University of Southern California Dissertations and Theses

Qualitative quantum chemical description of the structure-color relationships in phosphorescent organometallic complexes

(USC Thesis Other)

Qualitative quantum chemical description of the structure-color relationships in phosphorescent organometallic complexes

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ii

Qualitative quantum chemical description of the
structure-color relationships in phosphorescent
organometallic complexes

Patrick Saris
Mark E. Thompson Laboratory

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfilment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(CHEMISTRY)

May 2019
iii

TABLE OF CONTENTS
Chapter 1 - Introduction ............................................................................................................................... 1
Chapter 2 - Benzophenanthroline ............................................................................................................... 18
Chapter 3 - Computational methods .......................................................................................................... 40
Chapter 4 - Beyond phenylpyridine: novel platinum phosphors ................................................................ 64
Bibliography ............................................................................................................................................ 78

iv

List of tables
Table 2-1. Calculateda frontier orbital and triplet state energies of candidate triphenylenes calculated at
B3LYP/631G** level .................................................................................................................. 19
Table 2-2. Comparison of relevant properties of xyBzp 2IrPic and Firpic .................................................... 36
Table 3-1. Summary of calculated and experimental physical properties for 10b, 10c and 10c. ............... 53
Table 3-2. Number of - contacts and calculated mobilities for 10b, 10c and 10c. ................................ 57
Table 3-3. Efficiency parameters of the OLEDs for 10b, 10c and 10d. (η r was computed assuming x =1 and
n e = 0.2) ..................................................................................................................................... 59

v

List of figures

Figure 2-1. Phenylpyridine-like cyclometallation and representative triphenylenes ................................. 19
Figure 2-2. Retrosynthetic analysis of Bzp .................................................................................................. 21
Figure 2-3. Synthesis of Bzp (5) ................................................................................................................... 26
Figure 2-4. Synthesis of 5Pt(dpm) and crystal structures of 5 and 5Pt(dpm). ORTEP structures shown at
50 % probablility level ............................................................................................................ 27
Figure 2-5. Absorption and emission spectra of (a) 5 and iso-5; (b) 5PtDpm; (c) 10PtDpm. All absorption
spectra were collected in methylcyclohexane at 298 K. MeCy = methylcyclohexane; MeTHF =
2-methyltetrahydrofuran; PMMA = polymethylmethacrylate ................................................. 28
Figure 2-6. Proton NMR spectrum of (Bzq)(COD)(Acac)2Ir......................................................................... 32
Figure 2-7. Visual progression of the modified Nonoyama reaction .......................................................... 33
Figure 2-8. Absorption and emission spectra of phBzp2IrPic (left) and xyBzp2IrPic (right) ....................... 35
Figure 2-9. Emission spectra of facial (black) and meridonal (red) xyBzp3Ir in room temperature
methyltetrahydrofuran solution (solid trace) and 77 K frozen solvent matrix (dashed trace) 37
Figure 3-1. Predicted LUMO energies of 135 tetraazapentacenes ............................................................. 41
Figure 3-2. Predicted oxidation potential vs. reduction potential of all 201,021 'mega-run' azapentacenes
.................................................................................................................................................. 42
Figure 3-3. (top) Elemental enumeration of O, S, NH and CH in the “P ring” of the H2P structure. (bottom)
Lowest unoccupied molecular orbital energy vs. triplet energy for iterative
cyclopentaphenanthrene structures. ....................................................................................... 46
Figure 3-4. General scheme of H2P synthesis ............................................................................................. 47
Figure 3-5. LUMO vs. triplet energy for second tier iteration of aza substitution in phenanthrene section
the H rings of the parent phenantho[4,5 f]imidazole. Compounds 10a 10d¬ are illustrated by
colored circles. The identities of the other compounds in the scr .......................................... 49
Figure 3-6. Highest occupied molecular orbital and lowest unoccupied molecular orbital diagrams
calculated at the B3LYP/LACV3P** level of theory. The permanent dipole moment for each
molecule is illustrated in the images at the top. ...................................................................... 51
vi

