inhomogeneous transport of alq3

7/29/2019 Inhomogeneous Transport of Alq3

http://slidepdf.com/reader/full/inhomogeneous-transport-of-alq3 1/10

Synthetic Metals 156 (2006) 1108–1117

Inhomogeneous transport property of Alq3 thin films:Local order or phase separation?

I. Thurzo∗, H. Mendez, C. Iacovita, D.R.T. Zahn

Institute of Physics, TU Chemnitz, D-09107 Chemnitz, Germany

Received 31 January 2006; accepted 14 July 2006

Available online 7 September 2006

Abstract

Steady-state current–voltage ( I –V ) and impedance–voltage ( Z –V ) measurements were performed on in situ (UHV) prepared metal (Ag,Al)/Alq3 /indium-tin oxide (ITO) devices after exposure to air. When increasing the positive bias on the top metal electrode to a relatively well-

defined critical value, a transition from semiconducting to semi- or even insulating behavior of the contacted Alq 3 thin film is observed by means

of I –V measurements. The final insulating state remains stable when applying negative bias to the Ag electrode. In the case of the Al electrode,

there is a voltammetric current wave under a well-defined negative bias indicating a redox reaction of mobile ions at the Al electrode.

The Z –V measurements reveal a peculiar feature of ac transport through the Alq3 thin films, namely the equivalent series capacitance is equal to

its parallel counterpart in the frequency range from 100 to 1 MHz and amounts to only a fraction (0.3–0.5) of the expected geometrical capacitance

of the device. An equivalent electrical circuit has been developed, based on the existence of two parallel transport paths: an insulating (amorphous)

Alq3-phase shunted by a semiconducting (semi-insulating) one, both running into the impedance of the back contact. The equivalent circuit

model composed exclusively of frequency independent elements is useful for predicting the maximum frequency for retaining the full geometrical

capacitance. Even though the model is capable of describing the bias dependence of the impedance correctly, it does not shine light on the nature

of the (ordered) phase or domain responsible for the dielectric loss. The possibility of local order connected with dipole–dipole interaction in the

metal/Alq3 interface zone is discussed. In any case, the ordered portion of the organic material seems to form the huge interface dipole of about

1 eV with Ag or Al [M.A. Baldo, S.R. Forrest, Phys. Rev. B 64 (2001) 085201], the direction of the dipole promoting electron injection to Alq3.

Then the semiconductor-to-insulator transition could be initiated by a damage of the interface dipole under a critical positive dc bias of the metal,

preventing the flow of both dc and the real component of low-frequency ac current. The transition is not accompanied by any significant change

in the impedance of the back contact common to both phases.

© 2006 Elsevier B.V. All rights reserved.

Keywords: Organic semiconductor; Conduction; Dielectric loss; Inhomogeneity

1. Introduction

Besides polymers there is a wide class of materials based

on conjugated small organic molecules occuring in state-of-the-

art devices like organic light-emitting diodes (OLEDs), light-

emitting electrochemical cells(LECs)and field-effect transistors

with channels made of such small molecules (OFETs). In the

case of prototypic light generating devices usually two differ-

ent materials form a heterojunction, one of them transporting

∗ Corresponding author. Tel.: +49 371 531 3079; fax: +49 371 531 3060.

E-mail addresses: [email protected],

[email protected] (I. Thurzo).

injected electrons and the second one transporting holes to the

zone of recombination (electroluminescence). Therefore, it is

not surprising that there have been many attempts to understand

the physics of processes at metal/organic thin film interfaces

and the transport mechanism through the active organic lay-

ers. Now valuable reviews of this topic are available, mainly

by Brutting et al. [1] and Scott [2]. There is an intimate con-

nection between electroluminescence and charge transport in

organic light-emitting devices [3]. The organic semiconduc-

tor tris-(8-hydroxyquinoline) aluminum (Alq3) is known to be

an effective electron transporter in electro-luminescent devices.

Apart from practical usage of Alq3 it is also a suitable object for

modeling structural and electronic properties [4]. Depending on

preparation conditions and thermal history, several crystalline

0379-6779/$ – see front matter © 2006 Elsevier B.V. All rights reserved.

doi:10.1016/j.synthmet.2006.07.002

mailto:[email protected]


http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.synthmet.2006.07.002

http://localhost/var/www/apps/conversion/tmp/scratch_4/dx.doi.org/10.1016/j.synthmet.2006.07.002





I. Thurzo et al. / Synthetic Metals 156 (2006) 1108–1117 1109

modifications of Alq3 are observable, along with isomorphism

(meridional and facial isomers), as reported by Colle et al. [5–8]

and Martin et al. [9].

An important issue is the problematic stability and durability

of organic devices, prompting for studies of degradation phe-

nomena due to environmental and/or electrical stress [10–12].

The latter kind of impact is often treated in terms of mostly

unidentified traps responsible for the degradation of luminance.

