Chemical Physics 337 (2007) 151160

www.elsevier.com/locate/chemphys

Electromodulation of photoluminescence in vacuum-evaporated

lms of fac-tris(2-phenylpyridine)iridium(III)
Waldemar Stampor *, Jakub Mez_ yk
Department of Molecular Physics, Gdansk University of Technology, Narutowicza 11/12, 80-952 Gdansk, Poland

Received 16 May 2007; accepted 10 July 2007

Available online 20 July 2007

Abstract

Electromodulation of photoluminescence (EML) has been investigated in vacuum-evaporated lms of Ir(ppy)3 (tris(2-phenylpyri-
dine)iridium) commonly used as a phosphorescent emitter in organic light-emitting diodes. The quenching of photoluminescence up
to 30% is observed in the external electric eld applied to sandwich cells Al/Ir(ppy)3/Al/glass. The electric eld characteristics, the emis-
sion and excitation spectra of EML signals are interpreted in terms of eld-induced dissociation of excited states into free and/or trapped
charge carriers according to the Onsager theory. In addition, the existence of internal electric elds of (13) 105 V/cm in Ir(ppy)3 lms is
inferred from the (1x)EML signals measured at the rst harmonic of the applied electric eld frequency, in accordance with previous
electroabsorption results.
 2007 Elsevier B.V. All rights reserved.

Keywords: Electromodulation; Charge photogeneration; Onsager model; Iridium complex; OLEDs

1. Introduction processes responsible for the high eld drop in QE of

Cyclometalated iridium complexes have been recently
used in organic light-emitting diodes (OLEDs) to demon-
strate ecient electrophosphorescence with the near unity
internal quantum eciency (QE) [1,2]. Such high quantum
eciencies result from ecient intersystem crossing
because the iridium atom has a strong coupling between
its orbital angular momentum and its spin, which mixes
the singlet and triplet states of organic complex and thus
permits to harvest light from the all excitons both singlets
and triplets populated after an electron and hole combi-
nation in operating OLEDs.
Both uorescent and phosphorescent OLEDs typically
experience a pronounced decrease in the QE at large cur-
rent densities owing through devices subject to the high
electric eld [3]. There are three main exciton quenching
*
Corresponding author. Tel.: +48 583471568; fax: +48 583472821.
E-mail address: waldek@mif.pg.gda.pl (W. Stampor).
OLEDs: (i) excitonexciton interaction, (ii) excitoncharge
carrier interaction, and (iii) exciton dissociation. The
quenching mechanisms operating in iridium-based phos-
phor OLEDs are currently under intense debate [411].
The roll-o in QE observed in such devices was originally
attributed to triplettriplet annihilation [4,5]. However,
recent steady state [7] and transient [8,9] photolumines-
cence (PL) measurements on such materials at varying
intensities of the exciting light showed that excitonexciton
interaction is too weak to explain strong electrophospho-
rescence quenching (as high as 90% [4,9]) observed in Ir-
based phosphor devices. Instead, combinations of (i)(iii)
mechanisms have been employed to describe the quenching
eect. Thus, annihilation of excitons on charge carriers and
exciton dissociation were proposed to coexist in operating
OLEDs, with the former mechanism prevailing within the
lower-eld range [9,10]. Accordingly, the high eld
(>106 V/cm) quenching eects in electrophosphorescence
were attributed to the electric eld-enhanced dissociation
of electronhole pairs prior to exciton formation [6,9,10].

0301-0104/$ - see front matter  2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.chemphys.2007.07.016
152 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160

In OLEDs with Ir(z)3-doped emitting layer (z = uore- 2. Experimental and numerical details
nyl-pyrazolato derivative) [9] the decrease in QE as a
function of applied electric eld strength was tted satisfac- 2.1. Material and samples
torily using the PooleFrenkel model of charge separation
[12]. In turn, in OLEDs with Ir(ppy)3 (ppy = phenylpyri- The fac-Ir(ppy)3 material (for molecular structure see
dine) embedded in TPD (diamine derivative) matrix, the Fig. 1a) was prepared and puried as described in Ref.
Onsager model of charge pair dissociation [13] appeared [18]. Here fac stands for facial arrangement (with C3 sym-
to be more appropriate [6,10]. On the contrary, eld- metry) of phenylpyridine ligands where the three pyridine
induced exciton dissociation was excluded as a relevant N atoms are located opposite the three phenyl C atoms
process in OLEDs comprising Ir(ppy)3:TCTA (TCTA = in a pseudo-octahedral complex. The fac stereoisomer
carbazolyltriphenylamine) or Ir(piq)3:NPB (piq = phenyl- being more stable energetically [19] than a less symmetrical
isoquinoline, NPB = diamine derivative) systems because meridional form prevails in vacuum-evaporated layers of
the PL transients recorded on reverse-biased OLED cells Ir(ppy)3 [20,21].
were only slightly inuenced by high external electric eld The neat lms of Ir(ppy)3 were deposited by thermal
[11]. Consequently, an unied model taking into account evaporation (215232 C) in vacuum (103 Pa) at a rate
both triplettriplet and tripletcharge quenching mecha- of 10 nm/min onto room temperature quartz substrates.
nisms [11] was applied to reproduce the QE decrease in The quartz plates were cleaned in solvents and ultrasoni-
such OLEDs, with the former mechanism dominating in cally in an isopropanol bath before loading into a vacuum
the high current density range. The magnitude of the high deposition chamber. The thickness of the organic lms was
eld (or high current density) imposed QE reduction measured with a Tencor Alpha Step 500 Proler.
observed in various OLEDs [111] changes from as small The EML results were obtained with 100400 nm thick
as about 20% to as high as above 90%. On the basis of this lms in the sandwich cells supplied with two vacuum-evap-
short survey we can safely conclude that contributions orated semi-transparent aluminium electrodes, Al/Ir(ppy)3
from particular quenching mechanisms (i)(iii) in dierent lm/Al/quartz substrate (Fig. 1b). Aluminium electrodes
OLEDs depend on individual characteristics of iridium with rather poor charge injection ability were chosen to
phosphor emitters (ligand type, neat or doped lm, dopant
concentration, matrix type).
