This paper reports the results of a reliability investigation performed on four different groups of commerciallyavailablemid-powerwhite LEDs. In order to determine the robustness of this continuously growing class of light-ing devices we arranged an experiment of accelerated aging: the four groups of LEDs (from different manufac-turers) were submitted to a series of stress tests in environmental chambers with set-point temperaturesranging from 45 °C to 105 °C, in accordance to the IES LM-80-08 lumen maintenance measurement standard.The experimental data gathered all along the 4000 h of stress accumulated up to now suggest the presence ofmultiple degradation mechanisms that may limit the useful lifespan of the light-emitting diodes under test. Inparticular we observed the following phenomena: i) a decay of the luminous flux; ii) an increase in the reverseand forward leakage current; iii) theworsening of the chromatic properties of the emitted light; and iv) the pres-ence of a thermally activated degradation mechanism. The results provide a first insight into the reliability ofthose widely used LEDs; the results on the temperature-dependence of the degradation kinetics can be used asa guideline for the thermal design of modern distributed-light lamps.
Over the last years, Solid State Lighting (SSL) gained a considerablemarket share over conventional light sources like halogen, incandescentor Cold-Cathode Fluorescent (CCF) lamps. Thanks to the superiorlifetime, efficiency and color rendering capabilities of GaN-basedwhite light-emitting diodes (LEDs), this lighting approach is set tobe the reference standard for the nearest future.
A modern LED lamp designed for general lighting purposes usuallyfeatures a small number of High-Brightness devices (HB-LED) locatedin an appropriate frame, whose structure and material compositionare the result of rigorous thermal and optical analyses finalized tomax-imize both the heat and light extraction from the source devices. Typi-cally, these are state-of-the-art LEDs, supported by a solid backgroundof reliability studies and equippedwith a thermally advanced (ceramic)package. In the last few years, however, the need for cost reduction andfor a simpler design of distributed light systems pushed towards the im-plementation of the so-called mid-power LEDs in place of the less cost-effective HB devices.
Thosemid-power LEDs,whose (electrical) power rating ranges from0.2 to 0.5W, generally offer a good trade-off between lumen output andcost. This comes at the price of a reduced thermal and optical design ofthe package, which usually features a plastic housing that lacks of anoptical element and, in some cases, of a dedicated thermal pad. Thus,despite the lower power dissipation, the increased sensitivity of this
eneghini).
ng-term degradationmechanicrorel.2015.06.098
class of LEDs towards environmental temperature may turn out to bea major reliability issue.
The aim of this paper is to analyze the degradation mechanisms thatlimit the reliability of mid-power LEDs subjected to a high-temperatureoperating regime. To this aim, we arranged an accelerated aging experi-ment which revealed the following phenomena: 1) a strong and general-ized decay of the luminous flux for junction temperatures above 120 °C;2) an increase in the reverse and forward leakage current; 3) theworsen-ing of the chromatic properties of the emitted light; and4) the presence ofa thermally activated degradation mechanism.
The experimental results reported below and the proposed inter-pretation represent the first attempt to address the reliability of thisrapidly emerging class of white LEDs; this topic is not covered byprevious literature reports.
2. Experimental details
The mid-power LEDs analyzed within this paper have been se-lected based on the following criteria: a color temperature of 3000 K, anominal current of 100 ± 20 mA and a Color Rendering Index(CRI) N 80. Four families of LEDs (which from now on will be re-ferred to as groups A, B, C and D) fabricated by four different leadingmanufacturers were chosen. Three of the four groups (namely A,B andC) were selected within the highly popular 5630 (5.6 × 3.0 mm) class ofmid-power LEDs, while the fourth one, group D, presents a smaller-sized package (2.0 × 0.8mm) inherited from the typical devices designedfor backlighting purposes.
isms of mid-power LEDs for lighting applications, Microelectronics Re-
Fig. 1. Results of the thermal characterization performed on the four families of mid-power LEDs under analysis. On the left side: junction-temperature vs. thermal dissipationcurves at room temperature (25 °C) obtained through the diode's “forward voltagemethod” for one sample of each. On the right side: experimentally acquired values ofthe junction-temperature reached by the devices during the stress tests (maximum Tj isreported in Table 1).