Related studies devoted specifically to Alq3 were reported by

Papadimitrakopoulos and Zhang [13], Kim et al. [14] and Xu

and Xu [15]. It is possible to have different morphologies of

Alq3 thin films when changing the deposition rate [16–18].

In addition to the almost mandatory current–voltage ( I –V )

characterization of organic based devices or single organic lay-

ers, impedancespectroscopy is utilized to describe theirdynamic

behavior [19–27]. The outcome of such investigations is usually

presented as a sequence of artificially introduced zones (frac-

tional impedances or admittances), assigned to metal/organic

interfaces and bulk of the active layers. The zonal approach

works well as long as the injection level is negligible (dc biasclose to zero), spurious effects emerge if high forward biases are

applied, comprising a steep increase in capacitance at low fre-

quencies, negative capacitance, and even an inductive behavior

due to relaxation of the injected space charge, as demonstrated

by Martens et al. [28] f or the poly( p-phenylene vinylene) poly-

mer.

Our present contribution to anomalous transport behavior of

metal/Alq3 /ITO single-layer devicesbased on 100 nm thick Alq3

thin films is aimed at demonstrating instabilities of I –V curves at

a critical positivepotential of thetop metal electrode, as well as at

modeling the impedance leading to the experimentally observed

equivalent capacitance being fairly below the geometrical one(εr = 2). The results point to a laterally inhomogeneous mate-

rial, as a consequence of either phase separation or formation

of local ordered domains due to dipole–dipole interaction. The

latter phenomenon is detectable optically as a red shift of the flu-

orescence spectrum of Alq3 doped by the highly polar molecule

like DCM2 [29]. An independent evidence for the existence

of such domains in vacuum-deposited undoped Alq3 thin films

has been provided by Ishii et al. [30] on the basis of mapping

the profile of the persistent (in dark) internal built-in voltage.

Finally, a mechanismfor the observedsemiinsulator-to-insulator

transition,whilebiasing metalpositively, will be suggestedstart-

ing from a possible damage to the electron injection promoting

metal/Alq3 interface dipole (∆≈1 eV) reported by Baldo andForrest [31]. Stable forward I –V curves were observed for Ag,

Al cathodes whereas minimized hole injection was found when

Mg was used as anode [31].

2. Experimental

The 100 nm thick Alq3 thin films were deposited onto chem-

ically treated ITO substrates from a Knudsen cell in UHV

(base pressure≈10−9 mbar) at a rate of 0.15 nm/min. After

organic molecular beam deposition circular metal (Ag or Al)

dots of different areas (S = 2.82×10−3 cm2, M = 5× 10−3 cm2,

B = 7.85× 10−3

cm2

) were evaporated through a shadow mask.

Both I –V and Z –V measurements were performed in air (in dark)

at room temperature, using a HP-4061A Semiconductor Com-

ponent Test System. When measuring I –V curves in the first run,

i.e. after 10 min exposure to air, the bias was scanned at respec-

tive rates 0.01 and 0.02 V s−1, the measurement cycle starting

and terminating at a negative voltage of −1 or −2 V, respec-

tively. The delay t of each successive reading of the current

was always set to 1 s. For probing impedance, 20 mV (rms) of

ac voltage superimposed on the dc bias U scanned between 1

and−1 V was used (usually runs 2 and 3).

3. Results

3.1. I–V curves

A typical I –V curveof virginAg/Alq3 /ITO devicesis depicted

in Fig. 1. The device was initially in the semiconducting

(ohmic) state, characterized by a relatively high current den-

sity j > 1 A c m−2 at U Ag =−1 V. Upon return from U Ag = 1 V

towards −1 V a semiconductor-to-semiinsulator transition was

observed at U Ag≈ 0.9 V, the semiinsulating (SI) state being

usually characterized by a current density j between 10−6 and

10−5 A cm−2 at−1 V. Except for biases U close to zero, now the

j(U ) dependence obeys a power law of j = P0U n for a given polar-

ity of the bias, with P0 being a constant. This kind of transition

presumably stems from burning out of microscopic metallic fil-

aments carrying the initial high current, a phenomenon observed

earlier by Van Slyke et al. [32]. The symmetry of I –V curves in

the SI-state is not surprising, keeping in mind the almost iden-

tical work functions of Ag and ITO. A similar transition was

also observed in the case of Al/Alq3 /ITO diodes as evidenced

by Fig. 2a and b shows the fit of j in the SI-state as a functionof U . The spike in the current preceding the dominant transition

is a typical feature observed systematically for both Ag and Al

top electrodes, with the threshold voltage being slightly higher

for (partially oxidized?) Al.

Situations occurred when there were two current spikes and

two subsequent transitions in I –V curves at positive biases, see

Fig. 3. Thespikeat U Al≈1.4 V is followed by the creation of the

Fig. 1. Current–voltage characteristics of an Ag/Alq3 /ITO device showing a

semiconductor-to-semiinsulator transition, the size B of the dot is found in the

text.



