langevin bipolar charge and current distributions in

Upload: ajayakumarkavala

Post on 02-Jun-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    1/8

    Bipolar charge and current distributions in organic light-emitting diodes

    J. C. Scott, S. Karg, and S. A. Carter

    Citation: Journal of Applied Physics 82, 1454 (1997); doi: 10.1063/1.365923

    View online: http://dx.doi.org/10.1063/1.365923

    View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/82/3?ver=pdfcov

    Published by theAIP Publishing

    Articles you may be interested in

    Space charge induced electroluminescence spectra shift in organic light-emitting diodes

    J. Appl. Phys. 112, 014513 (2012); 10.1063/1.4736589

    Measurements of current spreading length and design of GaN-based light emitting diodes

    Appl. Phys. Lett. 90, 063510 (2007); 10.1063/1.2450670

    Use of anisotropic laser etching to the top n - Ga N layer to alleviate current-crowding effect in vertical-structured

    GaN-based light-emitting diodes

    Appl. Phys. Lett. 90, 041115 (2007); 10.1063/1.2432254

    Current crowding in GaN/InGaN light emitting diodes on insulating substrates

    J. Appl. Phys. 90, 4191 (2001); 10.1063/1.1403665

    Kinetics of charge carrier recombination in organic light-emitting diodes

    Appl. Phys. Lett. 72, 513 (1998); 10.1063/1.120805

    [This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to

    14.139.194.12 On: Fri, 24 Oct 2014 14:26:52

    http://scitation.aip.org/search?value1=J.+C.+Scott&option1=authorhttp://scitation.aip.org/search?value1=S.+Karg&option1=authorhttp://scitation.aip.org/search?value1=S.+A.+Carter&option1=authorhttp://scitation.aip.org/content/aip/journal/jap?ver=pdfcovhttp://dx.doi.org/10.1063/1.365923http://scitation.aip.org/content/aip/journal/jap/82/3?ver=pdfcovhttp://scitation.aip.org/content/aip?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/112/1/10.1063/1.4736589?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/90/6/10.1063/1.2450670?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/90/4/10.1063/1.2432254?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/90/4/10.1063/1.2432254?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/90/8/10.1063/1.1403665?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/72/5/10.1063/1.120805?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/72/5/10.1063/1.120805?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/90/8/10.1063/1.1403665?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/90/4/10.1063/1.2432254?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/90/4/10.1063/1.2432254?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/apl/90/6/10.1063/1.2450670?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/112/1/10.1063/1.4736589?ver=pdfcovhttp://scitation.aip.org/content/aip?ver=pdfcovhttp://scitation.aip.org/content/aip/journal/jap/82/3?ver=pdfcovhttp://dx.doi.org/10.1063/1.365923http://scitation.aip.org/content/aip/journal/jap?ver=pdfcovhttp://scitation.aip.org/search?value1=S.+A.+Carter&option1=authorhttp://scitation.aip.org/search?value1=S.+Karg&option1=authorhttp://scitation.aip.org/search?value1=J.+C.+Scott&option1=authorhttp://oasc12039.247realmedia.com/RealMedia/ads/click_lx.ads/www.aip.org/pt/adcenter/pdfcover_test/L-37/879969015/x01/AIP-PT/JAP_ArticleDL_101514/aplmaterialsBIG_2.jpg/47344656396c504a5a37344142416b75?xhttp://scitation.aip.org/content/aip/journal/jap?ver=pdfcov
  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    2/8

    Bipolar charge and current distributions in organic light-emitting diodes

    J. C. Scott,a)

    IBM Research Division, Almaden Research Center, 650 Harry Road, San Jose, California 95120

    S. KargPhysikalisches Inst., Universitat Bayreuth, 95440 Bayreuth, Germany

    S. A. CarterDepartment of Physics, University of California, Santa Cruz, California 95064

    Received 17 January 1997; accepted for publication 22 April 1997

    The electron and hole charge distributions and the luminance profile in organic light-emitting diodes

    OLEDs depend upon the bulk properties of the emissive layer, as well as on the injection

    characteristics at the anode and cathode interfaces. We address the problem of separating the

    relative contributions of hole injection, electron injection, and recombination to the overall

    performance of single layer OLED devices. Using the approach of Parmenter and Ruppel J. Appl.