In the present paper, we report on electromodulation
of photoluminescence (EML) in thin neat lms of
Ir(ppy)3. Applying low excitation intensities (<1015 pho-
tons/cm2 s) and weakly injecting electrodes allowed us to N
ignore excitonexciton and excitoncharge carrier interac-
tions, and thus to examine the electric eld-induced PL
quenching due to exciton dissociation alone. The EML Ir
is a well established method which enables to investigate N
N
in detail dissociation of excitons and charge separation
mechanism in organic solids (for the recent surveys of lit-
erature on this subject, see Refs. [1416]). The preliminary
EML results for Ir(ppy)3 lms have already been pre-
sented [6]. However, in our previous study only the
EML measurements in steady state (dc) electric eld were fac-Ir(ppy)3
carried out which restricted our data to a high electric
eld range. In the present paper, we study the quenching
eect applying a more sensitive ac electromodulation tech-
nique. In the (ac)EML experiments we could extend elec-
tric eld range and record the emission and excitation
spectra of EML signals. Comparing the (2x)EML emis-
sion spectrum for a Ir(ppy)3 lm with the recently
reported PL spectrum for a single crystal of Ir(ppy)3
[17] we discuss the possible involvement of crystal defect
states in exciton dissociation in Ir(ppy)3 lms. The all
observed EML results are satisfactorily rationalized on
the grounds of the Onsager model of charge separation.
Additionally, the analysis of the (1x)EML signals mea-
sured at the rst harmonic of the applied electric eld fre-
quency (x) enabled us to evaluate the internal electric Fig. 1. (a) The molecular structure of fac-Ir(ppy)3. (b) A schematic
elds built into Ir(ppy)3 lms. diagram of the sandwich cell arrangement used in EML measurements.
 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
eliminate electromodulation of photoluminescence due to
excitoncharge carrier interaction [6,10,14].
2.2. Electromodulation measurements
The photoluminescence was excited by a light beam
from a mercury (Narva, HBO 200W) or xenon (Osram,
XBO 450W) lamp followed by a SPM-2 Zeiss monochro-
mator. In the (ac)EML experiments the modulation of
photoluminescence was induced by a sinusoidal voltage,
U0 sin(xt), applied to the sample (typically x/2p =
175 Hz). The photoluminescence light was collected with
a quartz light guide followed by a monochromator
(SDMC1 Optometrics) and a set of appropriate cut-o
glass lters (Schott and Corning) and then detected by a
photomultiplier tube (EMI 9863QB). The ac PL signal con-
sisted of a steady-state (I0x) and alternating components
corresponding to the time-dependent PL intensity, I(t),
expanded in a Fourier series
It I 0x I 1x sinxt u1 I 2x sin2xt u2    1 PL intensity at the rst and second harmonics of the
The steady state signal, I0x, was measured with a dc digital
meter. The alternating signals, I1x(t) and I2x(t), were mea-
sured simultaneously by two lock-in ampliers (Princeton
Applied Research model 5210) referenced to the rst (1x)
and second (2x) harmonic of the applied external voltage
frequency, respectively. The quantities to be monitored
are dened here as
I 1x
1xEML ; 2
I 0x
I 2x
2xEML ; 3
I 0x
where I1x and I2x are the rst and second order Fourier
components (amplitudes) of the photoluminescence inten-
sity. The EML signals were stored, averaged and processed
by a computer on-line. The phase shift between the applied
voltage and the second harmonic response, I2x(t), was
found to be u2 = 90 10, which means photolumines-
cence quenching in a sense that the increasing electric eld
diminishes PL intensity. The emission spectra of I2x signals
were corrected for photomultiplier and monochromator
sensitivity. The phase shift (u1) between the modulating
voltage and the rst harmonic response, I1x(t), was equal,
within 10 error, to 0180, dependent on the sample po-
sition. When the sandwich cell was reversed to the opposite
position with respect to the direction of applied modulating
eld direction, the phase of (1x)EML signal changed by
180. The phase shifts of the EML signals were indepen-
dent of the amplitude of external voltage applied to the
sample. The EML signals (amplitudes and phases) did
not depend on the applied eld frequency within the acces-
sible in experiment frequency range (505000 Hz). The sen-
sitivity of the (ac)EML experimental setup was better than
104.
In the dc EML experiments the photoluminescence
quenching eciency,
153
DI IU 0  IU
dcEML  ; 4
I0 IU 0
was calculated from PL signals measured without (I0) and
with a dc voltage (U) applied to the sample (I). To reach
the reproducibility better than 10%, all PL intensities were
averaged during 60 s. In such experiments the sensitivity of
the setup was 102.
Electric eld-induced change in the number of photons
absorbed in Ir(ppy)3 lms (due to electroabsorption) is at
least two orders of magnitude smaller than the EML sig-
nals for all applied excitation light wavelengths and electric
eld strengths (confer Ref. [20]).
All measurements were carried out at room tempera-
ture. Further details of the experimental setup and electro-
modulation method are described elsewhere [14,22].
2.3. Numerical procedures
Applying sinusoidal electromodulation we measure the
applied electric eld frequency. The recorded (1x)EML
and (2x)EML signals (dened by Eqs. (2) and (3)) are com-
pared with those calculated on the basis of various theoret-
ical models as considered in Sections 3.2 and 3.3. The Inx
Fourier components:
Z T
x
I 0x IF t dt; 5a
2p 0
Z T
x
I 1x IF t sinxt dt; 5b
p 0
Z T
x
I 2x IF t cos2xt dt; 5c
p 0
are calculated numerically using a FORIF SSP procedure.
We note that for sinusoidal modulating eld, F(t) =
F0 sin(xt), according to Eq. (5b) the I1x Fourier component
disappears [14]. Therefore, any non-zero (1x)EML signals
detected at the rst harmonic mean the presence of an
internal electric eld (Fi) in the sample. Then the eective
modulating eld can be expressed as
F t F i F 0 sinxt: 6
The value of Fi can be evaluated from the magnitude of
(1x)EML signals as described in Section 3.3. It is also
worth to note that the (2x)EML signals are independent
of Fi for Fi  F0. The relation Fi  F0 is well fullled in
typical high-eld EML experiments on sandwich Al/organ-
ic lm/Al cells [22]. The analysis of (2x)EML eect, free of
internal elds Fi, allows to test the various theoretical EML
models and thus to quantify the PL quenching mechanism
as discussed in Section 3.2.