1110 I. Thurzo et al. / Synthetic Metals 156 (2006) 1108–1117

Fig. 2. (a) Current–voltage curves of an Al/Alq3 /ITO device before (solid line,

run 1) and during the semiconductor-to-semiinsulator transition. Runs 2 and 3

were utilized for recording initial impedance data while cycling the bias U from

−1 to 1 V and back. (b) Fitting the current density (run 4) upon the return from

U = 0 to −2 V as j = P0 exp(−U n).

SI-state, whereas the second spike precedes a SI-to-insulating

state (I) reaction, with the current density at −1 V dropping

below 10−11 A cm−2. To our knowledge, such a low current

density through Alq3 under effective electric field of 105 Vcm−1

has not been reported previously. The stepwise increase of the

Fig. 3. Current–voltage curves of still another Al/Alq3 /ITO device undergoing

a semiconductor-to-semiinsulator transition followed by a semiinsulator-to-

insulator reaction. An electrochemical “voltammetric current wave” is observed

at U ≈−1.5 V upon the return.

current at−1.5 V on the return was detected reproducibly for all

Al-contacted devices. Its origin will be discussed later. The last

feature to mention is the offset voltage of 0.3 eV for the current

reversal upon the return, which is caused by the interference of

the positive current through the device with either a comparable

negative background current of theinstrumentationor a transient

current through the device, the latter initiated by each negative

step of the decaying positive bias.

3.2. Impedance

Frequency domain investigations were always started by

probing both initial impedance ( Rs, C s) and admittance ( Rp, C p)

of the metal/Alq3 /ITO devices under zero bias in the frequency

range from 100Hz to 1 MHz. Surprisingly, both C s and C p were

equal one to another andindependent of frequency. Still a crucial

peculiarity is revealed after comparing the specific capacitance

(per unit area) value with that one calculated for the geometrical

capacitance of an organic layer with relative permittivity εr = 2

and thickness of 100 nm. Despite the perfect scaling with thearea of metal dots the capacitance values did not exceed one

half of the calculated value. Since at zero bias there is no mea-

surable dielectric loss due to dc conductivity as a possible cause,

one needs a non-standard physical model to tackle the paradox

at least semi-quantitatively. The basic idea (Model I) is that of

having a laterally inhomogeneous thin film and hence two par-

allel paths (channels) for the transport of ac current through the

organic material:

(i) an insulating (I) phase essentially free of dielectric loss

contributes the fractional capacitance (1− a)C to the total

geometrical capacitance C (0 < a <1);(ii) a semiinsulating (SI) phase contributes the complementary

capacitance aC in series with a non-linear resistor R(U )

representing the bias-dependent dielectric loss. The cor-

responding equivalent circuit for this Model I is drawn

schematically at the top of Fig. 4. Here, the resistor R1 was

added as the path resistance on ITO from the contacting pad

to the diode selected.

Basic features of Model I such as simulated impedance data will

be presentedin separate double-logarithmic plots of real ( Z r)and

imaginary ( Z i) parts of impedance Z against frequency f . The

same format shall later be used for showing fitted experimental

data. The quantities Z r and Z i inherent to the equivalent circuitI from Fig. 4 can be expressed via a set of equations (ω = 2π f ):

Zr = R1 +a2R

1+R2C2ω2a2(1− a)2, (1)

Zi =1+ R2C2ω2a2(1− a)

ωC[1+ R2C2ω2a2(1− a)2]. (2)

The impact of parameters a and R on Z r is illustrated in Fig. 5a

and b for given values of C and R1. It is worth mentioning

that for ω→0 Z r is approaching a2 R while the high-frequency

limit is R1. On the other hand, at sufficiently low frequencies

Z i approaches (ωC )−1

as demonstrated in Fig. 6. The frequency




Fig. 4. Basic (I) and extended model (II) to describe impedance of

metal/Alq3 /ITO devices.

of reaching the full capacitance C is a function of R and coin-

cides with that of reaching the plateau of Z r at low frequencies

(Fig. 5b). For sufficiently high values of R the equality of equiv-

alent capacitances C s and C p is confirmed, unless the transition

frequency is approached—Fig. 7. In figures that follow, Model

I will be labeled as “ R, R1,C ,a”.

Fig. 5. Double logarithmic Z r– f plot of the real part of impedance—simulation

by Model I using as parameter: (a) factor a and (b) resistor R.

Fig. 6. Double logarithmic Z i– f plot of the imaginary part of

impedance—simulation by Model I using the factor a as parameter. The

frequency of the transition to the full geometrical capacitance C is always given

by the time constant RC .