    Phys. 30, 1548 1959, and including Langevin recombination, expressions are derived for the

    currentvoltage and radiancecurrent dependencies in terms of electron and hole mobility,

    luminescence yield, and a current balance factor. When one carrier dominates the current flow,

    as in many practical cases, it is possible to obtain a simple asymptotic relationship which permits a

    test of the assumptions required to obtain the analytic solution. Experimental data from

    poly2-methoxy-52-ethylhexoxy-phenylenevinylene diodes fabricated with various anode and

    cathode materials are evaluated in the context of this analytical approach. 1997 American

    Institute of Physics.S0021-89799703315-X

    I. INTRODUCTION

    Organic and polymeric light-emitting diodes OLEDs

    have become the focus of considerable attention since the

    development of efficient multilayer organic devices1 and the

    subsequent discovery of electroluminescence in several

    classes of conjugated polymer.2 4 In spite of significant ad-

    vances in the understanding of the qualitative behavior of

    OLEDs, a fully quantitative description of charge injection at

    the electrodes, charge transport in the bulk, and electron

    hole recombination has yet to emerge. The analysis of bipo-lar, space-charge-limited current flow dates back at least to

    the 1950s, notably to work at the RCA labs.5 7 The difficulty

    of the problem arises from the necessity of determining self-

    consistently the electric fields at the cathode and anode in-

    terfaces, which simultaneously dictate the injection currents

    and satisfy the electrostatics of the hole and electron densi-

    ties. Since neither the interfacial injection characteristics for

    example, Ohmic, Schottky8 or FowlerNordheim9 are

    known a priori, and since the mobility of both signs of

    charge is known for very few materials of interest, it is by no

    means straightforward to design a series of experiments

    which will permit such a complete and self-consistent de-

    scription. Indeed, the experimentalist is frequently faced withconsiderable uncertainty regarding the chemical and mor-

    phological structure of the interface which may depend not

    only on preparation conditions, but also on the electrical his-

    tory of the device.

    Several recent papers have addressed the transport prob-

    lem for OLEDs, taking diametrically opposite approaches.

    For example, Parker10 used the electric field dependence

    to deduce that in poly2-methoxy-52-ethylhexoxy-

    phenylenevinylene MEH-PPV devices, especially

    those fabricated to show monopolar transport, the current

    voltage (IV) characteristics are controlled by Fowler

    Nordheim tunneling at the interfaces. Davids et al.11 modi-

    fied the FowlerNordheim equations to account for the low

    mobility of carriers in amorphous organic materials. In con-

    trast, Blom and coworkers12 have analyzed IVdata in mo-

    nopolar devices and conclude that the IVbehavior is bulk-

    controlled, with space-charge-limited hole current and trap-

    limited electron current. Both approaches appear to describethe relevant data equally well, which may reflect the fact that

    both injection and bulk limitation are active in the current

    regimes of interest and that the power law behavior of space-

    charge currents passes smoothly into the exponential law for

    tunneling. There may also be inherent difficulties in extrapo-

    lating data from hole-only and electron-only devices, which

    may be particularly sensitive to electrode preparation, to

    truly bipolar OLEDs. For example, a bottom calcium cath-

    ode is almost certainly oxidized in even the best glove box

    and depositing a gold top anode may lead to the diffusion of

    Au atoms into the polymer film. The approach taken in the

    current article permits, in principle, the derivation of the re-

    lation between interface field and injection current. Unfortu-nately in practice, the absence of a set of well-characterized

    material parameters prohibits such a completely quantitative

    evaluation.

    In the following sections, we present the details of the

    analysis necessary to determine the profiles of electric field

    and electron and hole densities across the thickness of a

    singlelayer OLED. The model is based on the solution of the

    bipolar injection problem due to Parmenter and Ruppel

    PR.5 Their approach contains several assumptions: the dif-

    fusive current due to carrier density gradients can be ne-aElectronic mail: [email protected]

    1454 J. Appl. Phys. 82 (3), 1 August 1997 0021-8979/97/82(3)/1454/7/$10.00 1997 American Institute of Physics[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to

    14.139.194.12 On: Fri, 24 Oct 2014 14:26:52

  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    3/8

    glected relative to drift in the applied electric field; deep

    traps for both electrons and holes are ignored; and the mo-

    bility of the carriers is assumed to be independent of electric

    field. One must question the validity of each of these as-

    sumptions in the present case, and it is one of the goals of

    this article to evaluate the extent to which each of them

    holds. In addition, we make one more assumption: recombi-

    nation kinetics can be described by the Langevin

    equation.13,14 This is not absolutely essential mathematically,

    but it greatly simplifies the evaluation of the integrals which

    appear in the PR solution, and in addition reduces by one the

    number of materials parameters which need to be deter-

    mined.

    We cast the theoretical predictions in a form which is

    amenable to comparison with experimental data, namely

    currentvoltage and radiancecurrent relations. In contrast

    to the photoconductive materials which motivated the RCA

    work, the connection between theory and experiment is im-

    proved in the present case by the fact that there is an addi-

    tional experimental observable, namely the intensity of the

    emitted light. The boundary conditions of the PR solution

    enter in the form of a current balance factor which is

    directly related to the radiance of the device. Hence in the

    present case, the emission of the diode itself helps to deter-

    mine the boundary conditions and thus further reduces the

    number of undetermined parameters. Further, when one sign

    of carrier dominates the current, true in many cases of prac-

    tical interest, then a function of current, voltage, radiance,

    and sample thickness is found which depends only on con-

    stant materials parameters. We use the constancy of this

    function to assess the validity of the model and its solution,

    independent of the boundary conditions i.e., of the charge

    injection mechanisms. It is found that the relationship in-

    deed holds reasonably well and therefore we conclude that

    the assumptions and approximations which permit analytic

    solution of the problem are not too severe.The manuscript is arranged as follows. The next section

    summarizes the PR solution and describes the extensions

    made on the basis of Langevin recombination and observable

    radiance. In Sec. III we reanalyze data from several samples

    on the basis of the predictions obtained. Section IV contains

    a discussion of the relevance and limitations of the results

    and the conclusions are ennumerated in Sec. V.