Figure 3-7. Absorption (black line), fluorescence (blue line), and cryogenic (77 K) spectra (red line) of (a)
10b, (b) 10d, and (c) 10c. In each subfigure, the bottom axis refers to wavelength and the top
axis denotes energy in eV. (d) Solid state phosphorescence .................................................. 52
Figure 3-8. Center of mass radial distributions [g(r)] from MD simulations of three host materials, obtained
by averaging over 30ns. ............................................................................................................ 55
Figure 3-9. Histogram plots showing the distribution of hole and electron hopping rates extracted from
the frontier dimer orbital splitting coupling calculations of both the exhaustive dimer set and
the smaller 20 dimer subset for the three host materials ....................................................... 56
Figure 3-10. OLED characterization: (a) Energy level diagram of the OLEDs (blue line is FIrpic), (b) J V L
characteristics, (c) plot of EQE versus current density. ............................................................ 58
Figure 3-11. Current-voltage plots for electron only devices, i.e. ITO/BCP (10 nm)/H2P (40 nm)/LiF:Al. . 61
Figure 4-1. Metal donating/withdrawing strategy for blue shifting phosphorescence ............................. 65
Figure 4-2. Emission spectra of [Pt]3Trippy in PMMA (red) and frozen methylcyclohexane (black) ......... 68
Figure 4-3. Excitation and emission spectra of Fac Czpy3Ir ........................................................................ 69
Figure 4-4. Rotationally restricted design strategy ..................................................................................... 70
Figure 4-5. Syntheisis of Snap3Ir ................................................................................................................. 70
Figure 4-6. Emisssion spectra of Pt complexes of CzPy and Snap in MeTHF .............................................. 71
Figure 4-7. Undesired side reaction resulting from attempted amidine condensation ............................. 73
Figure 4-8. Generic scheme of phosphine imide formation ....................................................................... 74
Figure 4-9. X-ray crystal structure of PtPie ................................................................................................. 75
Figure 4-10. Excitation and emission spectra of solid PtPie suspended in DMSO ...................................... 76
1

CHAPTER 1 - INTRODUCTION
Phosphorescence is a phenomenon derived from a property its namesake element,
phosphorus, from the Greek for “bearer of light”. White phosphorus, the element’s molecular
allotrope, was observed to emit a faint glow in the dark when it was first isolated in 1669, hence
its name. It is now known that this glow resulted from the reaction of P4 with aerobic oxygen, and
has been reclassified as chemiluminescence. Phosphorescence, however, still refers to the
persistent “glow-in-the-dark” luminescence exhibited by some materials after the absorption of
light. Phosphorescence from small molecules was determined in the 1960’s to arise from a
transition from a triplet excited state to a single ground state. This apparent “spin-flip” transition
is forbidden by a quantum mechanical spin selection rule. This rule is relaxed by spin-orbit
coupling which mixes singlet and triplet states, giving some allowedness to such a transition. Most
organic compounds exhibit very weak phosphorescence, if at all, with radiative lifetimes on the
order of seconds. Some organometallic complexes of heavy atoms phosphoresce many orders of
magnitude faster with high efficiency, and as such they are widely applied in electroluminescent
displays. Namely, that application is the organic light emitting diode (OLED).
An OLED consists of a thin film of dye-stuff sandwiched between two conducting
electrodes. A voltage applied to the electrodes causes a current to pass through the film, exciting
the dye and causing it to luminesce much is the way ultraviolet irradiation (e.g. from a black light)
causes photoluminescence from a dye. The fundamental difference between optical excitation and
electrical excitation is the nature of the excited state produced. Phosphors (as opposed to
fluorophores) may luminesce from the triplet states produced by electrical excitation. When an
oxidized species (cation, hole) meets a reduced species (anion, electron), they may recombine by
electron transfer to form two neutral species. While both resulting species may be in their ground
2

states, the extreme exothermicity of such an electron transfer lies well within the inverted region
of the Marcus electron transfer rate equation. Thus, a faster process dominates: electron transfer
recombination to form one ground state and one excited state. Since the excited state of interest
in OLEDs is the emissive state, only the lowest lying excited state manifold will be considered in
accordance with Kasha’s rule. With this consideration, holes and electrons may recombine to form
one of 4 possible excited states, each with equal probability of formation. The unpaired electrons
of the singlet state have antisymmetric (antiparallel) spins, with an overall spin quantum number
S = 0, while the unpaired electron spins in the three triplet sub-states have symmetric (parallel)
spins, each with an overall spin quantum number S = 1. The statistical branching ratio between
the aptly named singlet (one possible state) and triplet (three possible states) is therefore 25 % to
75 %.
Because luminescent molecular species generally have a closed shell singlet ground state,
only transitions from singlet excited states to the ground state are spin allowed by classical
quantum mechanics. The maximum internal quantum efficiency (IQE) of an OLED would be 25
% unless triplets may also contribute to luminescence. One mechanism of luminescence from
triplet excited states is P-type delayed fluorescence, arising from the recombination of two triplet
excitons to form one singlet exciton. The maximum IQE of an OLED based on P-type
fluorescence is 62.5 %, due to the loss of one exciton in the triplet-triplet recombination event.
Another mechanism of luminescence from triplet excited states is E-type delayed fluorescence,
arising from endothermic intersystem crossing from T1 to S1. This mechanism requires a small
singlet-triplet energy splitting (exchange energy), usually arising from a donor-acceptor charge
transfer type chromophore. E-type fluorescence, also known as thermally activated delayed
fluorescence, has recently gained attention for producing OLEDs with 100 % IQE. A third
3

