Ultra-Thin Ceramic Substrates for Improved Heatsinking for microLEDs

Abstract

MicroLEDs provide unrivaled luminance and operating lifetime, which has led to significant activity using devices for display and non-display applications. The small size and high power density of microLEDs, however, causes increased adverse heating effects which can limit performance. A new generation of electrically insulating high thermal conductivity materials, such as alumina, has been proposed to mitigate these thermal effects when used as a substrate as an alternative to glass. This strategy then could be used as a method of passive heatsinking to improve the overall performance of the microLED. In this work, a newly available material, an 80 micron thick alumina ceramic substrate, is shown to yield a 30 % improvement on average in the maximum current drive over a glass substrate.

Graphical Abstract

The small size of current microLEDs creates intensely localized heating during operation that can cause the microLED to degrade rapidly. While several active heatsinking options have been used, passive heatsinking by engineering the substrate can provide superior heatsinking without any power draw. Ultrathin ceramic substrates, like those available from Corning, are an ideal candidate for such a solution

1. Introduction

MicroLED technology based on Gallium Nitride (GaN) chiplets offers exceptionally high luminance, high luminous efficiency, and a long operating lifetime compared to competing approaches such as organic light emitting diodes. [1, 2] Due to these desirable characteristics, microLEDs are actively explored for applications in Augmented Reality (AR)/Virtual Reality (VR) displays, wearable devices, optogenetics, biomedical computational imaging, and visible light communications. [3, 4, 5, 6, 7, 8] The high luminance of microLEDs is in part due to the high current density operation achievable in LEDs; this high current density operation enables a high rate of radiative recombination yielding a high rate of photon generation, but also causes substantial self-heating resulting in increased device temperature. [9] This elevated temperature activates non-radiative recombination processes and has a detrimental effect on the efficiency of microLEDs. [10] Due to the small size of the devices, the onset and impact of the self-heating can be dramatic and lead to destruction of the device if not managed carefully.

Various recombination and transport mechanisms have been proposed as the potential reasons for this thermal efficiency reduction. Auger recombination has been shown to be one of the causes of temperature-dependent non-radiative recombination[11, 12]. Additionally, Shockley-Read-Hall (SRH) recombination due to the presence of trap states is also temperature dependent and causes a reduction in efficiency with an increase in operating temperature. The droop in efficiency due to temperature is exacerbated for high power LEDs with larger mesa dimensions. [9] The substantial self-heating of microLEDs has proved to be a major obstacle in realizing the full potential of the microLED technology for applications requiring high current density operation. Several works have attempted to understand the issue of efficiency droop caused by increased device temperature. [9, 10, 12, 13, 14, 15, 16] These works have demonstrated that an effective heat sinking/dissipation technology integration is required to overcome the detrimental performance effects of self-heating. One approach shown by Li et al. used cooling systems based on thermoelectric couple (TEC) for heat dissipation in high power LEDs. [17, 18] A recent study by Guo et al. enhanced heat dissipation in GaN based microLEDs by leveraging the piezo-phototronic effect. [19] Both of these approaches are active approaches to heatsinking, but a successful passive approach on an insulating substrate has not yet been realized.

One such method for passive heatsinking is to carefully engineer the substrate. High thermal conductivity materials can draw heat away from the active area to spread the heat laterally and provide passive heatsinking. Corning Alumina Ribbon Ceramic offers a promising option due to its high thermal conductivity (> 36 W/mK), small thickness (80 μ m), and large format. [20] The ultra-thin thickness of the ceramic substrate aids heat sinking by minimizing thermal mass. This substrate is an ideal test bench for using a high thermal conductivity material as a substrate for improved heat sinking in microLEDs. In this work, this ceramic substrate enables increased current drive and, therefore, higher density operation for the microLED by improved heatsinking.

2. Simulation

To illustrate the advantage of a substrate with high thermal conductivity for heatsinking, the heating generated by a power source, like a microLED, was simulated on a high and low thermal conductivity substrate.[21] The result is shown in Fig. 1. More simulation details are provided in the Experimental Section. The simulation shows substantially more heating for the low thermal conductivity sample than the high thermal conductivity sample, proving the feasibility of this approach. Alumina is one potential option for a high thermal conductivity substrate material with measured values of 11.0–36.0 W/(m·K), in comparison to glass which is 0.8 – 1.06 W/(m·K), but large size, thin alumina substrates can be difficult to manufacture with conventional approaches. [22] Corning Alumina Ribbon Ceramic offers a promising option due to its high thermal conductivity (> 36 W/mK), thin thickness (80 μ m), and large format. [20] The ultra thin thickness of the ceramic substrate aids heat sinking by minimizing thermal mass.

Figure 1:

Simulation of power source on (a) glass and (b) ultra-thin ceramic substrate for one minute of operation with corresponding infrared images after one minute of continuous operation on (c) glass and (d) ultra-thin ceramic substrate.

