Você está na página 1de 6

Journal of Information Display

ISSN: 1598-0316 (Print) 2158-1606 (Online) Journal homepage: https://www.tandfonline.com/loi/tjid20

Constant-stress accelerated life test of white


organic light-emitting diode based on least square
method under Weibull distribution

Jianping Zhang, Chao Liu, Guoliang Cheng, Xiao Chen, Jionglei Wu, Qunzhi
Zhu & Laichang Zhang

To cite this article: Jianping Zhang, Chao Liu, Guoliang Cheng, Xiao Chen, Jionglei Wu, Qunzhi
Zhu & Laichang Zhang (2014) Constant-stress accelerated life test of white organic light-emitting
diode based on least square method under Weibull distribution, Journal of Information Display,
15:2, 71-75, DOI: 10.1080/15980316.2014.889613

To link to this article: https://doi.org/10.1080/15980316.2014.889613

© 2014 The Korean Information Display


Society

Published online: 26 Mar 2014.

Submit your article to this journal

Article views: 210

View Crossmark data

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tjid20
Journal of Information Display, 2014
Vol. 15, No. 2, 71–75, http://dx.doi.org/10.1080/15980316.2014.889613

Constant-stress accelerated life test of white organic light-emitting diode based on least square
method under Weibull distribution
Jianping Zhanga∗ , Chao Liub , Guoliang Chengc , Xiao Chenc , Jionglei Wua , Qunzhi Zhua and Laichang Zhangd
a College of Energy and Mechanical Engineering, Shanghai University of Electric Power, Shanghai 200090, People’s Republic of China;
b Department of Automation Science and Technology, Xi’an Jiaotong University, Xi’an 710049, People’s Republic of China; c Shanghai
Tianyi Electric Co., Ltd., Shanghai 201611, People’s Republic of China; d School of Engineering, Edith Cowan University, 270
Joondalup Drive, Joondalup, Perth, WA 6027, Australia
(Received 9 November 2013; accepted 21 January 2014 )

It is currently hard to estimate the reliability parameters of organic light-emitting diodes (OLEDs) when conducting a life test
at normal stress, due to the remarkably improved life of OLEDs to thousands hours. This work adopted three constant-stress
accelerated life tests (CSALTs) to predict the life of white OLEDs in a short time. The Weibull function was applied to describe
the life distribution, and the shape and scale parameters were estimated using the least square method. The experimental test
data were statistically analyzed using a self-developed software. The life of white OLEDs predicted via this software is in
good agreement with that reported from the customers. The numerical results indicated that the assumptions of CSALT are
correct, and that CSALT can be used to predict the life of white OLEDs. This work confirmed that the life of white OLEDs
meets the Weibull distribution, and that the accelerated life equation conforms to the inverse power law. Furthermore, the
precise accelerated parameters were shown to be particularly useful in enabling the rapid estimation of white OLEDs’ life.
Keywords: white OLED; accelerated life test; LSM; Weibull distribution; constant stress

1. Introduction in a dual-plate OLED display structure protected by a


In recent years, the flat panel display (FPD) technology SiNx passivation film. Tests demonstrated that this struc-
made great progress with the emergence of the colored ture significantly improved the life of the white OLED. By
mobile phone, flat panel television, and computer. FPDs comparing the efficiency, heat resistivity, and reliability of
include a liquid crystal display, a plasma display panel, OLEDs with indium-tin oxide (ITO) and indium-zinc oxide
light-emitting diodes, and organic light-emitting diodes (IZO) as anode, Pinato et al. [6] considered the material
(OLEDs), among others. Unlike the other FPD technolo- selection of the OLED anode and carried out an accelerated
gies, the OLED is an initiative lighting device that does life test (ALT). The results showed that an OLED with IZO
not need backlight or environmental light. Owing to its as the anode has a higher reliability. Shi et al. [7] revealed the
advantages, such as self-emission, fast response time, wide factor affecting the life of the OLED lighting module in their
viewing angle, good low-temperature property, and high research on the relationship between the temperature distri-
luminance, the OLED display technology is becoming a bution and the photoelectric degradation characteristics.
hotspot in the current display field, attracting considerable With the innovation of the material design field, the
research attention [1,2]. reliability and life of OLEDs have been improved remark-
Life is one of the most important properties of OLEDs, ably. The life of most OLED display devices in massive
directly influencing its business prospect [3]. In the past commercial application has surpassed 10,000 h. For such
decade, the life and reliability estimation methods for a long life, it is hard to estimate the reliability param-
OLEDs were significantly investigated by researchers eters of OLEDs when carrying out a life test at normal
worldwide. Jankovic et al. [4] proposed a simple active- stress, because the life test may last for a very long time
matrix OLED pixel with conventional voltage driving and but only few samples fail [8]. The main life test methods
ability to compensate for the Vth shift of thin-film transis- include the accelerated degradation test [9–11] and the ALT.
tors (TFTs), and the stable pixel operation was achieved by ALT is a more commonly used method to reduce the test
employing a dual-gate TFT for OLED driving. Chang et al. time by increasing the test stress levels while keeping the
[5] put forward a white OLED with a color filter embedded failure mechanism unchanged. ALT can be classified into