Now let us inspect to what extent does Model I apply to the

experimental impedance data. The real part of the impedancedata of the sample from Fig. 1 is shown in Fig. 8a by full

squares. The corresponding fit using Model I is included as

solid lines. While matching very well the experimental data at

lower frequencies, there is an evident failure of Model I when

approaching and exceeding f = 104 Hz. Here, the data clearly

indicate another dispersion of Z r. This finding is the reason for

including an additional parallel R2C 2 circuit leading to Model II

at the bottom of Fig. 4. In order to keep the number of unknown

parameters as low as possible, when fitting data only the real

part R2s of the additional impedance was considered:

R2s=

R2

1+ ω2τ 2 , (3)

where τ = R2C 2. Instead of varying C 2, the characteristic

frequency f 0 of the dispersion, fulfilling the condition ω0τ = 1,

is obtained from the fit. Finally, knowing ω0 the unknown

capacitance C 2 is found after obtaining R2s. The dispersion due

to this faster relaxation process including R1, R2 and C 2 is plot-

ted in Fig. 8a as the dashed line. The full fit employing Model II

(“ R, R1,C ,a, R2,C 2”) is represented by circles. The capacitance

Fig. 7. The equivalent capacitances C s and C p are equal to each other unless the

transition frequency ( f ≈ 102

Hz) is reached (Model I).




Fig. 8. Double logarithmic Z r– f plot of the device from Fig. 1 in the semiinsu-

lating state is decomposed to individual components, lists of fitting parameters

are added for biases: (a) U = 0 V, (b) U =−1 V and (c) U = 1 V.

C =1.39× 10−10 F (εr = 2) corresponds to the geometrical

capacitance of a dot with the diameter B. An important point to

make is that the capacitance C 2s is not taken into account when

fitting Z i assuming C 2sC , a usually fulfilled condition.

When exploring theeffectsof externaldc bias, we restrict our-

selves to the ultimate biases (−1 V, 1 V) applied to the device

during impedance measuements within a complete cycle pro-

gressing in V = 0.1 V steps. There was no hysteresis in Z –V

curves in this range of biases. Device parameters reflecting

changes in Z r under application of thetwo biasescan be extracted

from Fig. 8b and c. As expected, the only remarkable change

with bias is that of R which is reduced by one order of mag-

Fig. 9. Double logarithmic Z i– f plot of the device from Fig. 1 in the semiinsu-

lating state is computed and compared to the experimental data for biases: (a)

U = 0 V, (b) U =−1 V and (c) U = 1 V.

nitude for U =−1 V (electron injection from the metal). Thescatter of the factor a as well as of the R2, C 2 values lies within

the reliability of the fitting procedure.

Adopting the parameters that fit Z r, the related imaginary part

of impedance Z i was calculated according to Model II and com-

pared with experimental data. The results of such a comparison

are shown in Fig. 9a–c. The equivalent capacitance C s is only

a fraction of the full capacitance C = 1.39× 10−10 F and inde-

pendent of bias. At low frequencies and applied biases a steep

increase of C s is observed, mainly under U =−1 V. The lower

R the higher is the excess capacitance, or put in other words,

the error in C s is due to the loss factor D = RsC s approaching

and even exceeding the value of 10. This loss due to dc current




under non-zero bias prevents the experimental verification of the

transition from the fractional aC capacitance to the full one of

C . Reducing the bias to zero causes R to increase to values close

to and higher than 108 , thereby reducing the low-frequency

ac current below the limit of detection (missing data).

With respect to the mechanism causing the transition from

the semiconducting to the semiinsulating state, an inspection of

changes in the equivalent circuit after the transition is of pri-

mary interest. Such an inspection was impossible for the above

treated system because of a too large factor D in the initial state.

From this point of view, the device Al/Alq3 /ITO from Fig. 2 is

much more suitable, as will be proved by the impedance analy-

ses before and after the transition. Considering the initial state

under zero bias the related Z r versus f plot is shown in Fig. 10a.

The initial value of R is approximately identical to the final one

of the Ag/Alq3 /ITO device treated above. One can estimate the

impact of applied bias upon the parameters of the equivalent

circuit when inspecting Fig. 10b and c. Owing to the reduction

of R by factors 30 and 20 under biases −1 and 1 V, respectively,

the leveling up of Z r at a2 R is proven. Despite the shift of thedominant relaxation to higher frequencies under the biases, the

essentially stable faster relaxation is still discernable. Note the

excellent reproducibility of the coefficient a, an evidence for the

stability (against bias) of the zone contributing to the fractional

capacitance aC . This finding can be directly confirmed by the

set of C s values obtained via fitting linear portions of Z i versus f

plots at different biases (Fig. 11a–c). The apparent (instrumen-

tal) giant increase of C s at low frequencies (by several orders

of magnitude) for non-zero biases is also evident from Fig. 11b

and c.

Letus turn ourattention to thechanges in impedancebehavior

caused by the switching from the initial semiconducting to finalsemiinsulating (SI) state of the device, evident from Fig. 2. After

fitting the corresponding Z r and Z i data, the parameters are listed

in Table 1.