3. Results and discussion
3.1. Absorption and photoluminescence spectra
The absorption spectrum of a neat Ir(ppy)3 lm (thick
solid line in Fig. 2) is nearly the same as that for Ir(ppy)3
154 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
PL
ABS
film
crystal
Intensity [ a. u. ]
EML
I2
EML
I2/I0
300 400 500
[ nm ]
Fig. 2. The excitation (squares) and emission (circles) spectra of the EML
signals for a 310 nm thick vacuum-evaporated lm of Ir(ppy)3. The rms
value of external electric eld Frms = 1.0 106 V/cm. The EML excitation
spectrum was recorded with emission light wavelength kem = 525 nm. The
EML emission spectrum was measured with excitation light wavelength
kexc = 365 nm. The absorption and photoluminescence spectra of vacuum-
evaporated lms are shown with solid lines. For comparison, the
photoluminescence spectrum of a Ir(ppy)3 single crystal taken from Ref.
[17] is displayed with the dotted line.
in solution as reported previously [20,21,23]. An intense
ultraviolet band in the range 250320 nm is assigned typi-
cally to the singletsinglet ligand-centred, pp* transition
(S0 ! 1LC). In turn, a weaker absorption band extending
into the visible region (320510 nm) is attributed to
metal-to-ligand-charge-transfer transitions (MLCT) [8,23
28]. More intense features with a main maximum at
385 nm are conventionally associated with singletsinglet,
dp* (S0 ! 1MLCT) transitions, and weakly manifesting
shoulders in the longer-wavelength range (450510 nm)
with singlettriplet, dp* (S0 ! 3MLCT) transitions and
their vibrational satellites. The transitions to 3MLCT states
being formally spin-forbidden acquire intensities by an
eective mixing with higher energy spin-allowed transitions
due to strong spinorbit coupling of iridium atom [29]. The
TD DFT (time-dependent density functional theory) calcu-
lations by Hay [19], neglecting spinorbit coupling,
revealed in the MLCT absorption region several pure
singlet and triplet transitions involving metal orbitals
(5d-Ir) signicantly mixed with p ligand orbitals. The two
lowest energy MLCT states are located at 492 nm
(2.52 eV) and 453 nm (2.74 eV) on the basis of electroab-
sorption measurements [20]. The rst state at 2.52 eV is cer-
tainly the lowest triplet state, T1, and the energy of the
second state (2.74 eV) corresponds well with the energy
of the singlet state, S1, calculated by the TD DFT method
[19,30]. Thus, the singlettriplet splitting energy in Ir(ppy)3,
DE(S1T1), can be evaluated as low as 0.22 eV (1800 cm1)
which is in good accordance not only with TD DFT results
(0.21 eV [19] and 0.25 eV [30]) and the evaluation based on
absorption data (0.19 eV [26]) but also with an ordering
scheme for organo-transition-metal compounds proposed
recently by Yersin [31]. For comparison, the magnitude
of DE(S1T1) for organic molecules involving only pp*
transitions is typically greater than 1 eV. A signicant
reduction of S1T1 splitting in organo-transition-metal
complexes is due to a large metal orbital contribution into
the relevant electronic wavefunctions [31]. An iridium core
of organic complex acts as an ecient electronic coupler
which extends the excitation over three ligands and induces
a rather large (on average) spatial separation between inter-
acting electrons. As a result strongly delocalized MLCT
excited states, containing charge transfer (dp*) and neutral
(pp*) congurations, with considerably reduced electron
electron exchange interaction are created. It is worth to
600 note here that due to a high degree of spinorbit interac-
tion, a spin labelling for MLCT states (singlet or triplet)
is often no longer adequate in such complexes [32]. This
is fully conrmed by the TD DFT calculations [33] where
including spinorbit coupling of 5d electrons as a perturba-
tion a strong mixing between 1MLCT and 3MLCT cong-
urations in Ir(ppy)3 has been demonstrated.
The photoluminescence spectrum of a 50 nm thick evap-
orated Ir(ppy)3 lm (thin solid line in Fig. 2) consists of a
broad band with its intensity maximum position (at
523 nm), red-shifted by several nm in comparison to that
measured in various solvents [18,2426]. The observed PL
emission is commonly identied as phosphorescence from
the 3MLCT state. This assignment originally based on PL
lifetime measurements and solvent dependence of PL spec-
tra [18] was later conrmed by quantum chemical calcula-
tions [19,30,33] and extensive experimental works [8,24
26,3436] including: (i) the temperature dependence of
the emission band structure [34], (ii) the rigidochromic
(blue) band shift of PL spectra on going from a room tem-
perature uid solution to a low temperature glass [24,26],
(iii) the Stokes shift and band structure analysis for
Ir(ppy)3 placed in Shpolskii crystals [35], (iv) the high mag-
netic eld eects on low temperature PL spectra [34] and
(v) the cyclic voltammetry measurements of reduction/oxi-
dation potentials [25,36]. Additionally, at low temperatures
a weak higher-energy PL band was observed (at 455 nm for
Ir(ppy)3 dispersed in PMMA [24] or at 396 nm for neat
lms [28]) which was attributed to emission from a 3LC
[24] or 1MLCT state [28]. All experimental data conrm
that intersystem-crossing in Ir(ppy)3 proceeds fast with
near unity eciency [28,38] which is consistent with the
recent femtosecond time-resolved absorption measure-
ments [37] and quantum mechanical calculations showing
very high density of spin-mixed states (about 70 states
within the lowest 1 eV) [33].
Maybe the most complete picture of the Ir(ppy)3 emis-
sive state has been emerged from a comprehensive study
by Finkenzeller and Yersin [34] where PL spectra and
decay times of Ir(ppy)3 dissolved in THF (tetrahydrofuran)
were registered in a wide temperature range and high mag-
netic elds. From the data analysis three PL active compo-
nents of the lowest 3MLCT state were identied revealing
the large value of the zero-eld-splitting (83 cm1) which
 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
is a consequence of strong spinorbit interaction intro-
duced by the Ir-5d electrons into the corresponding triplet
wavefunctions. The transition from the lowest triplet sub-
level to the ground state being strictly forbidden gains the
oscillator strength by coupling with HertzbergTeller
active vibrations. In turn, the spin-allowed transitions from
the two higher energy substates show a usual FranckCon-
don activity involving totally symmetric vibrations of the
ground state. At ambient temperature, the individual prop-
erties of triplet substates are smeared out and one nds
only an averaged behavior in a broad and rather featureless
PL band (Fig. 2).