Fig. 2. Luminous flux decay registered during the active stress at the highest temperature(Tamb = 105 °C) of all the four families of devices.
2 M. Buffolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx
Four single devices of each group were connected in series on aMCPCB (Metal-Core PCB) which was tightly anchored to an aluminumheatsink; a thin graphite layer served as thermal interface material.The peculiar series connection was adopted to provide a good signal-to-noise ratio during L–I measurements performed with our opticalcharacterization setup, which will be described later on. The electro-optical characterizationswere carried out on thewhole series, thereforeonly one effective sample per group and per stress condition wasanalyzed.
3. Thermal characterization and stress conditions
In order to define the thermal behavior of the four different LEDs'families under analysis,we carried out a complete thermal characteriza-tion on untreated samples. The junction-temperatures were estimatedfollowing the diode “forward voltage method” proposed by Xi andSchubert [1]. The results, summarized in Fig. 1, show the followingfacts: a) a quite good thermal resistance for the 5630 package-basedsamples (specific values range from about 9 to 20 °C/W); b) a thermaldisadvantage of D group's LEDs, due to the absence of a dedicated thermalpad and the reduced thermal design of the device; c) an almost linear re-lationship between environmental stress temperature and junction-temperature, whose maximum value of 152 °C is reached from group Dsamples stressed at Tamb = 105 °C. For this stress temperature, the fourfamilies of devices were operating outside the SOA (Safe-OperatingArea) for Tj advised from the manufacturers (see Table 1).
Together with electrical over-tress (EOS) events [2], the elevatedtemperatures reached by modern GaN-based light-emitting diodesduring operation are still the major limiting factor for their long-term reliability [3], both at chip or at package level. With the aimof testing the robustness of the mid-power LEDs under analysis
Table 1Main properties of the four different families of devices under analysis. Tj Max is themaximum junction-temperature reached during stress by a specific group. The lastcolumn reports the maximum Tj suggested on the datasheets.
Please cite this article as: M. Buffolo, et al., Long-term degradationmechanliability (2015), http://dx.doi.org/10.1016/j.microrel.2015.06.098
against thermally activated and/or current driven degradationmechanisms, the devices were actively stressed at nominal currentinside climatic chambers whose temperatures were set at 45, 65,85 and 105 °C. The stress cycles were regularly interrupted inorder to perform a complete electrical and optical characterization.For the latter we exploited a fully power and wavelength-calibrated setup constituted of a 65″ large integrating sphere(model LMS-650 from LabSphere) fiber-coupled to an Ocean OpticsUSB2000+ spectrometer. A Keithley 2614B source-meter providedthe polarization for the samples. On the other hand, the voltage–current characteristics were measured through an HP4155A param-eter analyzer, whose accuracy, even if affected by the non-ideal con-nection system used, let us spot current variations down to the pArange (see Fig. 4).
The measurement and stress methodologies described above fulfillthe common IES LM-80-08 lumenmaintenancemeasurement standard.
4. Results of the accelerated aging experiment
The degradation kinetics of the analyzed LEDs are reported in Fig. 2.Linear lifetime extrapolation for a 10% decrease of the output flux pro-duced expected values of useful device life between 2500 and 5000 h:considering that a common market requirement for a TTF70% (usuallyreferred to as L70) stands around 40,000 h, we can state that our
Fig. 3. Luminous flux trend for different stress temperatures relative to the smaller-sizedLEDs belonging to group D.
isms of mid-power LEDs for lighting applications, Microelectronics Re-
Fig. 4. Degradation of the electrical characteristic of the A group's devices stressed at thehighest ambient temperature (105 °C). On the left, in a semi-log scale, an increase in theforward leakage current is clearly visible, while its temperature dependent kinetics(measured at 4 V) are reported in the graph on the right hand. The measurements werecarried out on a series of four identical samples.