∗ Corresponding author. Email: jpzhanglzu@163.com

© 2014 The Korean Information Display Society


72 J. Zhang et al.

several categories [12]: constant-stress accelerated life test 2.3. Failure criterion and test end time
(CSALT), step stress accelerated life test, and progressive The life of the OLED device can be defined as the amount
stress accelerated life test. Among these three categories of time until the OLED luminance decreases to below 50%
of ALTs, CSALT is the most accurate and is also a bet- of the initial luminance. Based on this criterion, the failure
ter option if the amounts of the test samples and stresses time of each test sample was recorded. When all the samples
are quite small. Liu et al. [13] presented a design approach at the three stress levels failed, the tests were terminated.
for the sequential CSALT with an auxiliary acceleration
factor, which can further amplify the failure probability
of highly reliable test items at low stress levels while 3. OLED CSALT theory model
maintaining an acceptable degree of extrapolation for reli- 3.1. Basic assumptions
ability inference. Liu et al. [14] selected the accelerated Assumption 1 For each stress level Ii (i = 0, 1, 2, 3), the
mode and accelerated stress level in pneumatic solenoid white OLED life follows the Weibull distribution, and the
valve ALT by analyzing the sensitive stress in three kinds distribution function can be expressed as
of failure mechanisms (wear, fatigue, and aging), carried
  m i 
out ALT, and concluded that the mechanical performance t
of the solenoid valve is more stable than its electrical Fi (t) = 1 − exp − , t > 0, (1)
ηi
performance.
In this study, CSALT was performed to obtain the where mi is the shape parameter, which affects the geomet-
failure time and cumulative failure probabilities of white ric shape of the Weibull probability density curve, and ηi
OLED samples, and the least square method (LSM) was is the scale parameter that directly determines the curve’s
used to fit the curves at each stress level. Afterwards, the smoothness.
Kolmogorov–Smirnov tests were employed to determine if
the OLED life follows the Weibull distribution. The OLED Assumption 2 The OLED failure mechanism remains
curve of the life characteristic pattern can be obtained via unchanged while the stress ranges from I0 to I3 . As the
LSM, and it was confirmed that the OLED life characteris- mechanism of the white OLED is reflected by the shape
tic pattern meets the inverse power law. In addition, white parameter m of the Weibull distribution, the shape parame-
OLED life information can be acquired. ter at each stress level Ii should be unchanged, which means

m0 = m1 = m2 = m3 . (2)

2. OLED CSALT plan Assumption 3 The inverse power law can be used to
2.1. Selection of accelerated stress estimate the working life characteristics of electrical com-
OLED is a current-driving device, and current is the most ponents. The white OLED accelerated model satisfies the
important factor influencing the OLED life [8]. Therefore, inverse power law, namely that the relationship between
current was selected as the accelerated stress of ALT in scale parameter η and current stress level I is written as
this work.
ln η = α + β ln I , (3)

where α and β are the accelerated parameters.


2.2. Levels and numbers of accelerated stress
The levels and numbers of accelerated stress are two major
factors that ensure that the test results scientifically reflect 3.2. Estimation of the Weibull parameters
the OLED life. The difference between the highest and low- Equation (1) can be transformed as
est stresses should be enlarged under the condition that the
test precision and efficiency will be guaranteed. To prevent ln ln[1 − Fi (t)]−1 = mi ln t − mi ln ηi , (4)
the appearance of a new failure mechanism, the highest
if
stress should be made lower than the ultimate level deter-
mined by the material structure and the techniques of the y = ln ln[1 − Fi (t)]−1 , x = ln t,
product. (5)
a = mi , b = −mi ln ηi .
According to the principle of reliability, the minimum
number of accelerated stress in the life test should be no less Equation (4) can be simplified as the linear relationship form
than three. The M00071 white OLED mixed with red, green,
and blue colors were used in this work. The following three
y = ax + b. (6)
stress levels were selected: I1 = 9.64 mA, I2 = 17.09 mA,
and I3 = 22.58 mA. In addition, the normal working current After the sequencing of failure time tk,i at Ii (i = 1, 2, 3), the
of the test samples was I0 = 3.20 mA. corresponding cumulative failure probability F(tk,i ) can be
J. Inf. Disp. 73

Table 1. Failure time of the tested samples under constant stress at different currents.