The solid state eect on photoluminescence of neat
Ir(ppy)3 lms prepared by evaporation appears mainly in
considerable reduction of the PL quantum eciency and
additional inhomogeneous broadening of the PL spectrum
[8,21]. In comparison to measurements in dilute solutions
where nearly a single decay time of 12 ls is observed
[18,21], the PL decay in neat lms is about one order of
magnitude faster and non-monoexponential [21,39,40],
which indicates at additional non-radiative decay channels
in the solid state. It is regarded that intermolecular interac-
tions dictated by molecular packing in the solid state are
responsible for the observed quenching eects. The exis-
tence of weakly emissive luminescence quenchers such as
excimers, ground state aggregates and other energy traps
in Ir(ppy)3 lms was proposed [7,8,21,40]. The solid-state
self-quenching processes controlled by dipoledipole inter-
actions similar to those occurring in the Forster energy
transfer model were also considered in this context [41].
Due to diculty of growing quality Ir(ppy)3 crystals of
sucient size, only recently the full details of fac-Ir(ppy)3
crystal structure have been reported by Breu and cowork-
ers [17]. For our further reasoning it is useful to summarize
here the main conclusions of that work. The space group of
a fac-Ir(ppy)3 crystal turns out to be trigonal centrosym-
metric, P 3. Despite a very corrugated molecular shape
imposed by a propeller-like structure of octahedron com-
plex, Ir(ppy)3 molecules interpenetrate eectively in the
crystal lattice to attain a close packing controlled by inter-
molecular dispersive (London) and electrostatic (coulom-
bic) forces. The six molecules of the unit cell are stacked
rigorously in three columns running along c with the C3
molecular axis collinear with the c crystal axis. In the two
homochiral columns at 1/3, 2/3, z and 2/3, 1/3, z (Wyckho
positions, 2b and 2c, respectively) molecules with a
xed chirality, K and D, respectively, have permanent
dipole moments (parallel to a C3 molecular axis [19,20])
arranged in a head-to-tail position. In the third racemic
column at 0, 0, z (Wyckho position 2a) molecules alter-
nate their chirality successively and realize a tail-to-tail
arrangement of dipole moments. The IrIr distance within
each column is 8.402 A. From an alternative point of view,
the molecules 2b and 2c being at the same height 3.2. (2x)EML results
along c form two honeycomb-like hexagonal layers in the
ab plane at z = 0 and z = 1/2, respectively. The IrIr dis-
tance between adjacent molecules in the hexagonal layers
155
is relatively large (9.709 A) and the resulting cavity in the
center of the honeycomb is lled closely by two molecules,
at z = 1/4 and z = 3/4, respectively, belonging to the third
(racemic) column. The planar ppy ligands of adjacent mol-
ecules within the racemic columns are engaged in T-shaped
(edge-to-face) motifs. Although the described above mole-
cular arrangements minimize the total interaction energy of
the ideal lattice, the real crystals are liable to systematic
twinning (by merohedry) which interchanges the symmetry
related molecules in the racemic column and thus disrupts
the rigorous alternation of molecules along c. Conse-
quently, at twin boundaries a tail-to-tail alignment of
dipoles is converted into a more favorable head-to-tail
one and adjacent ppy ligands are no longer engaged in a
T-shaped but rather in a parallel displaced geometry (oset
face-to-face) with a shifted p-stack involving half of a
ligand. Such rearrangements of molecular packing and
the existence of many distorted regions at boundaries of
thin lamellar domains stacked along c profoundly aects
the photophysical properties of Ir(ppy)3. According to
Breu and coworkers [17], this inherent intrinsic disorder
is not only a reason for a nonlinear second order activity
of nominally centrosymmetric Ir(ppy)3 crystals, but also
has a pronounced eect on photoluminescence spectra.
The PL spectrum of a fac-Ir(ppy)3 single crystal (dotted
line in Fig. 2) is characterized by the appearance of a dis-
tinct shoulder (at 507 nm) at the shorter-wavelength side
of the main band (at 545 nm) which is not present in the
spectrum of the dissolved compound. While the main band
was assigned to the emission from regular domains, the
shorter-wavelength shoulder with its intensity varying
strongly from crystal to crystal was ascribed to the twinned
domains boundaries. The specic intermolecular interac-
tions occurring within the bulk and interface regions are
also displayed in distinctively dierent pressure-induced
shifts of either emission bands under high pressure applied
to crystal Ir(ppy)3 samples [17].
Recent AFM studies [21] revealed in vacuum-evaporated
Ir(ppy)3 lms many microdomains of rod-like shape resem-
bling the natural shape of Ir(ppy)3 single crystals [17]. This
fact strongly suggests that Ir(ppy)3 lms exhibit the poly-
crystalline character. It seems to be therefore reasonable
to compare PL spectra of Ir(ppy)3 lms with those for single
crystals. As can be seen in Fig. 2, the PL spectrum of a thin
Ir(ppy)3 lm exhibits a single inhomogeneously broadened
asymmetric band without any well-resolved features. The
PL intensity maximum is observed at the wavelength which
falls between the positions of the main peak and the shoul-
der in the PL crystal spectrum. A further insight into the
nature of emissive electronic states in solid Ir(ppy)3 can be
provided by investigations of the high electric eld eect
on PL spectra of thin Ir(ppy)3 lms.
The plot of the PL intensity detected at the second har-
monic of the electromodulating eld (I2x) as a function of
156 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
emission wavelength is shown by circles in Fig. 2. This emis-
sion EML spectrum for a Ir(ppy)3 lm was obtained at
kexc = 365 nm excitation with the rms electric eld strength
of 1.0 106 V/cm. For all wavelengths the PL intensity is
reduced by the applied electric eld. The EML spectrum
has its maximum red-shifted by about 10 nm and is notice-
ably sharper relative to the ordinary PL spectrum of a lm.
