Investigation of the long-term aging characteristics of chip-on-board LEDs

Abstract

There are four main classes of light-emitting diode (LED) package platforms that are currently used in solid-state lighting (SSL) technologies. These classes of packaging platforms are high-power light-emitting diodes (HP-LEDs), mid-power LEDs (MP-LEDs), chip-scale package (CSP) LEDs, and chip-on-board (COB) LEDs. Each MP-LED, HP-LED, and CSP LED package typically has only one to three LED die that are energized through two electrodes (i.e., an anode and a cathode). In contrast, COB LED packages can contain many more LED die in a single package, but still are energized through only two electrodes. As a result of the inherent high LED die count, the COB LED package platform offers the highest lighting flux density available in a thin, low-profile package. For example, it is possible to produce several thousand lumens (lm) from the light-emitting surface (LES) of a COB LED that occupies as little as 0.25 square inches (1.6 square centimeters [cm]). COB LEDs achieve this level of performance by integrating many MP-LED die into the LES and electrically connecting them through wire bond or solder interconnects. The LED array is directly mounted on an excellent thermal conductor such as a metal-core printed circuit board (MCPCB) or a ceramic substrate to manage the waste heat. COB LED packages are typically set to a fixed correlated color temperature (CCT) value. Some advanced COB LED packages can contain two or more LED primaries set to different CCT values and can be tuned between the CCT values of the LED primaries by adjusting the current to each. The COB LED package for such products contains separate electrodes for each LED primary to allow tunable-white operation.
The first SSL products typically used either HP-LEDs or MP-LEDs. Each of these packaging platforms were found to have particular mechanisms related to the LED package that caused either luminous flux depreciation or chromaticity shift. Because a decade of laboratory research and field performance data have been collected, analyzed, and are available, the degradation mechanisms of HP-LEDs and MP-LEDs are better understood today than in the early days of SSL products. However, COB LEDs are a relatively new LED packaging technology; therefore, the failure mechanisms specific to this LED package platform are not as well understood. The primary failure mechanisms associated with COB LEDs are thought to arise from two driving factors—the amount of heat generated in the COB LED package and the large area of silicone in the package, which can lead to moisture ingress.
This report is the first in a series of reports about the reliability of commercial COB LED products; therefore, it will focus specifically on thermal issues in COB LEDs. In this report, reliability is judged by either lights-out failures caused by electrical open circuits or parametric failures caused by excessive luminous flux depreciation or chromaticity shifts. Six different COB LED products from different Tier 1 manufacturers were included in this study. All of the COB LEDs were phosphor-converted LEDs (pc-LEDs), and the LES of the devices under test (DUTs) varied from 4.5 millimeters (mm) to 14.6 mm. Most DUTs were set to a fixed CCT value, but the CCT values of two classes of DUTs were tunable; therefore, this study offers new insights into the behavior of this type of LED products.
The initial part of this study consisted of a comparison of the construction techniques used in the different DUTs. Four out of the six classes of DUTs examined during this study were built on MCPCBs, with the remaining two classes of DUTs being built on ceramic substrates. Five out of the six classes of DUTs contained LED die that were interconnected by wire bonding, and the LED die in these products were mounted to the substrate with a thermally conductive adhesive. The LED die in the sixth class of DUTs were flip-chip bonded to an interposer layer on the aluminum nitride (AlN) substrate through solder bumps. The number of LEDs in the LES ranged from 12 (for a 4.5-mm LES) to 108 (for a 14.6-mm LED). The radiant efficiency of these devices ranged from 0.31 to 0.44, and the luminous efficacy of these samples ranged from 83 lumens per watt (lm/W) to 127 lm/W.
Because the LED die used in COB LED packages are similar in size to those used in MP-LEDs, a tunable-white MP-LED product was chosen for comparison purposes. The tunable-white MP-LED DUTs contained two independent circuits of one LED die each, with one circuit tuned to a nominal CCT value of 2,700 Kelvin (K) and the other circuit tuned to a nominal CCT value of 6,500 K. The MP-LED DUTs had the two LED die bonded on a silver lead frame and connected via wire bonds. As a whole, this structure is analogous to many of the COB LEDs examined during this study, but it operates at a lower power level and produces less waste heat. The MP-LED product was found to be more efficient than any of the COB LED products with radiant efficiency of the MP-LED product between 0.47 and 0.52 (depending on the CCT value), and luminous efficacies between 152 lm/W to 162 lm/W (depending on CCT value). The authors of this report attribute the higher efficiency of the MP-LED package to the smaller amount of waste thermal energy that must be dissipated and the smaller number of absorbing surfaces (e.g., other LED die, wire bonds) in the MP-LED package compared with the COB LED package.