Failure time/h
Current stress/mA t1 t2 t3 t4 t5 t6 t7 t8 t9 t10

I1 = 9.64 1691.50 2084.67 2100.32 2374.50 2421.50 2586.00 2621.50 2680.50 2868.00 2879.50
I2 = 17.09 601.50 689.67 697.33 716.50 785.50 854.50 889.50 1115.67 1131.33 1251.50
I3 = 22.58 406.00 440.50 463.50 532.50 555.50 643.67 651.33 716.50 762.50 –

calculated using the following median rank formula: According to Assumption 2 (Section 3.1), shape param-
eter m is obtained as the weighted average of mi , as follows:
k − 0.3
F(tk,i ) = , k = 1, 2, . . . , ni , (7) 3
ni + 0.4 ni mi
m = i=1 . (14)
where n1 = n2 = 10 and n3 = 9. Therefore, the test data 3
i=1 ni
can be written as

(tk, i , F(tk, i )), k = 1, 2, . . . , ni . (8) 4. Test data


OLED CSALTs were carried out according to the acceler-
According to Equations (5) and (8), one can obtain
ated test theory and plan. The failure time data of all the
(ln tk,i , ln ln[1 − F(tk,i )]−1 ) = (xk,i , yk,i ). (9) white OLED samples at Ii (i = 1, 2, 3) are listed in Table 1.

By applying LSM to Equation (9), coefficients ai and bi


(i = 1, 2, 3) can be expressed as 5. OLED life prediction
ni ni ni  Due to the calculation complexity in the aspect of life esti-
k=1 xk,i yk,i − k=1 xk,i k=1 yk,i /ni mation, a white OLED life test software developed by the
ai = n i 2 ni 2 ,
authors was used to process and analyze the test data. The
k=1 xk,i − k=1 xk,i /ni
ni ni (10) software has a series of advantages, such as simplicity,
y k,i a i x k,i speedy, accuracy, and generality.
bi = k=1 − k=1
.
ni ni

Combining Equations (5) and (10), shape parameter mi 5.1. Data processing for CSALT
and scale parameter ηi at Ii are, respectively, expressed as The constant-stress test data listed in Table 1 were processed
  by a self-developed software, and the statistics curves fit-
bi
mi = ai , ηi = exp − . (11) ted by LSM are plotted in Figure 1. Shape parameter mi ,
ai

The determination coefficients of xk,i and yk,i are writ-


ten as
n n  2
i i 
ni
xk,i yk,i − xk,i · yk,i ni
k=1 k=1 k=1
Ri =

2
n 
n  .
ni i 2
ni i 2
xk,i −
2
xk,i ni yk,i −
2
yk,i ni
k=1 k=1 k=1 k=1
(12)

3.3. Life estimation


If the OLED life conforms to the Weibull distribution, the
calculation formula of OLED average life μ and median
life t0.5 are, respectively, expressed as
   1/m
1 1
μ = η0  1 + , t0.5 = η0 ln , (13)
m 0.5

where (·) is the gamma function, and η0 is the scale


parameter estimated at I0 . Figure 1. Statistical curves of the constant-stress data.
74 J. Zhang et al.

scale parameter ηi , and determination coefficient R2i of the Table 2. mi , ηi , and R2i at each stress.
fitting curve at each current stress level can be obtained
Current
using Equations (10)–(12). These results are presented in stress (mA) I1 = 9.64 I2 = 17.09 I3 = 22.58
Table 2. It was found that the R2i calculated at each current
stress Ii (i = 1, 2, 3) approached 1, which demonstrated that mi 6.5764 4.2754 4.8350
ηi 2600.7568 959.9457 625.9725
the degree of linear fitting was high. R2i 0.9690 0.8909 0.9518