The peak position of the EML spectrum coincides well with
the position of a shoulder in the PL spectrum of a crystal. In
addition, from the EML data we can deduce that the PL
quenching eciency (DI/I) within the shorter-wavelength
spectral region being approximately constant (5% at the
electric eld of 106 V/cm) is several times larger in compar-
ison to that measured at the longer-wavelength part. This
means that some higher-energy emitting species, although
buried under the inhomogeneously broadened ordinary
PL band, are strongly aected by the electric eld and give
then the relatively large EML signals. We propose here that
the observed EML eect should be in large attributed to the
molecular aggregates similar to those responsible for the
higher-energy shoulder detected in the PL spectrum of a sin-
gle crystal (see Section 3.1). Further arguments for this
assignment are given by a quantitative analysis of the elec-
tric eld dependence of EML signals as will be discussed
below.
The square points in Fig. 2 show electric eld-induced
photoluminescence quenching eciency (I2x/I0x) as a func-
tion of the excitation wavelength (kexc). The PL quenching
gradually increases with the excitation photon energy
within the rst lower-energy (MLCT) absorption band in
Ir(ppy)3. A sharper increase appears in the spectral range
of the second absorption maximum connected with the
higher-energy (1LC) exciton states.
The electric eld dependence of PL quenching eciency
(calculated according to Eqs. (3) and (4)) measured at
kem = 525 nm and under excitation within the MLCT
absorption band is depicted in Fig. 3. The EML data
obtained by the dc and ac methods are shown for samples
with two dierent thickness (d = 150 nm and d = 310 nm).
Due to a poor sensitivity, the dc method could be applied
for electric eld intensities higher than 106 V/cm. As can
be seen, the values of the (2x)EML and (dc)EML signals
are in good agreement with each other and independent
of the sample thickness. The electroabsorption signals
(2x)EA measured at the same wavelength of absorbed light
as the excitation wavelength used in the EML experiments
are shown for comparison. Since the EA signals are about
two orders of magnitude smaller than the EML signals, the
electroabsorption contribution to the electromodulation of
photoluminescence can be safely neglected. The EML plot
in Fig. 3 distinctively departs from a second-order function
of the applied electric eld and a trend to saturation is
observed in the high eld range. The maximum quenching
of photoluminescence at an electric eld of 3 106 V/cm
reaches 30%.
The features of EML in Ir(ppy)3 cannot be rationalized
in the framework of the Stark eect where energy levels
100 Ir(ppy)3
10-1 em=525 nm
EML
/ 0
Frenkel
10-2 model
0
EA
2
Onsager
10-3
model
slope 2.0
10-4
105 106
F [ V/cm ]
Fig. 3. The relative PL intensity change (calculated according to Eqs. (3)
and (4)) as a function of the electric eld strength for Ir(ppy)3 lms
displayed in the loglog scale. The (2x)EML data (squares) and (dc)EML
data (circles) for lms with thickness d = 150 nm (open symbols) and
d = 310 nm (lled symbols) are shown. For (2x)EML data the abscissa F
stands for the rms value of the electric eld. The excitation and emission
light wavelengths are kexc = 405 nm or 436 nm, and kem = 525 nm. The
solid line is a plot representing the best t based on the 3D-Onsager model
with r0 = 15.5 A and g0 = 0.9. The dashed line is obtained according to the
PooleFrenkel model with er = 3.5 and APF = 107 (for denition of APF,
see Ref. [22]). The eld dependence of the electroabsorption signal (EA)
for k = 436 nm light passing through the sample (d = 310 nm) is shown by
the dotted line. The EA signal increases exactly with the square of the
applied electric eld, in accordance with the Stark eect.
shift or redistribution of oscillator strength between dier-
ent states induces the exact square-type eld dependence of
EML signals. The pronounced Zeeman interaction between
the substates of the lowest Ir(ppy)3 triplet state leading to
changes in PL intensities has been recently observed under
application of high magnetic elds at 1.5 K [34]. In princi-
ple, the Stark analogue of the mechanism proposed in Ref.
[34] is conceivable to operate in Ir(ppy)3 at low tempera-
tures and high electric elds. However, at room tempera-
ture a fast thermalization within the triplet manifold
occurs [34] and the photoluminescence is then governed
by the average behaviour of the equilibrated substates.
Besides, positive and negative signals should be typically
observed in the Stark-type EML spectra with intensities
scaled with F2 (see for example Ref. [42]) which is not the
case.
The tripletcharge carrier interaction can be also safely
ruled out as the process responsible for the strong PL
quenching observed in Al/Ir(ppy)3/Al system. As we
argued previously [6,10], the Al electrodes have only
weakly charge-injecting properties due to a relatively large
energy mismatch (above 1 eV, see Ref. [6] for details)
between the Al Fermi level and the relevant energy levels
of Ir(ppy)3 frontier orbitals.
Consequently, the eld-assisted dissociation of excitons
into free and/or trapped charge carriers is the most proba-
 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
ble process responsible for the photoluminescence quench-
ing eect observed in the solid Ir(ppy)3. To describe the
exciton dissociation, the three-dimensional Onsager model
of geminate recombination [13] has been applied. In this
model, dissociation of excitons is a two-step process. After
excitation charge carriers are rst separated to a certain
distance r0 creating a geminate eh pair (CT state). In a sec-
ond step, the charge carriers start a Brownian random walk
(in an isotropic continuous medium) under the combined
action of Coulomb and external electric elds. Assuming
that geminate pairs with a single thermalization distance
r0 are formed, the escape probability can be expressed as
follows [13,43]:
X1
XOns F 1  n1 P m rC =r0 P m n; 7 iment was attained with g0 = 0.9 0.1 and r0 =
m1
where Pm(x) is the incomplete gamma function of integral
order m and
er0 jF j
n :
kT
In formula (7), rC represents the coulombic capture radius
(Onsager radius) at which the Coulomb attraction energy is
equal numerically to the thermal energy, kT. In more
sophisticated approaches, some distribution of r0 distances
should be taken into account [44]. Since an electronhole
pair dissociation competes with its geminate recombination
into the emissive state, the PL intensity is expected to be a
function of the electric eld,
kf
IF 1  g0 XOns F I  ;
kf kn
where kf and kn are the rate constants for radiative and
non-radiative decay pathways of emitting states, respec-
tively, g0 is the yield of geminate pair formation, and I*
is the production rate of hot excitons. Except for the
escape probability (X) all parameters in formula (9) are
assumed to be electric eld independent. This also includes
the primary quantum yield g0, however, in some dierent
from Onsager approaches the eld dependence of this
quantity is also taken into account (see for example Refs.