Fig. 6. Normalized luminous flux plotted against normalized forward leakage current(measured at 4 V) for the samples of group A stressed at Tamb = 105 °C. The solid curverepresents the linear fit of the experimental data.
3M. Buffolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx
samples, probably, are not suited to be operated at junction tempera-tures above 120 °C, which, as already pointed out, is the averageupper bound of the SOA for the LEDs under test. The aforementionedlifetime values are intended to be used only for the evaluation of activa-tion energies, not for an accurate estimation of the expected lifetime ofthe device: not only thiswould requiremore data points, i.e.more hoursof stress, but also a more sophisticated extrapolation algorithm, like thewidely adopted IESNA TM-21.
Aswe can see from Fig. 3, the kinetics of flux decay are highly temper-ature dependent, which suggest the presence of a temperature activateddegradationprocess. Nonetheless, the luminousflux trendsprovide a use-ful guideline for an appropriate design of the lamp thermal management,which should guarantee a maximum junction-temperature below theextrapolated critical level. In order to identify the physical mechanismsresponsible for the flux decay, a deep investigation about the degradationof the chromatic and electrical characteristics was carried out. Forthe LEDs of group A, the increase in forward leakage current shows anapproximately linear correlation with the flux decay, as reported insemi-log scale in Fig. 6.
Since one of the main electronic transport processes in the low for-ward bias region is trap-assisted tunneling [4], and since the rate ofnon-radiative recombination is directly proportional to the concentra-tion of defect-related deep-levels inside the energy gap, the experimen-tal data suggest the presence of an ongoing process of crystal defectgeneration inside the active region of the device. The observed phenom-enon showed a remarkable temperature dependence.
Fig. 5. On the left: current–voltage characteristics of group D devices stressed at 105 °C(ambient temperature). On the right: kinetics of the leakage current, measured at −20 V(−5 V per LED), of the samples stressed at all the four environmental conditions.
Please cite this article as: M. Buffolo, et al., Long-term degradationmechanliability (2015), http://dx.doi.org/10.1016/j.microrel.2015.06.098
Another relevant variation of the current–voltage characteristic wasexhibited by D group LEDs (Fig. 5). The reverse leakage current of thosedevices experienced a temperature-dependent increase over time.Since leakage current strongly depends on the density of defects, a grad-ual worsening of crystal quality is to be expected [5,6].
Moreover, in the forward conduction region, those LEDs registered amaximum increase of the forward voltage of about 63 mV, well abovethe average 30 mV gained from the other samples under analysis. Theforward voltage increase was found to be temperature dependent andwell correlated with the gradual rise of the series resistance, as shownin Fig. 7. This suggests a partial degradation of the ohmic contacts and/or a resistivity increase of the quasi-neutral regions due to the high cur-rent density [7] and temperature of the operating regime. Such behav-ior, which may be the result of a “current-crowding” phenomenon [8],is critical when LEDs are employed in modules, where the voltage in-crease VF of a single device lowers the limited voltage headroom ofthe current driver by an amount nVF, where n is the number of devicesin series.
Group D LEDs suffered from a noticeable chromaticity shift, which,for instance, led to a maximum 55 K increase of the CCT point withinthe whole 4000 h period of aging. The chromatic degradation processtakes place mostly during the first stress hours, after which the yellowto blue emission peak ratio suddenly slows its decay. This phenomenonis clearly visible in Fig. 8. A similar behavior, but with different time
Fig. 7. Relative RON and voltage increase for D group samples aged at 105 °C.
isms of mid-power LEDs for lighting applications, Microelectronics Re-
Fig. 8. Relative emission intensities registered for the blue and yellow spectral peaks(see Fig. 9) of D group devices stressed at Tamb = 105 °C. PSD (Power Spectra Density)trend at 530 nm and yellow to blue ratio are included as well.
