Physical mechanism responsible for the stretched exponential decay behavior of aging organic light-emitting diodes
C.Féry,B.Racine,
D.Vaufrey,H.Doyeux,and S.Cinà
Thomson R&D France,1Avenue Belle Fontaine,CS17616,35576Cesson Sévigné,France
͑Received24June2005;accepted21September2005;published online15November2005͒
The main process responsible for the luminance degradation in organic light-emitting diodes
͑OLEDs͒driven under constant current has not yet been identified.In this paper,we propo an
approach to describe the intrinsic mechanisms involved in the OLED aging.Wefirst show that a
stretched exponential decay can be ud tofit almost all the luminance versus time curves obtained
under different driving conditions.In this way,we are able to prove that they can all be described
by employing a single free parameter model.By using an approach bad on local relaxation events,
we will demonstrate that a single mechanism is responsible for the dominant aging process.赤楠盆景怎么养
Furthermore,we will demonstrate that the main relaxation event is the annihilation of one emissive
center.We then u our model tofit all the experimental data measured under different driving
condition,and show that by carefullyfitting the accelerated luminance lifetime-curves,we can
extrapolate the low-luminance lifetime needed for real display applications,with a high degree of
accuracy.©2005American Institute of Physics.͓DOI:10.1063/1.2133922͔
OLED degradation is one of the main issues the industry
has to face in order to make this technology sufficiently re-
liable for mass production.Particularly,differential degrada-
tion between the three primary colors,sticking image effects,
as well as degradation under harsh storage conditions,need
to be more deeply understood.Despite the importance of the
issue,a limited number of studies have been published so
far,1–4and,as a conquence,the main mechanisms deter-
mining the degradation are still not well understood.
One problem still to be solved is also how tofit the
luminance-time͑L-t͒curve in order to estimate the half-life ͑defined as the time needed for the luminance to reach50% of its initial value,when driven at constant current͒at uful
luminance value.
Thanks to improvements achieved in both device struc-
ture and materials quality,the lifetimes͑LT͒at luminance
values needed for real display applications are today of the
order of100000h.In order to estimate the values in a
日本北海道reasonable amount of time,accelerate testing conditions,
both using higher luminance or higher temperature,are nor-
mally done.However,accelerations might introduce other
aging mechanisms,making the estimation of the LT a com-
plex issue.
Thefit of the L-t curves over a limited range of time,
also appears to be a critical problem.
A widespread approach is to describe the L-t curve by
using a combination of exponential decays,commonly using同学聚会讲话
two terms,thefirst-one accounting for the rapid initial decay,
the cond-one for the long-term degradation,as shown in
Eq.͑1͒,
L
L0
=ae−␣t+be−t,͑1͒
where L0is the initial luminance,and a,b,␣,,arefitting parameters.
However,this is extremely dependent on how and when thefit is done,as can be shown in Fig.1,where thefit with Eq.͑1͒has been done after approximately1000h and 2000h LT,extrapolating,respectively,3600and4400h half-life.
A far better way is to u a stretched exponential decay ͑SED͒,as previously reported,5,6and shown in Eq.͑2͒.How-ever,a physical justification for using the SED has so far never been found,
L
L0
=expͫ−ͩt
ͪͬ.͑2͒
In the following ctions we will prove that the SED can be successfully ud tofit all the L-t curves measured at differ-ent initial luminance.We will also show that the coefficient is constant when changing the initial luminance,reducing thefit to a single parameter͑͒.This will strongly improve the predictive nature of thefit,when ud to extrapolate lower and more realistic initial luminance curves.More im-portantly,we will explain the origin of the SE behavior, when ud to describe aging process with a constant input, as tho obrved when degrading an OLED at constant
current.
FIG.1.Typical L-t curve at L0=800cd/m2,showing the experimental data ͑full circles͒,together with thefit using Eq.͑1͒after1100h͑dotted line͒, and the2200h͑full line͒.
APPLIED PHYSICS LETTERS87,213502͑2005͒
0003-6951/2005/87͑21͒/213502/3/$22.50©2005American Institute of Physics
87,213502-1
体育节活动方案
Downloaded 19 Sep 2011 to 202.38.193.36. Redistribution subject to AIP licen or copyright; e apl.aip/about/rights_and_permissions
We ud a multilayer,green-phosphorescent structure,employing an n -doped electron transport layer ͑ETL ͒and a p -doped hole transport layer ͑HTL ͒,respectively,for elec-tron and hole injection.The OLED is deposited on glass with the following structure:ITO ͑150nm ͒/HTL ͑p -doped ͒80nm/EBL 7nm/EML ͑green doped ͒30nm/HBL 7nm/ETL ͑n -doped ͒20nm/Al 200nm.Where EBL is the electron blocking layer,EML is the emissive layer,and HBL is the hole blocking layer.
The samples were fabricated using a standard vacuum deposition process.The ITO substrates were cleaned by a wet process followed by O 2plasma treatment just before the HTL deposition.The devices were then encapsulated by cover glass and getters,in inert atmosphere.
Each sample contains six identical pixels.Four of them have been LT tested,while two pixels have been kept un-touched as a reference.Attention has been paid that no black spots had been developed during the OLED aging.
Figure 2shows the experimental L -t data at different initial luminance values,together with the fits using Eq.͑2͒at constant =0.53,determined by fitting the L -t at L 0=800cd/m 2.We can e that the SED with fixed can be ud to fit all the L -t curves with a high degree of accuracy.
