperature difference can increase if the operating power is increased, or the number of elements is increased ie; designing bigger light modules with more LED devices, in this study the operating power was elevated to

ip t

simulate such conditions, all devices were modeled at higher power values ~ 2 W. temperature results were obtained and used to simulate the performance of the LED device using RSoft

cr

LasreMODTM to see the thermal instability on the emission spectrum and output of the assembly

us

elements. Temperature difference in the case of assembly (A) reached about 13 oC, while it was about 1 oC for assembly (C) for the same high power values. This difference resulted in over 2

an

nm red shift and 4% intensity drop for the hot chips of assembly (A), whereas the resulting small difference in (C) kept the emission spectra identical for all elements of assembly (C). This output

M

stability in (C) ensure better performance and longer lifetime of the assembly and makes the

d

rectangular design with uniform element distribution the best option of the three investigated .

te

Figures 7 and 8 show photoluminescence (PL) emission spectrum and intensity for minimum and maximum temperature of chips in assemblies (A), and (C), respectively.

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

FIG. 7. PL emission spectrum for min/max temperature elements of assembly (A) at 2 W

Page 11 of 22

ip t cr us

an

FIG. 8. PL emission spectrum for min/max temperature elements of assembly (C) at 2 W

M

V. Conclusion

We have investigated thermal behavior of multiple high-power chip-on-plate packaged

d

LED design assemblies for solid state lighting, using the simulation and numerical analysis. The

te

research was aimed towards performance and thermal uniformity. Three assemblies were designed, and modeled for this study; two different sizes circular with radial element symmetry

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

and one with rectangular equal-spaced element distribution, assemblies were considered. Minimum and maximum chips temperatures were calculated in each assembly for different operating power values. By employing TRC model and heat diffusion equation, it was expected that rectangular assembly would achieve more thermal uniformity due to longer horizontal conductive thermal diffusion paths, and relatively similar initial boundary conditions in the heatsink base plane. As noted earlier, the three assemblies were modeled and simulated using ANSYS-13TM FFM software. Results showed that there is a temperature difference among LED chips in each assembly. Furthermore, the difference is an order of magnitude higher in the circular assembly than that of the rectangular one. The temperature results, thermal contours are

Page 12 of 22

consistent and come in agreement with our theoretical anticipation. Moreover, the effects of these findings were studied on the emission spectra for a comparative study. Those clearly show the effects of thermal non-uniformity on small red-shift and intensity drop observed in assembly (A),

ip t

but none appeared in assembly (C); which shows that more uniform element distribution leads to better thermal uniformity and device performance. This study provides a better understanding of

cr

the thermal behavior of LED assemblies and would be of significance designing much large-

an

us

numbered LED chip-assemblies and subsystems in lighting systems for various applications.

VI. References

bulk GaN”, Appl. Phys. Lett. (2001), 79, 711-712.

M

[1]. T. Nishida, H. Saito, and N. Kobayashi , “Efficient and high-power AlGaN-based ultraviolet light-emitting diode grown on

d

[2]. G. Zhang, S. Feng, Z. Zhou, J. Liu, J Li, and H Zhu,” Thermal Fatigue Characteristics of Die Attach Materials for Packaged High-Brightness LEDs” in Components, Packaging Manufacturing Technology IEEE vol. 2,pp 1346-1353.

te

[3]. J S Huang, Temperature and current dependences of reliability degradation of buried heterostructure semiconductor lasers. IEEE Trans Dev Mat Reliab,5(1), (2005).