The quantity Rdc was assessed from the I –V curves as

U /( j(U ) B). Two basic conclusions can be drawn from the table

when confronted with Figs. 10 and 11:

(i) the injection efficiency (represented by R)undertheultimate

biases is reduced by approximately one order of magnitude,

(ii) the fractional volume occupied by the capacitance aC is the

same as before the switching.

Table 1

Fitting parameters of the sample from Fig. 2, SI-state

Parameter U = 0 V U =−1 V U = 1 V

R () 3.6× 108 2× 108 1.5× 108

R1 () 78 80 80

C (F) 1.39× 10−10 1.39× 10−10 1.39× 10−10

a 0.65 0.68 0.68

f 0 (Hz) 3× 104 3× 104 5× 104

R2 ( 1200 1200 600

C 2 (F) 4.42× 10−9 4.42× 10−9 5.3× 10−9

C s (F) 5.22× 10−11 5.22× 10−11 5.22× 10−11

Rdc () – 6× 108 1.06× 108

Fig. 10. Double logarithmic Z r– f plot of the device from Fig. 2 in the initial

(semiconducting) state is decomposed to individual components, lists of fitting

parameters are added for biases: (a) U = 0 V, (b) U =−1 V and (c) U = 1 V.

The slightly reduced C s after reducing the bias in the SI-state

to zero remains a subject for speculation, for in general one

expects the capacitance to increase with the amount of excess

charge. As expected, there was no instrumental excess capaci-

tance C s in the SI-state at the lowest frequency used, irrespective

of applied bias.

In order to demonstrate the scaling of the anomalous capac-

itance with dot area, as well as to characterize the insulating

I-state after the SI→ I switching, the relevant impedance data

of the device from Fig. 3 (dot M) will be presented in terms of

the parameters in Table 2 (SI-state) and Table 3 (I-state).




Fig. 11. Double logarithmic Z i– f plot of the device from Fig. 2 in the initial

(semiconducting) state is computed and compared to the experimental data for

biases: (a) U = 0 V, (b) U =−1 V and (c) U = 1 V.

Table 2

Fitting parameters of the sample from Fig. 3, SI-state

Parameter U = 0 V U =−1 V U = 1 V

R () 3.3× 107 3× 106 3.6× 106

R1 () 1520 1490 1520

C (F) 8.85× 10−11 8.85× 10−11 8.85× 10−11

a 0.71 0.7 0.7

f 0 (Hz) 5× 104 5× 104 5× 104

R2 () 1000 1200 1200

C 2 () 3.18× 10−9 2.65× 10−9 2.65× 10−9

C s (F) 3.45× 10−11 3.41× 10−11 3.56× 10−11

Rdc () – 4× 106 6.4× 106

Table 3

Fitting parameters of the sample from Fig. 3, I-state

Parameter U = 0 V U = -1 V U = 1 V

R () >4× 108 >4× 108 >4× 108

R1 () 108 107 107

C (F) 8.85× 10−11 8.85× 10−11 8.85× 10−11

a 0.7 0.7 0.7

f 0 (Hz) 5× 104 5× 104 5× 104

R2 () 1200 1200 1400

C 2 (F) 2.65× 10−9 2.65× 10−9 2.27× 10−9

C s (F) 3.35× 10−11 3.37× 10−11 3.35× 10−11

Rdc () – 2.5× 1013 1.05× 1013

For calculating the parameter Rdc in a similar way as above

the dot area M instead of B was used. In the initial SI-state the

factor a amounts to about 0.7 when taking C = 8.85× 10−11 F

for the M-dot area, the bias-induced injection enhancement ratio

in the initial state is close to 10. The supposition C 2 (C 2s)C

is justified. Once again the apparent instrumental increase of C s

at low frequencies was present, now even for zero bias. Unfor-tunately, the onset of this instrumental error coincides with the

frequency where the transition to full capacitance is expected.

It is not easy to specify the consequences of the SI→ I

reaction starting from impedance measurements and fitting the

strongly limited Z r data inherent to the I-state taken when com-

posing the Table 3. With f = 103 Hz and zero bias, one arrives at

R≈4× 108 to be taken as a starting value for the remaining

Z r– f plots for the ultimate biases, where there are no data for

f < 104 Hz. The missing R-values at low frequencies justify the

condition R > 4× 108 found in Table 3. The latter finding is

a reason for claiming that in the I-state there is negligible bias-

induced charge injection into Alq3 from the top metal electrode

up to an effective electric field E of 105 V cm−1. On the other

hand there seems to be no significant change in the fraction

a≈0.7 of the still present slightly conducting phase if adopting

the Z i– f data for the I-state.

We close thepresentation of experimental data stating that the

imaginary part Z i of the impedance is always dominated by the

fractional capacitance aC unless an instrumental error coming

from dc conductivity occurs. Contrary to this, the real part Z rprovides correct data on R down to the lowest frequency used if

the value of R remains low enough for obtaining any data at all.