The estimated LTs are shown in Fig.3,together with the fit using the well known relation,
L 0n t 1/2=const,
͑3͒
where n is the acceleration coefficient,and t 1/2is the half-life.The best fit has been obtained with n =1.7.A LT value in
excess of 100000h is estimated at L 0=100cd/m 2.Although
not reported here,the same LT study has been performed for OLEDs with various architectures.We found that both and n depends on the materials t as well as on the device architecture.
We obrve that the residual PL follows the same trend as the residual EL,showing that the annihilation of the emis-sive centers ͑ECs ͒is likely to be one of the main reasons for the OLED degradation.The smaller relative decrea in PL compared with the EL ͑ϳ25%compared with ϳ50%͒can be attributed to the reason that both excitons formation and re-combination,only takes place on part of the EML.
To confirm this point,we have done the same experi-ment on a sample having a much thinner EML ͑ϳ16nm ͒.The results,shown in the inrt of Fig.4,clearly show that the relative change in PL and EL gets clor when the EL gets ,when the light emission takes place on a larger portion of the EML.
From what we have obrved,we can conclude that the degradation measured in EL is dominated by the annihilation of the ECs inside the EML.
In order to model the OLED aging,we consider other degradation process obrved in different systems,like degradation of SiO 2under constant pressure.It has been demonstrated that,in the systems,the aging rate decreas when the total aging increas.7
We formulate the hypothesis that,the probability to an-nihilate an isolated EC is proportional to exp−͑U 0/kT ͒,where U 0is the energy needed to degrade an isolated EC.
This probability is incread due to the prence of all the other ECs,as shown in Eq.͑4͒,
U ͑n ͒=U 0−͑n s −n ͒,
͑4͒
where n s and n are,respectively,the total number of ECs at t =0and the number of degraded EC at a time t and is a coupling constant.
We assume that once an EC is degraded,it does not play any role within the system,and that the probability of de-grading an EC decreas while aging the OLED.This is becau the number of EC that could contribute to the deg-radation of the others ECs,through process like radiative or nonradiative recombination,decrea while the OLED ages.Therefore,the rate equation for the numb
er of degraded EC,n ,can be written
as
FIG.2.L -t curves measured at various initial luminance values.Black lines are the fits with Eq.͑2͒,using a constant value for =0.53and using as the only fit
parameter.
FIG. 3.Estimated half-life using the SED with constant =0.53͑full squares ͒,together with the fit using Eq.͑3͒͑line ͒
.
FIG.4.Residual PL and EL measured on both the degraded and stored ͑L 0=0͒pixels.The inrt shows similar experimental data measured on a different sample with thinner EML.
前n项求和公式方法
Downloaded 19 Sep 2011 to 202.38.193.36. Redistribution subject to AIP licen or copyright; e apl.aip/about/rights_and_permissions
dn dt =
1
0exp͑−U͑n͒/kT͒−Cn,͑5͒
Cn is ud to accounts for ,dn/dt=0when t →ϱ.The constant1/0reprents the EC relaxation rate.
酸辣菜By combining Eqs.͑4͒and͑5͒,we obtain the differential Eq.͑6͒,who solution gives the number of damaged EC at time t as function ofand n s,
1d
小学教师实习周记dtͩn n S
ͪ=expͩ−A n n Sͪ−n n S exp͑−A͒,͑6͒
where
A=n S2 and
=0
n S
expͩU0−A kT
ͪ.
We define L͑t͒/L0=͑1−n/n s͒.This quantity should rep-rent the luminance degradation of an OLED,as tho shown in Fig.2.
In Fig.5wefind the same experimental data as in Fig.2, while thefits are now done by using L͑t͒/L0obtained by numerically solving Eq.͑6͒.
It has been demonstrated7that the numerical solution of the Eq.͑6͒family,can be very wellfitted by using the SED.
This explains why Eq.͑6͒and the SED give very similar results when ud on the same experimental data.It also shows,for thefirst time,the physical meaning behind the u of the SED tofit the luminance degradation of an OLED under constant driving conditions.
In summary,we have shown that the SED can be ud to accuratelyfit the OLED LT.We have also found out that the determination of low luminance LT starting from accelerated measurements,can be reduced to a single free parameter problem,greatly improving the accuracy of the half-life prediction.
Furthermore,we have shown for thefirst time that the annihilation of the ECs is the main mechanism responsible for the OLED degradation.This single mechanism is suffi-cient to account for both the initial rapid decay,as well as for the long term degradation obrved in the L-t curve.
We have propod a model that describes with a high level of accuracy the degradation process,and justifies the u of the SED tofit the LT experimental curves.
妙觉寺1I.D.Parker,Y.Cao,and C.Y.Yang,J.Appl.Phys.85,2441͑1999͒.
2E.Burrows,S.R.Forrest,T.X.Zhou,and L.Michalski,Appl.Phys.Lett. 76,2493͑2000͒.
3G.C.M.Silvestre,Appl.Phys.Lett.78,1619͑2001͒.
4H.Aziz and Z.D.Popovic,Appl.Phys.Lett.80,2180͑2002͒.
5W.E.Howard and O.F.Prache,IBM J.Res.Dev.45,115͑2001͒.
6M.Ishii,R&D Review of Toyota CRDL38.
7K.Trachenko and M.T.Dove,Phys.Rev.B70,132202͑2004͒
.
FIG.5.Experimental data as in Fig.2͑full dots͒.The continuous lines are fits with L/L0obtained by numerically solving Eq.͑6͒.
Downloaded 19 Sep 2011 to 202.38.193.36. Redistribution subject to AIP licen or copyright; e apl.aip/about/rights_and_permissions