[16,22]). The eld independence of g0 is in accordance with
the conventional formulations of the Onsager theory in
organic molecular solids [44]. It is assumed there that an
eh pair is formed indirectly by an autoionization of a
FranckCondon molecular state or directly via an intermo-
lecular charge-transfer (CT) transition where the both pro-
cesses do not require any electric eld assistance. This is
essentially a case of the present work because the EML re-
sults for Ir(ppy)3 were obtained with the excitation light
having photon energy at least 0.5 eV above the energy of
the lowest excitonic transition in this material. However,
since the fast intersystem crossing (S* ! T*) is considered
to take place in Ir(ppy)3 after photoexcitation [37], the vib-
rationally hot triplet (T*) rather than singlet (S*) states
should be regarded as precursors of geminate eh pairs
on the grounds of the Onsager model of exciton dissocia-
157
tion. On the other hand, this distinction may be less impor-
tant for MLCT states in iridium complexes where due to
strong mixing between 1MLCT and 3MLCT states the term
intersystem crossing has no well-dened meaning for
such states [32].
The Onsager model provides a good framework for a
quantitative description of the EML eect in Ir(ppy)3 lms.
The calculated values of I2x/I0x have been tted to the
measured (2x)EML signals using the numerical procedures
as described in Section 2.3 and Ref. [45]. In the tting pro-
cedure, the primary quantum yield g0 and initial separation
distance r0 were adjusted assuming the relative permittivity
er = 3. For the EML results displayed in Fig. 3 the good
agreement between theory (solid line in Fig. 3) and exper-
15.5 0.5 A. The obtained value of r0 is close to the twice
intermolecular distance in the columnar stack along c axis
in fac-Ir(ppy)3 crystal structure (16.8 A [17]) indicating that
the primary acts of exciton dissociation in Ir(ppy)3 take
place within the c columns rather than the ab planes. The
8
rationale for this fact is not only the smaller distance
between the adjacent molecules in the c columns in com-
parison to that in the ab planes in regular crystal domains
but also the face-to-face (laterally displaced) arrangement
of ppy ligands at twin boundaries which may favour the
charge transfer along the c direction (confer Section 3.1).
Furthermore, according to Ref. [17] the intermolecular dis-
tances at the twin boundaries should be slightly shorter
than the 8.4 A observed in the bulk of the lamellar domains
which goes well with our estimated values of r0 smaller
9 than 16.8 A corresponding to the two nearest neighbour
separation distances along the c axis. This fact taken
together with the previous assignment of the emission
EML spectrum (Fig. 2) suggests that the molecular aggre-
gates similar to those occurring at crystal twin boundaries
can be also involved in the exciton dissociation in Ir(ppy)3
lms. Therefore, our present results are in accordance with
the idea that a short range order of a crystal structure is
preserved in Ir(ppy)3 lms which is not unique but rather
typical for vacuum-evaporated organic lms (see for exam-
ple [46]).
The large value of g0 found in solid Ir(ppy)3 reects a
high degree of charge and energy delocalization of the
relevant excited states. A similar observation has been
recently made in star-burst amines [15] where highly delo-
calized excited states of those dendrimer systems exhibit
also large dissociation abilities. In the same way as in
Ref. [15] we speculate here that the ecient primary charge
separation step in the solid state is determined not only by
favourable electron exchange integrals between neighbour
molecules but also by ecient energy/charge delocalization
within the area of molecule itself. While the rst factor
entails naturally the molecular arrangements within the c
columns in Ir(ppy)3, the latter one should be ascribed to
strong spinorbit coupling introduced by heavy atoms in
such organo-metallic complexes [31]. A continuous
increase of the EML eect with excitation photon energy
158 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
(squares in Fig. 2) shows that in the solid state the primary
dissociation of hot excitons is a process competing
against their conversion and/or vibrational deactivation
to the lowest emissive state. In turn, a more abrupt (step-
like) increase of the EML eect occurring within the higher
photon energy range (where the LC states are excited) sug-
gests that dissociation of the hot LC excitons contributes
also to the total EML eect due to the favourable arrange-
ment of adjacent ligands as argued above.
The importance of specic intermolecular interactions in
neat Ir(ppy)3 lms can be easily recognized if we compare
the present EML results with those for Ir(ppy)3 dispersed
in solid matrices [10,11,47]. In the latter case the quenching
of Ir(ppy)3 emission under application of high electric elds
is signicantly weaker. For example, in a TPD:Ir(ppy)3 sys-
tem the Ir(ppy)3 emission intensity is only reduced by 2% at
an electric eld of 2 106 V/cm [10]. In this context we
recall also the recent transient EML experiments where
PL decays aected by external electric elds are measured
in TCTA:Ir(ppy)3 [11] and PVK:Ir(ppy)3 [47] systems. In
contrast to stationary measurements, the transient EML
experiments allow to distinct amplitude-type quenching
(when population of excited states is decreased) from
rate-type quenching (when the lifetime of emissive state is
reduced) [48,49]. Interestingly, both in TCTA:Ir(ppy)3
and PVK:Ir(ppy)3 systems the decay time of Ir(ppy)3 emis-
sion was independent of electric eld and the amplitude-
type PL quenching was observed. In particular, an ampli-
tude decrease of 5% has been detected at a eld strength
of 2.5 106 V/cm in the former system [11]. That result is
in full accordance with the Onsager model of exciton disso-
ciation where the population of emissive states is estab-
lished by hot exciton dissociation events and then the
decay time of relaxed emissive states should not depend
on electric eld. It would be instructive to make similar
experiments for neat Ir(ppy)3 lms and we address this
issue for our future research.