    II. THEORY

    For the convenience of the reader, and to establish nota-

    tion, we reproduce here the essential points of the PR paper.5

    The relevant equations give the electron and hole currents

    JnXNeeF, 1a

    JpXPehF 1b

    as functions of the charge densities N and P , and electric

    field F. Here we use capital letters for physical dimen-

    sionedvariables and will use lower case for reduced, dimen-

    sionless variables. In steady state, the total current, JJnJp, is independent of position, X, between the cathode

    (X0) and anode (XL). The charge carrier mobilities,

    e and h, are taken to be constants, independent of electric

    field. Electronhole recombination is described by bimolecu-

    lar kinetics:

    1

    e

    Jn

    XrN P

    1

    e

    Jp

    X . 2a,b

    For Langevin recombination, which is expected to apply in

    this case where charge transport is due to hopping, the re-

    combination coefficient is given by14

    r e0

    eh 2e00 . 3

    Here, is the dielectric constant of the organic material, and

    this equation defines0 , the recombination mobility. Lastly,

    the electric field satisfies the Poisson equation

    F

    x

    e

    0NP . 4

    This form implies that all carriers are mobile and that none

    are frozen in deep traps. On the other hand, shallow trapping

    events may be included in the definition of a trap-controlled

    mobility, where the carriers are in equilibrium with a trap

    distribution which is no wider than a few times their thermalenergy.

    Following PR, we introduce several new variables:

    aeh0

    20LJ F2, 5

    BJn/J, 6a

    CJp/J. 6b

    Note that this definition ofa differs from PRs A in that it is

    rendered dimensionless by normalizing to the sample thick-

    ness, L. Otherwise we retain the PR notation. B(x) is the

    fraction of the current carried by electrons at normalized

    position xX/L (0 x1) . The dimensionless mobilitiesare

    ee/0 , 7a

    hh/0 , 7b

    where now, because of the Langevin recombination relation

    eh2, 8

    and

    0e,h 2. 9

    The set of Equations 1,2, and 4 then becomes

    BC1, 10

    B

    x

    C

    x

    BC

    a , 11a,b

    a

    x 2e BeC. 12

    The general solution is most conveniently written in terms of

    B(x) which is the solution of the differential equation

    B

    xK1B 1e 1B e1, 13

    1455J. Appl. Phys., Vol. 82, No. 3, 1 August 1997 Scott, Karg, and Carter[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to

    14.139.194.12 On: Fri, 24 Oct 2014 14:26:52

  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    4/8

    where Kis a constant of integration

    K1/Ba

    B cdB Be1 1B 1e. 14

    Then a is given by

    aKBe 1B 2e 15

    and C follows from Eq. 10. The boundary conditions Bcand B a are the fractions of the current carried by electrons at

    the cathode and anode, respectively.

    The variables of physical interest are the charge densities

    and the electric field

    nNL 3j 1/2 2ee

    1/2 B

    a1/2, 16

    pPL 3j 1/2 e2e

    1/2 C

    a 1/2, 17

    fF/ 2e0L

    2j 1/2 ae 2e

    1/2

    , 18

    where the normalized current density is given by

    jJ/ 2e2

    00L

    5 . 19Also of interest is the n p product which gives the profile of

    recombination and hence, in the absence of self-absorption

    and of enhanced interfacial quenching, of luminance

    n pjBC

    a . 20

    In the typical experiment with OLEDs, one generally

    measures the current, voltage, radiance behavior. Devices

    use different metals for anode and cathode and there is there-

    fore a built-in potential which is given to a first approxima-

    tion by the work-function difference of the electrodes. Thebuilt-in potential must be added as a constant of integration

    to the voltage, V. The currentvoltage characteristic is then

    given by

    VVbi

    J1/2 20L

    3

    0eh

    1/2

    K3/2Bn

    B cdBB3/2e1 1B 23/2e

    21

    and the radiance RpN PdX i.e., the emitted optical

    power per unit area is given by

    Rp

    J

    e

    fB cB afb, 22

    where is the mean photon energy and the fluorescence

    yield, f, is the fraction of recombination events which re-

    sults in an externally emitted photon. The form of Eq. 22

    simply states that the quantum efficiency is given as one

    might have deduced by a straightforward argument requiring

    no differential equations by the product of the fluorescent

    yield and the fraction of the electron current which, having

    left the cathode, fails to reach the anode. It suggests the

    usefulness of introducing the current balance factor bBcBa. Equivalently, by Eq. 10, bCaCc, in obvious

    notation. It also provides a means of using experimental

    data to deal with at least one of the boundary conditions

    necessary for the explicit evaluation of the integrals in the

    currentvoltage relation, Eq. 21.