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

[4]. M. Khizar, and M. Y. A Raja," Optical output power degradation of AlGaN-based deep-UV Light Emitting Diodes by plasma treatment" Proc. IEEE-SouthEastCon 2007, pp 584-589 [5]. L. Marona , P. Wisnewski, P. Prystawko,I. Grzegory, T. Suski, S. Porowski, et al. “Degradation mechanisms in InGaN laser diodes grown bulk GaN crystals”. Appl. Phys Lett. 8, (2006). [6]. N. Kudsieh, M. Khizar, A. Shah, and M.Y.A Raja , “Transient Thermal Analysis of lnGaN/GaN Laser Diodes” , Proc High-Capacity Optical Networks and Enabling Technologies (HONET), 2010 , pp.121-127 [7]. P. Mashkov, T. Pencheva, and B. Gyoch,” Lamp with Rigid LED Strip-Design and Thermal Management,” in Electron. Technol. (ISSE) 2011, pp. 39-44 [8]. J-C. Shyu “Performance of the LED array panel in confined enclosure”, in Microsystems Packaging Assembly and Circuit Technology Conference, 2010 5TH Int’l, pp.1-4 [9]. S.J. Chang, C.S. Chang, Y.K. Su , C.T. Lee , W.S Chen, C.F Shen, Y. P Hsu, S.C Shei, H.C. Lo “Nitride-Based Flip-Chip ITO LEDs”, Advanced Packaging, IEEE Transactions on Vol: 28 ,2005 pp 273 – 277

Page 13 of 22

[10]. M.Y. Tsai, C.H. Chen, and C.S. Kang “Thermal Analyses and Measurements of Low-Cost COP Package for High-Power LED” presented at Electronic Components and Technology Conference (ECTC) 2008, 58 TH pp. 1812-1818. [11]. N. Kudsieh, M. Khizar, and M. Y. A. Raja, “Thermal Modeling of Specialty Heat-sinks for Low-cost COP Packaging for

ip t

High-power LEDs”, High Capacity Optical Networks and Enabling Technologies (HONET), 9th, 2012 pp. 087- 091 [12].W Sun Lee, K-Y Byun “The availability of the Thermal Resistance Model in Flip-chip Packages” in Electronic Material and Packaging EMP 2007,pp. 1-5

cr

[13].N. Kudsieh, M. Khizar, and M. Y. A. Raja, “Controlling thermal stability among LED chips assembly packages for highpower solid-state”, High-Capacity Optical Networks and Enabling Technologies (HONET), 2013, pp 78-83

te

d

M

an

[15]. ANSYS, Inc, thermal analysis guide, ANSYS, Inc release 12, 2009.

us

[14]. G.Dhatt., E Lefranois, and G Toutzo, Finite element method John Wiley & sons, Dec 2012

Ac ce p

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65

Page 14 of 22

ip t

Figure file

cr

LED chip 10µm

Ag solder 40µm

us

Sapphire 90µm

an

Al plate (submount) 300µm

M

Heatsink

Ac ce p

te

d

Figure 1. Cross-sectional view of COP LED package (not to the scale) and typical thicknesses of layers are denoted

Page 15 of 22

te

d

M

an

Heat-sink

cr

ip t Al plate

us

Chip

Ac ce p

Figure. 2. Thermal paths for one LED assembly element

Page 16 of 22

ip t cr us an M d te Ac ce p Figure. 3. 3D models of different LED assemblies/modules. (i) Lighting-module (A), (ii) Lighting-module (B), and (iii) Lighting-module (C)

Page 17 of 22

ip t cr us an M d te Ac ce p Figure. 4. 3D meshed models of different LED assemblies. (i) Module/Assembly (A), (ii) Module/Assembly (B), and (iii) Module/Assembly (C)

Page 18 of 22

ip t cr us an

Ac ce p

te

d

M

Figure. 5. Chip temperature difference ΔT of assemblies (A), (B), and (C) at different operating powers

Page 19 of 22

ip t cr us an M d te Ac ce p Figure. 6. Temperature contours in heatsink at 1.25 W power (i) assembly (A), (ii) assembly (B), (iii) assembly (C)

Page 20 of 22

ip t cr us an M Ac ce p

te

d

Figure. 7. PL emission spectrum for min/max temperature elements of assembly (A) at 2 W

Page 21 of 22

ip t cr us an M d

Ac ce p

te

Figure. 8. PL emission spectrum for min/max temperature elements of assembly (C) at 2 W

Page 22 of 22