Finally, we note that the PooleFrenkel (PF) model [12]
is inappropriate to describe the charge separation in
Ir(ppy)3 lms. In the PF formalism a geminate eh pair is
assumed to dissociate by a single thermally-activated jump
over the coulombic barrier lowered by the external eld in
the down-eld direction. At high electric elds the one-
dimensional version of the PF model is usually applied
[44]. The dissociation process is then described in terms
of the rst order kinetic rate constant,
 p
bPF F
k eh k 0 exp ; 10
kT
where k0 is the zero-eld rate constant and the basic
parameter of the model, bPF depends on the relative per-
mittivity er of the material (for detailed description of the
PF model and its application to EML see Refs. [14,22]).
The theoretical curve obtained on the basis of the PF mod-
el (broken line in Fig. 3) deviates distinctively from the
experimental data and gives the stronger electric eld
dependence of EML signals in comparison to the Onsager
model. Failure to describe charge separation mechanism
according to the PF model is frequently reported in organic
solids (see for example Refs. [14,50]). The PF formalism
rests on premise that a charge carrier escapes Coulomb
attraction of its counterpart jumping over the potential
barrier in a one step process. This assumption is dicult
to realize in organic solids where the Onsager radius is
equal to many intermolecular distances (rC = 170 A for
typical value of er = 3) and the electronic wavefunctions
are strongly localized in space. The possible explanation
of the apparent success of the Onsager model originates
just from the fact that it takes into account the multi-step
character of charge separation process in terms of carrier
diusion. How valid is the Onsager formalism of carrier
continuous diusion in organic solids was widely discussed
in the context of geminate recombination on a discrete lat-
tice using the Monte Carlo simulation method [50] and the
master equation approach [51].
3.3. (1x)EML results
When electromodulation is induced by the sinusoidal
voltage without any external steady state component
(UDC = 0 V), as takes place in the present measurements,
(1x)EML eect measured at the rst harmonic of the
applied electric eld frequency should not appear in fully
symmetric and isotropic systems due to symmetry reasons.
Therefore, the nonzero (1x)EML signals (Fig. 4) indicate a
real asymmetry of Al/Ir(ppy)3/Al/glass cells prepared by
vacuum-evaporation. It is well established that the elec-
tronic structure of the sandwich devices with nominally
symmetric conguration is not in fact symmetric. Electro-
4
Ir(ppy)3
em =525 nm
5
Fi [10 V/cm]
I1 / I0 [ 10-2]
0
1.0
2.0
-4
3.0
0 1 2
Frms [106 V/cm]
Fig. 4. The (1x)EML signal (calculated according to Eq. (2)) as a function
of the rms value of modulating electric eld for a 310 nm thick Ir(ppy)3
lm. The circles stand for experimental data and the solid lines are
theoretical curves calculated according to the Onsager model for dierent
values of internal electric elds as indicated in the gure. The values of the
Onsager model parameters are the same as in Fig. 3.
 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
modulation of absorption (EA) or luminescence (EML)
spectroscopy oers an excellent opportunity to investigate
and quantify the eect by probing the average internal elec-
tric elds (Fi) built into the samples (for detailed back-
ground information, see Refs. [45,52]). The combined
analysis of the (1x)EA and (2x)EA response has been
recently applied to a wide range of molecular organic mate-
rials which made it possible to evaluate values of Fi
between 4 104 and 3 105 V/cm for sandwich arranged
cells, Al/organic lm/Al/glass substrate [52]. The observed
features of Fi were explained by the asymmetrical distribu-
tion of space charge in the near electrode regions associated
with dierent defect (of structural and/or impurity origin)
characteristics on the substrate and non-substrate side of
the samples. The similar approach can be also applied to
the present (1x)EML results for Ir(ppy)3.
In Fig. 4 the (1x)EML signal versus the rms value of
modulating electric eld is displayed for a Al/Ir(ppy)3/Al/
glass system. Theoretical plots are calculated according to
the Onsager model, inserting F(t) = Fi + F0 sin(xt) into
Eq. (9) with dierent values of Fi as indicated in the gure.
The results show that the internal electric eld Fi is not con-
stant and increases with the amplitude of the applied eld
from Fi = 1 105 V/cm under the low eld regime up to
above 2 105 V/cm at the higher elds. From EA measure-
ments on Ir(ppy)3 lms [52] we obtained Fi = (2 1)
105 V/cm approximately independent of the applied eld.
This apparent inconsistency can be explained by the dier-
ent time scales of phenomena involved in EA and EML.
While electroabsorption proceeds during the very short
interaction time of light with molecule (10161015 s),
the EML eect is averaged during the exciton lifetime
which is orders of magnitude longer. Therefore, the inter-
nal electric elds monitored in EA and EML experiments
can be of dierent characters as observed also for
other organic systems [45]. It is worth to note here, how-
ever, that electric elds Fi of (13) 105 V/cm are too weak
to aect noticeably the (2x)EML eect described in Section
3.2.
4. Conclusions
Electromodulation of photoluminescence in vacuum-
evaporated lms of Ir(ppy)3 has been investigated in sand-
wich arranged cells Al/Ir(ppy)3/Al/glass. The electric eld
characteristics, the emission and excitation spectra of
EML signals are rationalized in terms of the eld-assisted
dissociation of excitons into free and/or trapped charge
carriers. The charge separation mechanism is described
well by the three-dimensional Onsager model which indi-
cates that charge carriers overcome the potential barrier
by the diusive motion rather than one-step jumps. An
analysis of (1x)EML signals provides information on inter-
nal electric elds of (13) 105 V/cm occurring in Ir(ppy)3
lms, in accordance with previous electroabsorption mea-
surements for this material.
159
References
[1] C. Adachi, M.A. Baldo, M.E. Thompson, S.R. Forrest, J. Appl. Phys.
90 (2001) 5048.
[2] S. Watanabe, N. Ide, J. Kido, Jpn. J. Appl. Phys. 46 (2007) 1186.
[3] J. Kalinowski, Organic Light-Emitting Diodes: Principles, Charac-
teristics, and Processes, Marcel Dekker, New York, 2005.
[4] M.A. Baldo, S. Lamansky, P.E. Burrows, M.E. Thompson, S.R.
Forrest, Appl. Phys. Lett. 75 (1999) 4.
[5] M.A. Baldo, C. Adachi, S.R. Forrest, Phys. Rev. B 62 (2000) 10967.
[6] J. Kalinowski, W. Stampor, J. Mez_ yk, M. Cocchi, D. Virgili, V.