    It may be argued for many combinations of organic and

    electrode materials that one or another sign of carrier domi-

    nates the current. This is almost certainly true at very low

    current levels, where only one electrode injects, but may be-

    come less true at emitting current levels, since clearly both

    signs of carrier must be present. In a previous publication15

    we have described experiments which lead to the conclusion

    that, for polyaniline/MEH-PPV/Ca structures, holes are the

    dominant carrier. With these ideas as motivation, we now

    evaluate the integrals in Eqs. 14 and 21 for small values

    of the current balance factor, b1. Note, however, that the

    equations are completely symmetric with respect to inter-

    change of electrons and holes, and that one could perform an

    identical analysis for small hole current. The asymptotic ex-

    pansion yields an approximate currentvoltage relation

    VVbi

    J1/2 20L

    3

    0eh

    1/2 1

    b1/22 43eB c

    1 1e B c3/2 . 23

    First, let us examine the third factor containing B c1

    and e2. Unless the denominator approaches zero, this has

    a value of order unity, but if, for example, e1 and

    Bc1 then the fraction would diverge. However, in this

    case the numerator is negative, and therefore unphysical. In-

    deed, one might make the hypothesis, based on the fact that

    the numerator must remain positive, that for vanishingly

    small electron mobility, the fraction of the current carried by

    electrons at the cathode, B c, never exceeds 1/2, which is

    physically reasonable. Hence the denominator will never be-

    come vanishingly small and the factor remains of order

    unity.

    Combining Eqs. 21 and 23, we arrive at one of the

    central results of this article

    VVbiRp1/2

    J 20L

    3

    0eh

    e f

    1/2

    O 1 . 24

    The combination of mobilities which enters this expression

    is an effective, or reduced mobility

    eh

    20e

    1h

    11r, 25

    the value of which is dominated by the lowerof the electron

    and hole mobilities.

    To the extent that the initial assumptions regarding dif-

    fusion, trapping, the field-independence of the mobilities, the

    absence of interfacial quenching, and the lack of self-

    absorption are valid, then the right-hand side of Eq. 24 is

    constant for a given sample of thickness L and therefore

    serves as a good experimental test of these assumptions. Un-

    fortunately, to our knowledge, both electron and hole mobili-

    ties are known for none of the materials of experimental

    interest, and it is therefore not yet possible to test fully the

    validity of the result. In the next section, we examine some

    experimental data in light of the above analysis and of the

    materials parameters available to date.

    1456 J. Appl. Phys., Vol. 82, No. 3, 1 August 1997 Scott, Karg, and Carter[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to

    14.139.194.12 On: Fri, 24 Oct 2014 14:26:52

  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    5/8

    Ideally, one wishes to use the available experimental

    data to understand in detail the behavior of a given diode

    structure. The complete solutionEqs.21and 22give the

    three observable quantities I, V, and Rp in terms of sixparameters Vbi , e, h, f, B c , and B a. In principle,

    several of these may be determined by independent experi-

    ments. For example, the built-in voltage is related to the

    work-function difference of the anode and cathode metals. It

    may be even more accurately determined, in the same

    sample as used for IVRp, by electroabsorption

    measurements,16 or from the bias voltage which nulls the

    photocurrent. The carrier mobilities can be determined by

    time-of-flight methods,17 from field-effect transistor

    measurements,18 or from the single carrier space-charge

    limited current.19 The external fluorescent yield may be esti-

    mated from the photoluminescence efficiency if one makes

    the additional and perhaps questionable20 assumption thatthe singlet/triplet statistics introduce an additional factor of

    1/4. If these four parameters are known, then both B c(J) and

    B a(J) can be determined from the experimental data and the

    solution is completely determined. Equation 18 gives the

    electric field throughout the layer, and particularly at both

    electrode interfaces. Hence the charge injection mechanism

    could be unambiguously determined. Unfortunately, we

    know of no diode structure for which such complete infor-

    mation is available.

    III. COMPARISON WITH EXPERIMENT

    Figure 1 shows radiancecurrent data for a series of

    MEH-PPV diodes with polyaniline anodes and several dif-

    ferent cathode metals; Ca, Al, and Ag. The data are plotted

    as external quantum efficiency (Rp/J)( e/) vs logJ. The

    efficiency is, according to Eq. 22, proportional to the cur-

    rent balance factor, b , since the fluorescence yield is as-

    sumed to be constant. Both the current density and b are seen

    to vary by several orders of magnitude. These same data are

    replotted in Fig. 2 in the form suggested by Eq. 24. For

    each sample, the function (VVbi)Rp1/2/J is seen to vary by

    less than one order of magnitude. Note that the hysteresis

    obtained during a voltage cycle is comparable to the varia-

    tion itself. An apparent failure of the model is that, in spite ofthe common active layer, MEH-PPV in all three devices,

    there is more than an order of magnitude difference between

    the Ca and Ag samples. We return to these points below.