mechanism of luminescence from triplet excitons is phosphorescence, the focus of this account.
Phosphorescence arises from the mixing of singlet and triplet states by spin-orbit coupling, giving
allowedness to transitions from nominally triplet excited states to nominally singlet ground states.
Although phosphorescent materials allowed OLEDs to achieve 100% IQE over 15 years ago, and
are currently incorporated in all high efficiency commercial OLED displays, challenges still
remain. Blue emitting phosphorescent OLED materials suffer from greatly reduced operational
lifetimes compared to their green to red emitting counterparts. This account will discuss the
structure-property relationships of phosphorescent OLED dopants, with particular focus on color
tuning towards blue emission, while maintaining favorable properties such as chemical stability,
high radiative rates, and favorable electrochemical properties.

The following physical properties are important to the performance of an OLED dopant:
1. Color
The emission color of an OLED is fully described by its emission spectrum, from
which several useful simplifying descriptors may be derived. Qualitative color can be defined by
a coordinate on the CIE color gamut. While CIE coordinate defines the OLED’s appearance to
the eye, it fails to capture the effect of the emission spectral shape on the performance of the OLED.
Narrow spectral line shapes are generally favorable for display applications since they appear as
fully saturated colors, like the red, green, and blue color vertices of NTSC or PAL display gamuts.
Broad emission line shapes are desirable for lighting applications, since they appear as unsaturated
colors like sunlight, leading to higher color rendering indices. An important color parameter to
consider when designing an OLED dopant is the onset wavelength of the emission, since this
4

represents the energy of the excited state. For spectra with clearly defined vibronic progressions,
the highest energy peak is assigned as emission to the zeroth vibrational level of the ground state,
and therefore corresponds to the energy of the excited state (E0-0). Assignment of E0-0 is less
rigorous for featureless emission spectra, so an onset wavelength is generally considered, usually
where the emission intensity rises to 10 % of the maximum.
2. Phosphorescence
Due to the statistical exciton branching ratio of electrogenerated singlet and triplet excited states
(1:3), an OLED dopant must be able to emit from triplet excited states in order to achieve an IQE
over 25 %. Several mechanisms allow triplet excitons to produce photons have been mentioned
above: P-type fluorescence, E-type fluorescence, and phosphorescence. Here we will rationalize
how the structural characteristics of organometallic dopants tune their phosphorescent properties.
3. Photoluminescent quantum yield
The radiative efficiency of a luminophore is defined as the ratio of number of emitted
photons to number of excitons (excited molecules). For common fluorophores, the
electroluminescent quantum yield is lower than the photoluminescent quantum yield, due to the
formation of nonradiative triplet excited states. For the phosphorescent dopants discussed here,
intersystem crossing is much faster than fluorescence, and so all excitons are rapidly converted to
the triplet spin state regardless of whether electro- or photo- generated. Thus, photoluminescent
quantum yield is a good indicator of the electroluminescent quantum yield of a dopant inside a
working OLED. PLQY may equivalently be expressed as a competition of rates for deactivation
of the excited state. Namely, the PLQY is the ratio of the rate of radiative relaxation (kr) to the
total rate of relaxation, both radiative and non-radiative (Knr). Therefore: PLQY (= Kr/Kr+Knr).
5

While maximizing PLQY is always favorable for an OLED, values over 80 % are considered
highly efficient.
High PLQY is a crucial property for OLED dopants, achieved by maximizing radiative
rates while minimizing non-radiative rates. However, a high PLQY due to both rates being slow
is not desirable. High absolute radiative rates improve the efficiency of OLEDs, particularly at
high brightness. A striking example is an OLED made with platinum octaethylphorphyrin as the
dopant. While PtOEP is an efficient phosphor, due to low non-radiative rates, its low radiative
rate and therefore long excited state decay time leads to increased bimolecular quenching
mechanisms at high current densities.
4. Transport energy levels
In almost every host-dopant pair, the dopant is a thermodynamic trap to one or both
charges. The level of doping tends to be high enough, however, that percolation pathways allow
charge transport to be mediated by the dopant itself. It is therefore important to consider the
energies of the highest occupied and lowest unoccupied molecular orbitals of a dopant material
with respect to the host matrix.