Fattori, P. Di Marco, Phys. Rev. B 66 (2002) 235321-1.
[7] J. Kalinowski, J. Mez_ yk, F. Meinardi, R. Tubino, M. Cocchi, D.
Virgili, J. Appl. Phys. 98 (2005) 063532-1.
[8] W. Holzer, A. Penzkofer, T. Tsuboi, Chem. Phys. 308 (2005) 93.
[9] R.J. Holmes, S.R. Forrest, T. Sajoto, A. Tamayo, P.I. Djurovich,
M.E. Thompson, Org. Electron. 7 (2006) 163.
[10] J. Kalinowski, W. Stampor, J. Szmytkowski, D. Virgili, M. Cocchi, V.
Fattori, C. Sabatini, Phys. Rev. B 74 (2006) 085316-1.
[11] S. Reineke, K. Walzer, K. Leo, Phys. Rev. B 75 (2007) 125328-1.
[12] J. Frenkel, Phys. Rev. 54 (1938) 647.
[13] L. Onsager, Phys. Rev. 54 (1938) 554.
[14] W. Stampor, Chem. Phys. 256 (2000) 351.
[15] W. Stampor, Chem. Phys. 315 (2005) 259.
[16] V.I. Arkhipov, H. Bassler, in: W. Brutting (Ed.), Physics of Organic
Semiconductors, Wiley-VCH, Weinheim, 2005, p. 183.
[17] J. Breu, P. Stossel, S. Schrader, A. Starukhin, W.J. Finkenzeller, H.
Yersin, Chem. Mater. 17 (2005) 1745.
[18] K.A. King, P.J. Spellane, R.J. Watts, J. Am. Chem. Soc. 107 (1985)
1431.
[19] P.J. Hay, J. Phys. Chem. A 106 (2002) 1634.
[20] W. Stampor, J. Mez_ yk, J. Kalinowski, Chem. Phys. 300 (2004) 189.
[21] E.B. Namdas, A. Ruseckas, I.D.W. Samuel, S.C. Lo, P.L. Burn, J.
Phys. Chem. B 108 (2004) 1570.
[22] J. Kalinowski, W. Stampor, P. Di Marco, J. Chem. Phys. 96 (1992)
4136.
[23] T. Tsuboi, T. Tanigawa, Thin Solid Films 438439 (2003) 301.
[24] M.G. Colombo, T.C. Brunold, T. Riedener, H.U. Gudel, M. Fortsch,
H.B. Burgi, Inorg. Chem. 33 (1994) 545.
[25] A.B. Tamayo, B.D. Alleyne, P.I. Djurovich, S. Lamansky, I. Tsyba,
N.N. Ho, R. Bau, M.E. Thompson, J. Am. Chem. Soc. 125 (2003)
7377.
[26] A. Tsuboyama, H. Iwawaki, M. Furugori, T. Mukaide, J. Kamatani,
S. Igawa, T. Moriyama, S. Miura, T. Takiguchi, S. Okada, M.
Hoshino, K. Ueno, J. Am. Chem. Soc. 125 (2003) 12971.
[27] T. Tsuboi, N. Aljaroudi, Opt. Mater. 27 (2005) 1859.
[28] T. Tsuboi, J. Lumin. 119120 (2006) 288.
[29] M.G. Colombo, A. Hauser, H.U. Gudel, Top. Curr. Chem. 171
(1994) 143.
[30] E. Jansson, B. Minaev, S. Schrader, H. Agren, Chem. Phys. 333
(2007) 157.
[31] H. Yersin, Top. Curr. Chem. 241 (2004) 1.
[32] G.A. Crosby, K.W. Hipps, W.H. Elfring Jr., J. Am. Chem. Soc. 96
(1974) 629.
[33] K. Nozaki, J. Chin. Chem. Soc. 53 (2006) 101.
[34] W.J. Finkenzeller, H. Yersin, Chem. Phys. Lett. 377 (2003) 299.
[35] S. Kodate, I. Suzuka, Jpn. J. Appl. Phys. 45 (2006) 574.
[36] W. Finkenzeller, P. Stossel, M. Kulikova, H. Yersin, SPIE Proc. 5214
(2004) 356.
[37] K.C. Tang, K.L. Liu, I.C. Chen, Chem. Phys. Lett. 386 (2004) 437.
[38] J.N. Damas, G.A. Crosby, J. Am. Chem. Soc. 92 (1970) 7262.
[39] C. Adachi, M.A. Baldo, S.R. Forrest, M.E. Thompson, Appl. Phys.
Lett. 77 (2000) 904.
[40] J. Kalinowski, W. Stampor, M. Cocchi, D. Virgili, V. Fattori, P. Di
Marco, Chem. Phys. 297 (2004) 39.
[41] Y. Kawamura, J. Brooks, J.J. Brown, H. Sasabe, C. Adachi, Phys.
Rev. Lett. 96 (2006) 017404-1.
160 W. Stampor, J. Mez_ yk / Chemical Physics 337 (2007) 151160
[42] N. Ohta, Bull. Chem. Soc. Jpn. 75 (2002) 1637.
[43] M. Tachiya, J. Chem. Phys. 89 (1988) 6929.
[44] M. Pope, E.C. Swenberg, Electronic Processes in Organic Crystals
and Polymers, Oxford University Press, New York, 1999.
[45] J. Kalinowski, W. Stampor, P. Di Marco, J. Electrochem. Soc. 143
(1996) 315.
[46] R. Elermann, G.M. Parkinson, H. Bassler, J.M. Thomas, J. Phys.
Chem. 87 (1983) 544.
[47] Y.B. Li, Y.B. Hou, F. Teng, X.R. Xu, Chin. Phys. Lett. 21 (2004) 1362.
[48] Z.D. Popovic, M.I. Khan, A.M. Hor, J.L. Goodman, J.F. Graham, J.
Phys. Chem. B 106 (2002) 8625.
[49] A. Nollau, M. Homann, T. Fritz, K. Leo, Thin Solid Films 368
(2000) 130.
[50] B. Ries, H. Schonherr, H. Bassler, M. Silver, Philos. Mag. 48 (1983)
87.
[51] H. Scher, S. Rackovsky, J. Chem. Phys. 81 (1984) 1994.
[52] W. Stampor, Chem. Phys. 334 (2007) 216.