    The order of magnitude of the constant (V

    Vbi)Rp1/2/Jis between 0.1 and 10 VW1/2cm/A for each of

    the samples shown in Fig. 2. The value of the right-hand side

    of Eq. 24, using values see discussion section below for

    both electron and hole mobility of 104 cm2/Vs, a thickness

    of 100 nm, and a quantum yield of 1% is approximately

    1 VW1/2cm/A. Thus the quantitative agreement between

    theory and experiment is also encouraging.

    In order to test the predicted thickness dependence we

    replot in Fig. 3 the data of Ref. 15, Fig. 3 in the form ( V

    Vbi)Rp1/2/J L 3/2, following Eq. 24. The data collapse to-

    FIG. 1. External quantum efficiency of three MEH-PPV LEDs with polya-

    niline anodes and different cathodes. For all three devices the thickness of

    the MEH-PPV layer is 100 nm and the highest applied voltage is 5 V. FIG. 2. (VVbi)Rp1/2/Jplotted as a function of current density for the same

    three devices as in Fig. 1. The values used for the built-in potentials are

    given in the legend.

    FIG. 3. (VVbi)Rp1/2/JL 3/2 for several polymer LEDs. These are the same

    data as shown in Fig. 3 of Ref. 15.

    1457J. Appl. Phys., Vol. 82, No. 3, 1 August 1997 Scott, Karg, and Carter[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to

    14.139.194.12 On: Fri, 24 Oct 2014 14:26:52

  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    6/8

    wards a common curve, varying within only a factor of 3

    over nearly four decades in current.

    The agreement between the predictions of the model and

    the experimental data, although far from perfect, is quite rea-

    sonable and indicates that the approximations made are not

    too severe. The variation with different electrode materials,

    and even between batches of MEH-PPV is rather large and

    needs to be explained. If we take the model at face value,

    then according to Eq. 24, for a given device, (V

    Vbi)Rp1/2/J is proportional to (f/r)

    1/2. This quantity

    may be quite sensitive to impurities which quench emission

    and/or provide shallow traps which modify the mobility of

    the charge carriers. Different electrode materials, particularly

    the cathode which is deposited last, may be expected to af-

    fect the polymer layer in different ways. Indeed, recent stud-

    ies have shown that Ca and Al each react with PPV and its

    derivatives2123 and moreover that the nature of the reaction

    products differs between the two metals.24 It is then likely

    that during deposition, metal atoms penetrate the polymer

    layer to a different extent creating different concentrations of

    trapping and quenching defects. Alternatively, the last factor

    in Eq. 23 may be sufficiently different for the different

    cathodes to account for the data. For example, a CaO inter-

    face layer can serve to block holes from entering the cathode,

    makingB c closer to unity. Additional work on other materi-

    als is required to understand these observations in more de-

    tail.

    Another potential source of discrepancy lies in the pos-

    sibility of an inhomogeneous current distribution and there-

    fore inhomogeneous emission across the area of the device.

    Since the radiance and current density enter Eq. 24 with

    different exponents, the value of the left-hand side depends

    on the active area of the device. Although our LEDs look

    relatively uniform to the naked eye, we have not examined in

    detail the pattern of emission of the devices reported here.

    The data of Figs. 1, 2, and 3 were obtained from a volt-age cycle on nearly new diodes and reveal significant hyster-

    esis. This behavior reflects physical and/or chemical changes

    in the diode structure that have been previously noticed and

    sometimes called juvenile aging.25 It is not yet clear

    which parameter f, r, or Vbi is most affected in this

    process, but we note that for Fig. 3, we were forced to use a

    value of the built-in potential for the ITO/Ca couple of 1.8 V,

    rather less than the accepted work-function difference,

    b(ITO)b(Ca4.82.91.9 V, in order to prevent un-

    physical negative values of the abscissa. Thus it is possible

    that some of the aging is due to the development of interfa-

    cial dipole layerswhich modify the built-in potential across

    the bulk of the layer. This is the subject of ongoing investi-gation.

    IV. DISCUSSION

    We start this section with a more detailed examination of

    the assumptions made during the analysis of Sec. II. In light

    of the reasonably good agreement between theory and ex-

    periment, it appears that these assumptions and approxima-

    tions are quite well obeyed in the samples considered to date.

    The field independence of the mobility is somewhat sur-

    prising since most disordered organic charge transport mate-

    rials exhibit a mobility26,27 which varies exponentially as the

    square root of field (exp F). However, the strength ofthis dependence varies with both concentration of chargeable

    sites and with temperature.28 becoming weaker at high con-

    centration and even changing to negative slope above a cer-

    tain material-dependent temperature.29,30

    The experimental data seem to suggest that the reduced

    bipolar mobility,r, is not much less than 104 cm2/Vs and

    therefore that the electron mobility is not much less than the

    hole mobility.31,32 This is in contradiction to many attempts

    to measure the electron mobility by transient

    electroluminescence33,34 EL or by time-of-flight methods35

    which have led to the suggestion that e108 cm2/Vs.