4. Discussion

When modeling the experimental data on impedance of

metal/Alq3 /ITO diodes by means of the equivalent circuit

(Fig. 4b) we have tacitly assumed that there are several spa-

tially separated zones throughout the device. Nevertheless, there

is an urgent need for a physical assignment of the equivalent,

frequency independent elements. After doing so one is able to

discuss the mechanism of the SI→ I transition in the fractional,

more or less conducting branch in Alq3 thin films. We have

considered the situation predicted by Baldo and Forrest [31] to

apply to theconducting phase represented viathe serial R, aC cir-

cuit interfaced with Ag or Al electrodes. Baldo and Forrest [31]

detected interface dipoles amounting to ∆≈1 eV, oriented in a




Fig. 12. Energy band diagram of a metal/Alq3 interface in the semi-insulating

(SI: top) and insulating (I: bottom) state (after Baldo and Forrest [31]). Electron

injectionto LUMOis promotedby theGaussiandensity of interfacestates(DOS)

in both cases, the I-state being the result of a damage to the true interface dipole

∆, caused possibly by injection of holes under positive biases on metal.

direction suitable for pushing down the energy bands of organicsemiconductor with respect to the vacuum level of metal, a sit-

uation illustrated by the energy diagram depicted at the top of

Fig. 12. Here, injection of electrons to the transport band of Alq3

(LUMO) is promoted by the Gaussian band (DOS) of interface

states. Keeping in mind that the fractional capacitance aC is

independent of bias and in parallel with the completely insu-

lating phase capacitance (1−a)C provides the full geometrical

capacitance C of the organic layer, we suggest that the resistor R

stands for the contact resistance of the interface region occupied

by the interfacedipole. A thorough study on the bias dependence

of R might contribute to an advanced model of charge injection

through the interface dipole region.

Now we are ready to handle this region in terms of the localorder model developed by Baldo et al. [29] and applied success-

fully to doped Alq3. For achieving the local order a strongly

polar material should be embedded in an otherwise undoped

polar Alq3, the dipole moment of a meridional Alq3 molecule

lying between 4.1 Debye [30] and 5.3 Debye [29]. The facial

phase molecule of Alq3 shouldpossess a largerdipole momentof

7.1 Debye [30], yet its close-to-surface content in Alq3 deposited

by UHV thermal deposition is probably negligible. Then the

crucial role is possibly played by the true interface dipole pos-

sessing a giant dipole moment of 30 Debye as deduced from

I –V curves by Baldo and Forrest [31]. If this picture is realistic,

it would lead to hole injection from the metal to the interface

region as the possible mechanism of destroying the interface

dipole through neutralizing its anion constituent. As a result of

this event the barrier height for electron injection (and extrac-

tion?) at the metal/Alq3 side of the device would increase by

∆≈1 eV (bottom of Fig. 12), thereby decreasing the injection

probability by a factor of

exp

∆

kT

≈ 6× 1016. (4)

Such a strong injection reduction would make the interface

completely insulating, a transition compatible with that shown

in Fig. 3 for the dc current. This is likely the case of the ac

current in the I-state, too, when looking at the missing low-

frequency data or at the persistent fractional capacitance C s.

However, the lone 103 Hz data point measured in the I-state at

zero bias might indicate different transport mechanisms for dc

and ac currents, e.g. through LUMO (extended) and the DOS

(via hopping) states, respectively.

At first sight any confrontation of I –V data with impedanceones expressed in terms of Model II appears impossible, since

the model does not explicitly provide any path for dc current.

Yet, a comparison of respective data on R and Rdc from tables

shown above suggests a correlation between the two quantities.

In our opinion, there is an implicit dc-transport path mediated

by the equivalent parallel resistance Rp (not shown) of the sub-

circuit of the Model I. For this configuration, the equality of C pand C s is documented (Fig. 7). Moreover, our experimental data

provide evidence for the equality of Rp and Rs as well. This is

easily understood if we define for both serial and parallel con-

figurations a unique time constant τ = RpC p = RsC s for Model I.

When searching for the transport mechanism in the dc limit, itis convenient to recall the results of I –V measurements, in par-

ticular the values of the exponent n in the power-law expression

given above. With reference to the standard knowledge [1], the

branches of I –V curves with n≈ 2 reflect the high-quality of the

organic material carrying space-charge limited currents (SCLC).

There have been a few cases when n approached values larger

than 3, a possible indication of the case of trap-limited current

flow [1].