Fig. 10. Relative yellow to blue peak emission ratio for the D group samples stressed at allthe four environmental conditions.
4 M. Buffolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx
constants, was experienced from the samples stressed at lower temper-atures (Fig. 10): the strict correlation between stress temperature andthe kinetics of the observed chromatic shift suggests the presence of athermally activated degradation process, which probably involves theplastic package [9], the yellow phosphors [10] or the encapsulatingma-terial [11].
The simultaneous presence of a high-temperature environment(N100 °C) and of a low-wavelength optical radiation is a well-knowncondition that triggers both the decay of the phosphor conversionefficiency and the worsening of the optical properties of the polymericencapsulants. As a matter of fact, those materials tend to assume anamorphous phase above their “glass-transition” temperature (TG),which not only promotes the discoloring by VOCs (Volatile OrganicCompounds) but also reduces their optical transmissivity in the low-wavelength range. Since our samples did not seem to suffer from this“red shift”, and since the power-normalized spectra of untreated sam-ples showed a remarkably different phosphor composition for D groupLEDs, the “blue shift” of the chromaticity point, which affected in a rel-evant way only the aforementioned devices, is probably due to a con-version efficiency loss of the phosphor species.
One last package-related degradation mechanism, which was notconsidered so far, is the tarnishing of the Ag reflective coating [12].This metallic layer, covering the lead frame and/or the plastic housing,
Fig. 9. Spectral measurements of D group devices stressed at Tamb = 105 °C. Inset graphsprovide amore accurate vision of the specific degradation of the phosphor's (yellowpeak)and GaN chip's (blue peak) emissions during the stress.
Please cite this article as: M. Buffolo, et al., Long-term degradationmechanliability (2015), http://dx.doi.org/10.1016/j.microrel.2015.06.098
is usually employed in mid-power LEDs in order to increase the overalllight extraction efficiency. However, sulfur and chlorine-rich pollutants,released from different sources, may diffuse through the siliconeencapsulant and react with the Ag atoms, forming Ag2S and AgClcomplexes. This may lead to an unrecoverable chromaticity shift and/ora decay of the emitted optical power. In order to evaluate this hypothesis,we de-processed the stressed samples with a depolymerizer agent andcompared them with untreated samples. Visual inspection did notreveal any particular sign of sulfurization, though in-depth analyses, likeEDS, may provide us more accurate results.
Inspection of the de-processed devices belonging to the other threegroups did not reveal proofs of degradation either, except for thesamples of group A. As we can see in Fig. 11, those two-dice basedLEDs showed a little darkening of the chip area near the lower sur-face contact. While this may be an evidence of an ongoing currentcrowding process, no correlation with the (slight) forward-voltageincrease was found.
As additional analysis, we tried to estimate the activation energies ofthe registered lumen decay processes. From the L90 lifetime data, weextrapolated energy values of the degradation process ranging from0.38 to 0.6 eV. The Arrhenius plots, reported in Fig. 12 for group Cdevices, did not show any bending for high temperature values,which suggests that no new processes were activated for junction-temperatures above 120 °C.
Fig. 11. De-processed samples of A group: untreated sample (left) and device stressed at105 °C for 2500 h (right). Left and right chips are respectively shown.
isms of mid-power LEDs for lighting applications, Microelectronics Re-
Fig. 12. Arrhenius plot of the 90% TTF (Time To Failure) for C group's LEDs. The failurecriterion, which corresponds to a 10% decrease of the luminous flux at nominal current,has been chosen in order to guarantee the best possible coherence between the (linearly)extrapolated lifetime and the experimental data acquired so far, whose linear fit is repre-sented by the red line showed above.
5M. Buffolo et al. / Microelectronics Reliability xxx (2015) xxx–xxx
5. Conclusions
In summary,within this paper, we have presented thefirst extensivestudy of the degradation mechanisms of mid-power white LEDs. Theresults allowed us to identify the following degradation processes:generation of defects within the active region of the devices, worseningin the optical properties of the package/phosphor system, and the deg-radation of the electrical characteristics of the devices. The degradationprocess was found to be thermally activated with activation energies inthe range from 0.38 to 0.6 eV.