    However, time-of-flight mobility measurements are typically

    made in the small-charge-density regime in order to mini-

    mize space charge effects, and transient EL includes trap

    filling effects before the steady state current is established.

    When the currents flow in light-emitting diodes, it is very

    likely that the steady state density is sufficiently high that all

    available traps and sites of long charge residence time are

    filled. Thus the relevant mobility is not the trap-controlledmobility but the trap-filled mobility which may be sev-

    eral orders of magnitude greater, and may also have a weaker

    field dependence.

    This leads us next to the issue of neglected trapping,

    which may seem to be at variance with the ideas presented in

    the previous paragraph. However, the manner in which

    trapped charge density affects the results of the analysis is to

    alter the electric field distribution across the organic layer. If

    this trapped-charge field is much less than the sum of the

    applied field and the space-charge field due to mobile carri-

    ers, then the effect will be negligible. One must also remem-

    ber that both electrons and holes may be trapped, so that

    some amount of neutralization takes place. From Eqs. 16and17, one may estimate the mobilecharge density at the

    lowest current levels (105 A/cm2) to be of order

    1015 cm3. Note that this is at the point where the emission

    first becomes detectable, in our case at approximately

    106 W/cm2. Thus the approximation of no trapping re-

    quires that the concentration of uncompensated deep traps be

    less than, say, 1014 cm3, which is low, but not unreason-

    able.

    In order to examine the magnitude of the error induced

    by ignoring diffusion, consider the charge density gradients.

    Within the scope of the present model, using Eqs. 16,17,

    and 18, one can, in principle, estimate the relative magni-

    tudes of the hitherto neglected diffusion current and the

    drift current in terms of the poorly characterized diffusion

    constant and mobilities. The result, as is well known,36 scales

    according to j 1/2 and is therefore most important at low cur-

    rent density. For both electrons at the cathode and holes at

    the anode, the ratio of the drift currents is proportional to

    b3/2 and therefore the question of divergence depends on the

    relationship between b and j in the limit of low current. This

    is still an open issue and the subject on ongoing work.

    The last assumption to reexamine is that of Langevin

    recombination, which is based on the concept that if two

    1458 J. Appl. Phys., Vol. 82, No. 3, 1 August 1997 Scott, Karg, and Carter[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to

    14.139.194.12 On: Fri, 24 Oct 2014 14:26:52

  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    7/8

    oppositely charged carriers, with sufficiently short mean free

    paths, approach each other within the Coulomb radius,

    rce2

    40kT, 26

    then they will inevitably recombine. For this to apply in the

    present case, the Coulomb radius must be small compared to

    the thickness of the sample, and also the electron and hole

    densities must be small enough that there is a low probability

    of finding multiple carriers within the volume of the Cou-lomb sphere. At room temperature and for a dielectric con-

    stant of 3.5, typical for conjugated polymers, the Coulomb

    radius is approximately 20 nm, sufficiently less than the

    usual sample thicknesses which are about 100 nm. At the

    very highest current levels 0.1 A/cm2 in the present work,

    and up to 1 A/cm2 in other experiments, the charge densities

    can be estimated Eqs. 16 and 17 to approach

    1017 cm3 and therefore the number within the Coulomb

    sphere is no more than 3. For most experimental data this

    limit is rarely exceeded and hence the Langevin assumption

    is generally valid.

    The data plotted in Figs. 2 and 3 does seem to show a

    consistent trend in its deviation from constancy: all thecurves are convex, and many show a maximum in mid- to

    upper range of current. This is presumably due to break-

    downs in the assumptions and approximations used in deriv-

    ing the analytical solution. At this stage in our investigations,

    it is not yet clear which particular assumptions might ac-

    count for the observed shape. It is also worth mentioning that

    care must be taken to eliminate or account forleakage cur-

    rents at the level of tens of pA, as well as the series resis-

    tance of the electrodes which may give an additional voltage

    drop at high current. Lastly we mention the uncertainty in the

    built-in potential. As noted above, this is clearly not just the

    difference in workfunctions of the electrodes, but may in-

    clude the effects of interfacial dipole layers. Moreover, weare currently investigating the possibility that it can change

    with operation and may therefore account for the observed

    hysteresis.

    The solution given by PR embodies the boundary condi-

    tions, i.e., the charge injection at the electrode interfaces, in

    the parametersB c andB a. It will prove useful in subsequent

    discussions to use a diagram for reference. In Fig. 4, we have

    sketched the various points of interest in the (Bc,Ba) plane.

    Both B c and Ba are fractions less than one. Since electrons

    are injected at the cathode and may recombine before they

    reach the anode, B aB c. Thus the region of physical inter-

    est is the triangle bounded by B a0, B c1, and B aB c.

    The origin represents the condition of hole-only current and

    the point 1,1 corresponds to electron-only current. Along

    the line B aBc there is no recombination ( b0) and lines

    of equal b are parallel and to the right of this. The locus

    B a1B cCc describes balanced injection where the frac-

    tion of the current carried by electrons at the anode equals

    that carried by holes at the cathode. Finally, the point 1,0

    represents the ideal condition of balanced injection and com-

    plete recombination. The behavior of the integrals in Eqs.