5.2. Kolmogorov–Smirnov test


According to the K–S theory [15], a greater significance The result also confirmed that the accelerated model fully
level (α = 0.2) was selected due to the limited samples. obeys the inverse power law.
The K–S test was employed to determine if the OLED life
follows the Weibull distribution at each current stress level Ii
5.4. OLED life prediction at the normal working stress
(i = 1, 2, 3). The values of the K–S test at three accelerated
stresses were calculated using a self-developed software, After substituting I0 = 3.2 mA into Equation (15), the
as follows: α1 = 0.142 < D0.2,10 = 0.323, α2 = 0.186 < OLED scale parameter at normal working stress can be
D0.2,10 = 0.323, and α3 = 0.126 < D0.2,9 = 0.339. obtained as η0 = 16511.9541 h. Combining mi (i = 1, 2, 3)
Hence, it was confirmed that the K–S tests at each stress in Table 2 with Equation (14), the OLED shape parame-
Ii (i = 1, 2, 3) passed. ter at normal working stress m0 = 5.2425 can be obtained.
Therefore, according to Equation (13), average life μ and
median life t0.5 of the white OLED can be calculated as
5.3. Accelerated life equation follows: μ = 15202.2 h and t0.5 = 15397 h.
Data points (ln Ii , ln ηi ) (i = 1, 2, 3) can be plotted and fitted The market performance demonstrates that the life of
into a linear curve by LSM, as shown in Figure 2. The accel- the white OLEDs, which were selected as the samples in
erated parameters obtained via LSM were α = −1.6834 the accelerated test, is about 16,000 h. Therefore, the white
and β = 11.6699. As such, the accelerated life equation is OLED life estimated through CSALTs in this work is very
expressed as close to the client feedback data, with a difference in struc-
ture, material, and techniques between the OLED samples
ln η = −1.6834 ln I + 11.6699. (15) and the products excluded.

Determinant coefficient R2 in Figure 2 is 0.9991, very close


to 1, which shows that the degree of linear fitting was high. 6. Conclusions
From the three CSALTs and the statistical analysis for
OLED, the following main conclusions can be drawn:

(1) The results of the test statistic analysis confirmed


that the white OLED life satisfies the Weibull dis-
tribution, and that the accelerated life model follows
the inverse power law.
(2) The self-developed software can be used to simplify
the statistical analysis of the OLED ALT results,
and to precisely predict the white OLED life.
(3) The acquired life information can provide engi-
neers with important guidelines for improving the
reliability and quality of OLED products.

Funding
This study was financially sponsored by the Shanghai Subject
Chief Scientist Program (type B) [13XD1425200]; the Founda-
tion of Shanghai’s Committee of Science and Technology, China
[11160500600], [11dz2281700], and [09DJ1400204]; the Innova-
tion Program of the Shanghai Municipal Education Commission
[14ZZ154] and [13ZZ130]; the Key Fund of the Shanghai Science
Technology Committee [13160501000]; and the National Natural
Figure 2. Curve of the OLED life characteristic pattern. Science Foundation of China [50706025] and [51201097].
J. Inf. Disp. 75

References [7] Y.F. Shi, J.X. Yang, W.Q. Zhu, S.Z. Li, K.J Ma, F.
[1] Y.Y. Liu, W.D. Geng and Y.P. Dai, J. Cent. South Univ. 19, Wang, and M.L. Zhung, Chin. Illumin. Eng. J. 23, 107
1276 (2012). (2012).
[2] J.I. Hai, H.O. Kyong, L. Inhwan, H.R. Do, M.C. Sang, N.K. [8] J.P. Zhang, F. Liu, Y. Liu, W. Helen, W.L. Wu, and A.X.
Keum, D.K. Hye, and K.K. OH, IEEE Trans. Electron. Dev. Zhou, IEEE Trans. Electron. Dev. 59, 3401 (2012).
57, 3012 (2010). [9] I.P. Jong and J.B. Suk, IEEE Trans. Reliab. 59, 74 (2010).
[3] T.H. Chang and S.Q. Peng, Vac. Cryogenics 14, 115 (2008). [10] F.K. Wang and T.P. Chu, Microelectron. Reliab. 52, 1332
[4] N.D. Jankovic and V. Brajovic, Electron. Lett. 47, 456 (2012).
(2011). [11] X. Liu and L.C. Tang, Qual. Reliab. Eng. Int. 26, 863 (2010).
[5] W.H. Chang, K.K. Hwa, S.P. Hee, H.P. Sung, J.S. Chang, [12] J.P. Zhang and R.T. Wang, J. Test. Eval. 37, 316 (2009).
S.C. Hong, and C.K. Woo, J. Disp. Technol. 5, 541 (2009). [13] X. Liu and L.C. Tang, J. Stat. Plan. Infer. 140, 1968 (2010).
[6] A. Pinato, M. Meneghine, A. Cester, N. Wrachien, [14] D.F. Liu, Z.Y. Tang, Z.C. Pei, and J. Peng, Mach. Tool
A. Tazzoli, E. Zanoni, and G. Meneghesso, IEEE Hydraulic. 38, 26 (2010).
CFP09RPS-CDR 47th Annual International Reliability [15] Y. Kang, H.J. Cai, and S.B. Song, J. Hydraul. Eng. 32, 5
Physics Symposium Montreal (2009), p. 105. (2013).

Você também pode gostar