Please cite this article as: M. Buffolo, et al., Long-term degradationmechanliability (2015), http://dx.doi.org/10.1016/j.microrel.2015.06.098
References
[1] Y. Xi, E.F. Schubert, Junction-temperature measurement in GaN ultraviolet light-emitting diodes using diode forward voltage method, Appl. Phys. Lett. 85 (2004)2163–2165.
[2] M. Dal Lago, M. Meneghini, N. Trivellin, G. Mura, M. Vanzi, G. Meneghesso, E. Zanoni,Hot-plugging of LED modules: Electrical characterization and device degradation,Microelectron. Reliab. 53 (9–11) (September–November 2013).
[3] M. Meneghini, M. Dal Lago, N. Trivellin, G. Meneghesso, E. Zanoni, DegradationMechanisms of High-Power LEDs for Lighting Applications: An Overview, IEEETrans. Ind. Appl. 50 (1) (Jan.–Feb. 2014) 78–85.
[4] X.A. Cao, E.B. Stokes, Diffusion and tunneling currents in GaN/InGaN multiple quan-tum well light-emitting diodes, IEEE Electron Device Lett. 23 (9) (Sept. 2002)535–537.
[5] J. Cho, A. Mao, J.K. Kim, J.K. Son, Y. Park, E.F. Schubert, Analysis of reverse tunnellingcurrent in GaInN light-emitting diodes, Electron. Lett. 46 (2) (January 21 2010)156–158.
[6] F. Rossi, M. Pavesi, M. Meneghini, G. Salviati, M. Manfredi, G. Meneghesso, A.Castaldini, A. Cavallini, L.R.U. Strauss, U. Zehnder, E. Zanoni, Influence of shortterm low current dc aging on the electrical and optical properties of AlGaN blueLight-Emitting Diodes, J. Appl. Phys. 99 (2006) (053 104-1-053 104-7).
[7] B. Hamon, B. Bataillou, l. Mendizabal, A. Gasse, G. Feuillet, N-contacts degradationanalysis of white flip chip LEDs during reliability tests, IEEE Int. Reliab. Phys.Symp. (2014) FA.1.1–FA.1.6 http://ieeexplore.ieee.org/xpl/articleDetails.jsp?re-load=true&arnumber=6861140.
[8] M. Meneghini, A. Tazzoli, G. Mura, G. Meneghesso, E. Zanoni, A Review on the Phys-ical Mechanisms That Limit the Reliability of GaN-Based LEDs, IEEE Trans. ElectronDevices 57 (1) (Jan. 2010) 108–118.
[9] E. Jung, M.S. Kim, H. Kim, Analysis of Contributing Factors for Determining the Reli-ability Characteristics of GaN-Based White Light-Emitting Diodes With Dual Degra-dation Kinetics, IEEE Trans. Electron Devices vol. 60 (1) (Jan. 2013) 186–191.
[10] M. Meneghini, M. Dal Lago, N. Trivellin, G. Meneghesso, E. Zanoni, Thermally Acti-vated Degradation of Remote Phosphors for Application in LED Lighting, IEEETrans. Device Mater. Reliab. 13 (1) (March 2013) 316–318.
[11] P. Appaiah, N. Narendran, Indika U. Perera, Yiting Zhu, Yi-wei Liu, Effect of thermalstress and short-wavelength visible radiation onphosphor-embedded LEDencapsulantdegradation, Opt. Mater. 46 (April 2015) 6–11.
[12] G. Mura, G. Cassanelli, F. Fantini, M. Vanzi, Sulfur-contamination of high powerwhite LED, Microelectron. Reliab. 48 (8–9) (August–September 2008) 1208–1211.
isms of mid-power LEDs for lighting applications, Microelectronics Re-