    14 and 21, can be mapped onto this diagram, with the

    reduced electron mobility, e, as an implicit parameter. The

    problem to be solved is then to trace in detail the path taken

    on this diagram by the cathode-organic-anode system as the

    total current density is increased.

    In Sec. II, we described the behavior of the voltagecurrent relation, Eq. 21, for values of B c and b near the

    origin. There is a square-root divergence in V(b1/2) as

    b tends to zero, for all nonzero Bc. At the origin hole-only

    current the divergence is suppressed. In this case, and for

    he1, the last two factors of Eq. 21 can be evaluated

    exactly to yield the finite value, 2/3, corresponding to the

    single carrierChilds lawsolution.19 We have confirmed by

    numerical integration that the square-root divergence persists

    for small b deviations along the entire line of no recombina-

    tion except at the endpoints. At the point of balanced injec-

    tion and complete recombination, 1,0, again the integrals

    can be exactly evaluated for he1, yielding /8. These

    analytical and numerical results, along with other numericalevaluations which we do not report here, reveal that the fac-

    tor K and the other integral in Eq. 21 are slowly varying

    functions both ofe and of position on the (Bc,Ba) diagram

    except near the line of no recombination. Thus, for most

    values of the current balance factor which are not too small

    the currentvoltage relation is well approximated by the fa-

    miliar so called Childs law relation, J V2, independent

    ofb . Thus one expects to find a crossover from the predic-

    tion of Eq. 24 to Childs law behavior, albeit with a nu-

    merical constant different from the usual 9/8 and with an

    effective mobility, as the current density is increased.

    It is clear from the foregoing discussions that, in general,

    there is an electron current at the anode as well as the cath-

    ode, and similarly, a hole current at both electrodes. Consid-

    erable attention10,11,37 has been given to the injection currents

    electrons at the cathode and holes at the anode and their

    dependence on field and/or voltage. A complete description

    of the problem also necessitates consideration of the extrac-

    tion currents. These will depend not only on the nature of

    any interfacial barrier and the electric field across it, but also

    on the density of carriers at each interface. The equations

    given above permit the calculation of these densities and the

    electric field, but for a complete solution the boundary con-

    FIG. 4. Diagramatic representation of the boundary conditions appropriate

    to the PR solution.B c and B a are the fractions of the total current carried by

    electrons at the cathode and anode, respectively. Their difference, b , is the

    current balance factor.

    1459J. Appl. Phys., Vol. 82, No. 3, 1 August 1997 Scott, Karg, and Carter[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to

    14.139.194.12 On: Fri, 24 Oct 2014 14:26:52

  • 8/10/2019 Langevin Bipolar Charge and Current Distributions in

    8/8

    ditions must be determined self-consistently for any given

    model of injection and extraction.

    V. SUMMARY AND CONCLUSIONS

    We have analyzed a model for the charge distributions in

    OLEDs by adapting the work of Parmenter and Ruppel to

    include Langevin recombination, and have taken advantage

    of the fact that, to a good approximation, the recombination

    rate is proportional to the observed luminous emission. The

    results are contained in the two equations Eqs. 21 and22 which describe the currentvoltage and current

    radiance behavior in terms of the fraction of the current car-

    ried by electrons at the cathode and anode. These boundary

    conditions remain to be more fully determined by further

    examination of experimental data, but, by means of an

    asymptotic expansion valid when one sign of carrier domi-

    nates in the low current regime, we predict that the combi-

    nation of variables (VVbi)Rp1/2/JL 3/2 should be constant.

    The prediction is borne out to a reasonable degree of accu-

    racy in experimental data obtained on MEH-PPV diodes

    with various electrode materials and for various thicknesses.

    Hence we conclude that the assumptions inherent in the PR

    analysis, namely negligible diffusive currents, field-independent mobilities and few deep traps, are not too se-

    vere.

    The concept of a current balance factor arises naturally

    from the PR solution, and we have shown that it is directly

    proportional to the emission. We have also introduced a dia-

    grammatic method for locating the relevant boundary condi-

    tions and as a guide to describing the behavior of devices as

    the current density increases.

    In this article we have not addressed in any detail the

    question of the charge injection processes at the anode and

    cathode interfaces. Because of the universal nature of the

    square-root singularity at small values ofb , and because the

    carrier mobilities are not yet known with sufficient accuracy,it is not yet possible to use the available data to locate the

    exact operating conditions on the (B c,B a) map. This there-

    fore becomes the focus of future work. We also hope that the

    present analysis can be used to help identify the nature of the

    changes in device performance with aging by separating

    electrode degradation i.e., the boundary conditions implicit

    in the current balance factor and bulk degradation reflected

    in the function (VVbi)Rp1/2/J which depends only on mo-

    bility and fluorescence yield. Finally, it is clear that a fully

    quantitative solution may need to include some of the effects

    neglected here, and will have to be obtained numerically.