Turning back to Model II, the R2, C 2 couple is tentatively

located at the bottom Alq3 /ITO interface. Its stability within the

bias range from −2 to 2 V is not surprising if one imagines the

relatively small portion of the dc bias absorbed by R2 R. Still

an explanation of having this almost ohmic back contact remainsa riddle. It remains to go back to the I-state represented by the

portion of the dashed j–U curve from Fig. 3 exhibiting the dis-

persion of the residual dc current, the inflexion point lying close

to U =−1.5 V. We consider the residual current to include a con-

siderable ionic component well known from electrochemistry as

“voltammetric wave”. Its origin resides in a redox reaction of

mobile ionic species at the metal electrode. For voltammetric

waves in solid electrolytes we refer to the review paper on solid-

state voltammetry by Kulesza and Cox [33]. In our case, we tend

to assign the wave to at least partially mobile cations (originally

participating in the interface dipole) reduced at the Al electrode,

mentioning again the absence of the current wave when using




Ag electrodes on top. One cannot exclude ionic species from the

surrounding atmosphere from playing a role in this slow charge

relaxation. This may be the right moment to touch the kinetics

of the equilibration of in situ Alq3 with atmosphere. The semi-

conducting or semiinsulating states stabilize on a time scale of

minutes after taking out devices from UHV. The bias-induced

I-state can be maintained for days, at least when the bias does

not exceed the interval used during the experiments presented.

Once the I-state has been reached, it is impossible to restore the

(non-ohmic) S- or SI-state without destroying the device.

A confrontation of present results with directly related pub-

lished work [27,28] deserves attention. Even though related

steady-state I –V curves of either PPV- or Alq3-based devices

were not shown in the publications, the polarity of bias may

correspond to the forward direction in each case. The data corre-

spond rather to admittance measurements (R−1p , C p) at different

frequencies and biases and exhibit pronounced dependence of

both components on frequency at a given bias. Despite quite dif-

ferent organic materials the published zero-bias data exhibit a

common distinct feature not compatible with our present results:the relative permittivities εr, as deduced from the overall high-

frequency capacitances, correspond to 2 (PPV) and 4 (Alq3),

respectively, while always showing a continuous increase when

lowering the frequency. The zero-bias conductanceG(ω) = R−1p

reported by Berleb andBrutting [27] f or E ≈105 V cm−1 may be

viewed as signaling the I-state. Under applied bias an apparent

negative contribution to capacitance is reported and assigned to

the relaxation of the injected space charge [27,28]. When pass-

ing to low frequencies, there is an apparent minimum (due to the

negative component) followed by a steep increase of the capaci-

tance, the minimum resulting from the interference between the

possibly negative component and the increasing dielectric lossdue to dc current. Berleb and Brutting [27] accepted the ear-

lier suggested analysis of the negative component [28] in terms

of a distribution of times-of-flight of injected excess carriers,

classifying the steep increase of capacitance as an indication

of dispersive transport, to be characterized by the dispersion

parameter ␣. Our results corresponding to even lower electric

fields are contradictory in the sense that one deals with an instru-

mental artifact due to excess conduction loss of whatever origin.

Moreover, our results and analysis presented above can be dealt

with an equivalent circuit composed of frequency independent

elements and εr≈ 2, the latter permittivity being in accordance

with the results on PPV by Martens et al. [28]. Nonetheless, even

some of their low-frequency admittance data are suspected of becoming inaccurate due to dc leakage of the diodes.

5. Conclusions

Evidence is provided for lateral inhomogeneity in the mor-

phology of Alq3 manifesting itself as a phase separation

deduced from the current–voltage and impedance behavior of

metal/Alq3 /ITO devices prepared in situ by thermal evapo-

ration and measured in air. Two phases seem to coexist in

the Alq3 thin films: an (instrumentally) completely insulating

(amorphous) phase along with a more or less conducting phase

building an interface dipole with the top metal (Ag, Al) elec-

trode. Such a coexistence is revealed by modeling the experi-

mentally observed fractional geometrical capacitance (1− a)C

(a = 0.5–0.7) of the devices as one component of the equiva-

lent circuit composed of frequency independent elements. The

interface dipole zone is represented by a bias-dependent contact

resistance in series with the complementary geometrical capac-

itance aC of the insulating bulk of Alq3. There seems to be a

correlation between this ac contact resistance and the dc resis-

tance as determined from I –V measurements. The dc current

obeys a power-law dependence ∝ U n on bias U , the coeffi-

cient n covering the range from 2 to 3.5 (space-charge to trap-

limited currents). The local formation of the zone is interpreted

within the framework of the metal/Alq3 energy band diagram

developed by Baldo and Forrest [31], along with the model of

dipole–dipole interaction leading to a local order (Baldo et al.

[29]). It is suggested that the local order is due to an interac-

tion of the giant interface dipole (30 Debye) with the intrinsic

dipoles of Alq3 molecules. The semiconducting phase under-

goes an irreversible semiinsulator-to-insulator transition undera critical positive potential of the top metal electrode. This tran-

sition is tentatively interpreted as coming from hole injection

and subsequent neutralization of the anion of the giant interface

dipole, thereby destroying the latter. The Alq3 /ITO barrier lead-

ing to an observable dispersion of the real part of impedance

at high frequencies is modeled via a linear parallel RC circuit

independent of the related absorbed portion of applied dc bias.