Prior to testing the COB LED DUTs, the samples were mounted on appropriate heat sinks intended to keep the case temperature (Tc) of the COB LED package within the manufacturer’s specifications. Room temperature operating life (RTOL) tests were conducted on these COB LED DUTs (mounted to a heat sink by using a silicone thermal grease as a thermal interface material [TIM]) in still air, and the Tc values varied between 69 degrees Celsius (°C) and 113°C. The luminous flux maintenance (LFM) data were analyzed by using an exponential decay model as indicated in the American National Standards Institute (ANSI) and Illuminating Engineering Society (IES) technical memorandum (TM) ANSI/IES TM-21-19. The LFM values in RTOL for many DUTs increased during the first 6,000 hours (hrs) of operation, indicating an improvement in efficiency, and they were assigned a decay rate constant (α) value of 2.0 × 10-6 as required by ANSI/IES TM-21-19. Minimal chromaticity shift was observed for the devices during RTOL tests. Therefore, neither “lights out” nor parametric failures were observed through 6,000 hrs of the RTOL tests.
Separate populations of some of the DUTs were also operated in elevated ambient environment of 75°C, during 75°C operational life (75OL) tests in an environmental chamber with high air circulation. The 75OL tests proved to be a higher stress environment that produced α values significantly higher than those measured in RTOL tests even though the Tc values were not that different. The reason for the higher stress levels in 75OL is not clear since the Tc values are expected to be similar in 75OL (with high air recirculation) and RTOL (with no forced air circulation). One possible cause of this difference in α values was a change in the thermal resistance of at least one interface (e.g., the TIM) over time because of degradation, but this could not be confirmed experimentally. The 75OL condition did accelerate luminous flux depreciation, but all DUTs exhibited LFM values well above the parametric failure threshold (i.e., LFM ≤ 0.7) during the 6,000-hr test duration. In addition, the chromaticity shifts of the COB LED DUTs during the 75OL test were similar to those observed during the RTOL test, but the magnitude of the shifts was greater in 75OL. As a result, parametric failure because of excessive chromaticity shift (i.e., Δu'v' ≤ 0.007) occurred for some DUTs in 75OL but not all.
All DUTs examined during this study exhibited either shifts in the blue or green direction or a combination of the two during the test period; either shift is indicative of chromaticity shift mode (CSM) behavior that can be classified as CSM-1 or CSM-2. CSM-1 is usually caused by a relative increase in emissions from the LED emitter, whereas CSM-2 is usually caused by a change in the emission spectrum of warm white phosphors. In keeping with these previous findings, CSM-1 behavior was observed exclusively for the cool white DUTs examined during this study; in contrast, many of the 2,700 K DUTs included in this report exhibited mostly CSM-2 behavior. In some instances, both CSM-1 and CSM-2 behaviors were observed simultaneously, which suggests that both CSMs may be active. No DUTs evaluated during the RTOL test exhibited a chromaticity shift that exceeded the normal parametric thresholds from chromaticity shift failure (e.g., Δu'v' = 0.007, Δu'v' = 0.004), but both the neutral white and cool white settings of the tunable COB product can be classified as parametric failures because of excess chromaticity shift during the 75OL test.
In order to understand the performance of the LEDs used in the mid-power package, several of the product types investigated in this report were decapped to remove the silicone and phosphor layer. The decapping results indicated that the external quantum efficiencies (EQE) of the COB LEDs ranged between 0.46 and 0.67. This lowest EQE value, 0.46, was observed for a violet-pumped COB LED, and the reduced EQE value may be related to the LED epitaxy or the alumina (Al2O3) substrate. The best performing COB LED product had a EQE of 0.67 and was the only flip-chip bonded product examined during this study. The performance of the flip-chip COB LED product was similar to that of the MP-LED package that was used as the benchmark, which had an EQE of 0.66.
COB LEDs are an emerging LED packaging platform for SSL products that offers products in high lighting density and in a thin profile. Recent advances in packaging of COB LED products have also provided tunable light capabilities in this package platform either as a dim-to-warm product or as a white-tunable product. However, this initial report underscores the importance of thermal management with COB LED technologies because of the high density of waste heat that is generated by the small package. During room temperature testing, where the Tc was properly controlled, the luminous flux maintenance performance of the COB LED products was excellent, with most decay rate constants calculated to be at the minimum value allowed by ANSI/IES TM-21-19. COB LED packages offer new capabilities to the SSL industry in a denser package than those offered by other platforms. The high light density in a thin profile may help to create new SSL products with advanced lighting features such as tunable white or unique optical patterns for light delivery. The key to capitalizing on these new capabilities is proper design of the luminous containing the COB LED, especially the thermal management pathways.