MicroLEDs were placed on both glass and ultra-thin ceramic substrates as described in the Experimental Section to compare real-life operation heating with the simulated results. The heating as observed with a FLIR infrared camera for two representative samples on glass and ultra-thin ceramic substrate after one minute of driving at 3V are shown in Figs. 1c and 1d. These results reflect those of the simulation shown in Fig.1. These measurements show a maximum operating temperature difference of 17.0 ± 3.4 °C between the two substrates when the error in the FLIR camera measurement, as discussed in the supporting information, is taken into account.

3. I-V Curve Speed Testing

In order to quantify the improved performance of the ultra-thin ceramic substrate as a heatsink for microLEDs, several tests were performed utilizing the I-V curves of three microLEDs on each substrate (glass and ultra-thin ceramic). The first test involved increasing the time between points taken in the I-V curve, thus extending the overall time to take the I-V curve and the heating experienced by the microLED. The delay between subsequent points was increased from 100 ms to 20000ms. Infrared images of a device on a glass and ultra-thin ceramic substrate at the end of three representative speeds (200ms, 1000ms, 10000ms) are shown in Fig. 2. All infrared images taken can be seen in the Supporting Information in addition to the calibration data for the FLIR camera. A significant temperature difference, in one case exceeding 20 °C, was seen between the microLEDs on the glass substrates and the microLEDs on the ceramic substrates.

Figure 2:

Infrared images of devices on ultra-thin ceramic substrates at end of each ramp test: a) glass, 200 ms b) glass, 1000 ms c) glass, 10000 ms d) corning, 200 ms e) corning, 1000 ms f) corning, 10000 ms

In addition to the I-V curves, the power of the microLED at each point was observed in order to obtain the external quantum efficiency (EQE). All I-V curves taken with corresponding EQE for the “ON” region are shown in the Supporting Information. The change in maximum EQE for increasing time delay between I-V curve points for the three devices tested on a glass substrate is shown in Fig. 3a. All devices show decreasing EQE above 2000 ms, with a maximum decrease of over 60 %. The corresponding plot for the devices on ultra-thin ceramic substrates is shown in Fig. 3b. For the devices on ultra-thin ceramic substrates, no such decrease is observed, even at the largest time delay between points. Additionally, several of the devices on glass were destroyed by this testing, whereas the samples on the ultra-thin ceramic substrate continued to function normally.

Figure 3:

Change in EQE as a result of ramped I-V curve testing for microLEDs on (a) glass and (b) ultra-thin ceramic substrates

4. Drive Time Tests

A second experiment was performed to observe the effect on the overall EQE of the device after increasing drive times causing heating in the device. Three devices on glass and ultra-thin ceramic substrates were driven at 3V for increasing times from 1 minute to 30 minutes. After the drive time was complete, an I-V curve was collected at the maximum speed allowed by the Keithley and LabVIEW interface to minimize heating incurred by the I-V curve collection. A power meter was once again used to determine the EQE during the I-V curve. Infrared images were taken for a device on glass and two devices on the ultra-thin ceramic substrate at the end of each drive interval before I-V curve collection. A representative set of these images are shown in Fig. 4. All images taken are shown in the Supporting Information. There is decrease in heating at the longest drive times likely due to degradation of the microLED for the sample on the glass substrate.

Figure 4:

Infrared images of samples at end of each drive time test on glass: a) 1 min b) 5 min c) 10 min and ceramic tape d) 1 min e) 5 min f) 10 min

All I-V and EQE curves are shown in the Supporting Information. The change in EQE for devices on a glass substrate as a result of each drive interval is shown in Fig. 5a. The corresponding results for devices on the ultrathin ceramic substrate are shown in Fig. 5b. Two of the three devices on the ultra-thin ceramic substrate show more degradation than the third. There are a few potential explanations for this behavior. One such explanation is manufacturing tolerances between the purchased microLEDs. The placement technique used in this work can create minor differences in overall resistance and thermal contact between the device and substrate. These small differences may play a larger role during long drive times causing increased heating. The infrared images for each experiment shown in the Supporting Information do illustrate that substantially more heating is incurred in this test. The maximum temperature observed for the devices on glass in the ramped I-V curve test was 47.4°C, whereas for the drive tests it was 53.6°C. However, even with these variations, all devices on ultra-thin ceramic substrates show less degradation than those on glass. The maximum degradation observed for the devices on the ultra-thin ceramic substrates is 50.4 % and the minimum degradation of the devices on glass is 53.1 %, extending up to 76.9 %. This provides an improvement then of at least 6 % and up to 52.9 %.

Figure 5:

Change in EQE as a result of drive time tests for microLEDs on (a) glass and (b) ultra-thin ceramic substrates

5. Destruction Testing

To probe if the operational limits of the microLEDs was improved by the heatsinking capabilities of the ultrathin ceramic substrate, three devices on each substrates were driven with increasing current up to 0.5A and the point at which each device failed was observed. This test was used to understand the advantage the ultra-thin ceramic substrate provides for high current drive applications. The optical power of each microLED as measured by the power meter as a function of the increasing drive current is shown in Fig. 6. The power curves for the devices on the ultra-thin ceramic substrate can be shown to extend further along the drive current axis.