High-sensitivity impedance instrumentationis needed to acquire

low-frequency data on insulating Alq3, thereby enabling a com-

parison of transport mechanisms in frequency- and time-domain

(dc limit).

References

[1] W. Brutting, S. Berleb, A.G. Muckl, Org. Electron. 2 (2001) 1.

[2] J.C. Scott, J. Vac. Sci. Technol. A 21 (2003) 521.

[3] P.E. Burrows, Z. Shen, V. Bulovic, D.M. McCarty, S.R. Forrest, J. Appl.

Phys. 79 (1996) 7991.

[4] A. Curioni, M. Boero, W. Androni, Chem. Phys. Lett. 294 (1998) 263.

[5] M. Colle, R.E. Dinnebier, W. Brutting, Chem. Commun. (2002) 2908.

[6] M. Colle, S. Forero-Lenger, J. Gmeiner, W. Brutting, Phys. Chem. Chem.

Phys. 5 (2003) 2958.

[7] M. Colle, J. Gmeiner, W. Milius, H. Hillebrecht, W. Brutting, Adv. Funct.

Mater. 13 (2003) 108.

[8] M. Colle, W. Brutting, Phys. Status Solidi (a) 201 (2004) 1095.

[9] R.L. Martin, J.D. Kress, I.H. Cmpbell, D.L. Smith, Phys. Rev. B 61 (2000)15804.

[10] H. Aziz, Z.D. Popovic, N.-X. Hu, A.-M. Hor, G. Xu, Science 283 (1999)

1900.

[11] D.Y. Kondakov, J.R. Sandifer, C.W. Tang, R.H. Young, J. Appl. Phys. 93

(2003) 1108.

[12] D.Y. Kondakov, J. Appl. Phys. 97 (2005), 024503.

[13] F. Papadimitrakopoulos, X.-M. Zhang, Synth. Met. 85 (1997) 1221.

[14] S.H. Kim, H.Y. Chu, T. Zyung, Y.S. Yang, ETRI J. 26 (2004) 161.

[15] M.S. Xu, J.B. Xu, Thin Solid Films 491 (2005) 317.

[16] L.F. Cheng, L.S. Liao, W.Y. Lai, X.H. Sun, N.B. Wong, C.S. Lee, S.T. Lee,

Chem. Phys. Lett. 319 (2000).

[17] H. Mu, H. Shen, D. Klotzkin, Solid State Electron. 48 (2004) 2085.

[18] C.B. Lee, A. Uddin, X. Hu, T.G. Andersson, Mater. Sci. Eng. B 112 (2004)

14.

[19] I.H. Campbell, D.L. Smith, J.P. Ferraris, Appl. Phys. Lett. 66 (1995) 3030.




[20] Y. Li, J. Gao, G. Yu, Y. Cao, A.J. Heeger, Chem. Phys. Lett. 287 (1998) 83.

[21] G. Yu, Y. Cao, Ch. Zhang, Y. Li, J. Gao, A.J. Heeger, Appl. Phys. Lett. 73

(1998) 111.

[22] W. Brutting, H. Riel, T. Beierlein, W. Riess, J. Appl. Phys. 89 (2001) 1704.

[23] G. Paasch, S. Scheinert, Synth. Met. 122 (2001) 145.

[24] G.L. Pakhomov, V.I. Shashkin, D.E. Pozdnyaev, C. Muller, J.-M. Ribo,

Org. Electron. 3 (2002) 93.

[25] Y.S. Yeong, D. Troadec,A. Moliton,B. Ratier, R. Antony, G. Veriot, Synth.

Met. 127 (2002) 195.[26] P. Stallinga, H.L. Gomes, M. Murgia, K. Mullen, Org. Electron. 3 (2002)

43.

[27] S. Berleb, W. Brutting, Phys. Rev. Lett. 89 (2002) 286601.

[28] H.C.F. Martens, H.B. Brom, P.W.M. Blom, Phys. Rev. B 60 (1999)

R8489.

[29] M.A. Baldo, Z.G. Soos, S.R. Forrest, Chem. Phys. Lett. 347 (2001)

297.

[30] H. Ishii, N. Hayashi, E. Ito, Y. Washizu, K. Sugi, Y. Kimura, M. Nivano,

Y. Ouchi, K. Seki, Phys. Status Solidi (a) 201 (2004) 1075.

[31] M.A. Baldo, S.R. Forrest, Phys. Rev. B 64 (2001) 085201.

[32] S.A. Van Slyke, C.H. Chen, C.W. Tang, Appl. Phys. Lett. 69 (1996)2160.

[33] P.J. Kulesza, J.A. Cox, Electroanalysis 10 (1998) 73.

inhomogeneous transport of alq3

Documents