    The results of the analytic solution should then be useful in

    estimating the values of numerical parameters.

    ACKNOWLEDGMENTS

    We are grateful to Barbara Jones for insights into the

    properties of the integrals of Eqs. 14 and 21, to Phil

    Brock and Marie Angelopoulos for providing the materials

    used in this work, and to Weidong Chen for critical reading

    of the manuscript.

    1 C. W. Tang and S. A. Van Slyke, Appl. Phys. Lett. 51, 913 1987.2 J. H. Burroughes, D. D. C. Bradley, A. R. Brown, R. N. Marks, K.

    Mackey, R. H. Friend, P. L. Burns, and A. B. Holmes, Nature London

    347, 5391990.3

    G. Grem, G. Leditzky, B. Ulrich, and G. Leising, Synth. Met. 51

    , 3891992.4 Y. Ohmori, M. Uchida, K. Muro, and K. Yoshino, Jpn. J. Appl. Phys. 30,

    L19411991.5 R. H. Parmenter and W. Ruppel, J. Appl. Phys. 30, 15481959.6 M. A. Lampert, RCA Rev. 20, 682 1959.7 M. A. Lampert, Phys. Rev. 125, 126 1961.8 W. Schottky, Zeits. f. Physik118, 5391942.9 R. H. Fowler and L. Nordheim, Proc. R. Soc. A119, 173 1928.

    10 I. D. Parker, J. Appl. Phys. 75, 1656 1994.11 P. S. Davids, Sh. M. Kogan, I. D. Parker, and D. L. Smith, Appl. Phys.

    Lett. 69, 2270 1996.12 P. W. M. Blom, M. J. M. de Jong, and J. J. M. Vleggar, Appl. Phys. Lett.

    68, 33081996.13 P. Langevin, Ann. Chem. Phys. 28, 289 1903.14 M. Pope and C. E. Swenberg, Electronic Processes in Organic Crystals

    Oxford University Press, New York, 1982, p. 502.15 S. Karg, J. C. Scott, J. R. Salem, and M. Angelopoulos, Synth. Met. 80,

    1111996.16 I. H. Campbell, T. W. Hagler, D. L. Smith, and J. P. Ferraris, Phys. Rev.

    Lett. 76, 1900 1996.17 F. K. Dolezalek, Photoconductivity and Related Phenomena, edited by J.

    Mort and P. M. Pai Elsevier, Amsterdam, 1976, p. 27.18 F. Garnier, A. Yassar, G. Horowitz, and F. Deloffre, Mol. Cryst. Liq.

    Cryst. 228, 811993.19 M. A. Lampert and P. Mark, Current Injection in Solids Academic, New

    York, 1970.20 L. J. Rothberg, M. Yan, F. Papadimitrakopoulos, M. E. Galvin, E. W.

    Kwock, and T. M. Miller, Synth. Met. 80, 411996.21 Y. Gao, K. T. Park, and B. Hsieh. J. Appl. Phys. 73, 7894 1993.22 E. Ettedgui, H. Razifitrimo, Y. Gao, B. R. Hsieh, W. A. Feld, and M. W.

    Ruckman, Phys. Rev. Lett. 76, 299 1996.23 P. Broms, J. Birgersson, N. Johansson, M. Logdlund, and W. R. Salaneck,

    Synth. Met. 74, 179 1995.24 W. R. Salanek and J.-L. Bredas, Adv. Mater. 8 , 481996.25 S. A. Carter, M. Angelopoulos, S. Karg, P. J. Brock, and J. C. Scott, Appl.

    Phys. Lett. 70, 20671997.26 P. M. Borsenberger and D. S. Weiss, Organic Photoreceptors for Electro-

    photographyMarcel Dekker, New York, 1993.27 H. Bassler, Phys. Status Solidi A 140, 91993.28 L. B. Schein, Philos. Mag. B 65, 7951992.29 A. Peled and L. B. Schein, Chem. Phys. Lett. 153, 422 1988.30 P. M. Borsenberger, L. T. Pautmeier, and H. Bassler, J. Chem. Phys. 94,

    54471991.31 H. Meyer, D. Haarer, H. Naarman, and H. H. Horhold, Phys. Rev. B 52,

    25871995.32 M. Gailberger and H. Bassler, Phys. Rev. B 44, 86431991.33 H. Vestweber, J. Oberski, A. Greiner, W. Heitz, R. F. Mahrt, and H.

    Bassler, Adv. Mater. 2, 1971993.34 S. Karg, V. Dyakonov, M. Meier, W. Riess, and G. Paasch, Synth. Met.67, 165 1994.

    35 D. Haarerprivate communication.36 M. A. Lampert and P. Mark, Current Injection in Solids Academic, New

    York, 1970.37 S. Karg, W. Riess, M. Meier, and M. Schwoerer, Mol. Cryst. Liq. Cryst.

    236, 791993.

    1460 J. Appl. Phys., Vol. 82, No. 3, 1 August 1997 Scott, Karg, and Carter[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to