Figure 6:

MicroLED power as a function of current during destruction testing for three microLEDs on glass and ultra-thin ceramic substrates

The destruction point of each of these devices as taken from Fig. 6 is shown in Table 1. All devices on the ultrathin ceramic substrate outperformed those on glass. The difference in performance ranged from 0.02A up to 0.14A, with an average change of 0.093A. This average change represents a 30% improvement overall. The drive current at the breaking point of the glass samples, 0.31A, is a very high current drive for these microLEDs. The improvement offered by the ceramic substrate then is significant for applications taking advantage of the high luminance of microLEDs, where high current drive is common.

Table 1:

Table of destruction points for microLEDs on glass and ultra-thin ceramic substrates.

Substrate	Device	Destruction Point (A)	Average (A)
	1	0.3
Glass	2	0.33	0.31
	3	0.3
	1	0.44
Ultra-thin Ceramic	2	0.35	0.40
	3	0.42

6. Conclusion

MicroLEDs placed on the ultra-thin ceramic substrate outperform all devices placed on glass devices due to the high thermal conductivity of the ultra-thin ceramic substrate. This finding bolsters the hypothesis that high thermal conductivity materials can be used successfully as a substrate for microLEDs for passive heat sinking. Future experiments could be homogenized by improving the method of placing the microLEDs, such as the use of a pick-and-place machine. Other avenues for future work would be to explore the heat sinking effect of the ultrathin ceramic substrate for microLEDs with varying wavelengths or larger microLED arrays.

7. Experimental Section

7.1. Fabrication

Glass substrates (5 cm by 5 cm) were procured from LumTec Inc and ultra-thin ceramic substrates (Corning Alumina Ribbon Ceramic, 5 cm by 5 cm) were provided by Corning. All samples were cleaned by sonication in DI water, acetone, and isopropyl alcohol for ten minutes each and dried with a nitrogen gun. To create a footprint for the microLEDs, Cr/Au (5 nm/100 nm) was deposited by e-beam evaporation, and then patterned using standard lithographic methods before wet etching Cr/Au. Remaining photoresist was stripped using AZ 400T stripper. In order to attach the microLEDS (CREE C527DA2432-0017-A) to the footprint, silver paste was placed on the footprint using a tungsten probe tip, and then a probe tip and micromanipulator is used to move the microLED into place. The sample was then placed on a hot plate at 150C for at least 30 minutes to cure the silver paste.

7.2. Simulation

Energy2d (https://energy.concord.org/energy2d/) was used to make 2D models of a power source on glass and alumina substrates. The value for the power source used was chosen as the maximum power (0.3 W) used by a microLED in this experiment. The simulation time for each device was approximately one minute. The thermal conductivity of alumina used was approximately ten times that of glass to reflect the value mismatch of the materials shown in literature. Commonly accepted values for glass and alumina were used for all other parameters. The boundary condition chosen between the heat source, meant to represent the microLED, and the substrate was constant heat flux as the microLED is expected to emit a constant power.

7.3. I-V Curve Collection

For all I-V curve collection, the footprint of the microLED was probed using tungsten probe tips. The probes were connected to a Keithley Source Measure Unit (Keithley 24000 series) controlled by LabVIEW to measure the I-V curve and record the resulting data. An optical power meter (Thorlabs) is used simultaneously to record the optical power from the microLED to then calculate the EQE of the microLED. I-V curves were taking for two types of testing explained below. All I-V curves were swept between −1 and 3 V in steps of 0.05 V. Baseline data for each microLED was taken before both ramp and time testing with the time between points set at 100 ms. Five sets of data were taken and averaged to create an accurate baseline for each sample. All tests were run on three devices on both the ultra-thin ceramic and glass substrate.

7.4. I-V Curve Ramp Tests

For the I-V curve ramp tests, the time between points taken in the I-V curve was increased from 100 ms to 20000 ms with the following intervals: 100 ms, 200 ms, 500 ms, 1000 ms, 2000 ms, 5000 ms, 10000 ms, and 20000 ms. Infrared images captured with a FLIR handheld camera showing the microLED operating temperature and the collected I-V curves are shown in the Supporting Information for both glass and ultra-thin ceramic substrates. A calibration for the FLIR camera measurements is also shown in the Supporting Information.

7.5. Drive Time Tests

For the drive time tests, the microLED was set to a constant bias for set amounts of time. The microLED was driven at 3V for increasing times between 1 and 30 minutes. The time intervals used were 1, 5, 10, 15, and 30 minutes. After each interval, an I-V curve was taken between 1 and 3V with a step size of 0.05V and a time delay of 100 ms between points. Infrared images and I-V curves from this experiment are shown in the Supporting Information.

7.6. Destruction Testing

Destruction testing was done by increasing the current drive across microLEDs on glass and ultra-thin ceramic substrates until the microLED stopped emitting light, defined as the point at which the power observed by the power meter goes below 1E-8 W. This point was chosen as it is at least three orders of magnitude lower than the measured power of any microLED tested.