5. Chemical stability (the blue problem)
Operational lifetimes of working OLEDs depend on many factors, including contamination
by impurities during fabrication, diffusion of environmental oxygen or moisture into the device,
and physical reorientation of the amorphous organic layers such as crystallization or second order
phase transitions. These problems have been largely solved, allowing for commercial application
6

of long lived green and red OLEDs. Blue OLEDs, however, still suffer greatly in operational
lifetime due to chemical decomposition within the emissive layer. It stands to reason that the
higher energy excited states within a blue emitting device approach the bond dissociation energy
of the organic components. While homolytic bond dissociation energies are well above the energy
of a blue photon, heterolytic cleavages may be accessible. For instance, oxidized or reduced
species, which are present in the active layer as charge carriers, have far lower heterolytic bond
dissociation energies than their neutral counterparts. Furthermore, bimolecular recombination of
two excitons each at 2.75 eV (blue) can produce an exciton at 5.0 eV (deep UV). Thus, robust
chemical structures are necessary in order to withstand the presence of high energy (blue, or deep
UV) excitons together with polarons.

- excited state decay lifetime-
Kr
Spin-orbit coupling (SOC), the principal requirement for phosphorescence, is a relativistic effect
which mixes pure singlet and triplet spin states into states consisting of some part singlet and
triplet. Specifically, SOC couples the otherwise orthogonal quantum numbers l (angular
momentum) and s (spin) into one quantum number (ls). Though this is a relativistic effect, it can
be qualitatively described by a simple Bohr-model atom and classical electromagnetism (insert
figure). The angular motion of the electron around the positively charged nucleus can be seen in
the inertial frame of the electron as the movement of the nucleus around the electron. This moving
positive charge induces a magnetic moment that couples to the spin of the electron. The degree of
7

SOC is integral to the performance of a phosphorescent dopant in an OLED, and the relationship
between molecular structure and degree of SOC will be discussed in this account.
While spin orbit coupling is an effect of atoms with high atomic number, large spin orbit coupling
requires more than just the mere presence of heavy atoms. For example, heavy atoms present in
solvent around a chromophore can lend some SOC to the system. The degree of SOC tends to be
low in this case, leading to fast intersystem crossing but slow phosphorescence. A common
implementation of this “external heavy atom effect” is the use of methyl iodide as solvent or
additive to sensitize triplet excited states in fluorophores. Atoms which are part of the luminophore
may also be considered “external” if they are not involved in the triplet excited state. Halogenated
analogues of fluorophores tend not to be luminescent, and are used as triplet sensitizers. Rose
bengal, for instance, is a halogenated derivative of fluorescein used to excite oxygen in
photodynamic therapy.
In the case of organo heavy metal phosphors, the degree of metal character in the emissive
state is key to the degree of SOC and therefore the radiative rate constant. Four types of transitions
will be considered: ligand centered (3LC), metal to ligand charge transfer (3MLCT), ligand to
ligand charge transfer (3LLCT), and metal centered (3MC). The emissive state of most
organometallic phosphors consists of some linear combination of these types of transitions.
Hypothetical deconvolution of the triplet excited state into a combination of these “localized
states” is known as the localized state approximation, which is useful in gaining a qualitative
understanding of structure property relationships and developing molecular design principles.
Each state contributes to the properties of the emissive state. The degree of contribution from each
of the states depends on the structural characteristics of the compound, to be discussed in section
??. The energy of the emissive state is equal to, or less than, the energy of the lowest lying localized
8

state. Therefore, in order to maximize the emission energy of (blue shift) the emissive state, the
energy of the localized states must be maximized. Furthermore, each localized state contributes
differently to the radiative rate of the emissive state, with the lowest lying state(s) dominating the
emissive kinetics.
The 3LC state is a direct analogy to the external heavy atom effect. It is characterized by
vibronically structured emission, and little to no metal character in the excited state. Similarly to
halogenated organic fluorophores, organometallic compounds with a 3LC dominant emissive state
undergo rapid intersystem crossing, but slow phosphorescence. Thus, the 3LC state tends to be
weakly emissive. In cases of closely lying singlet and triplet states, because phosphorescence is
slow but ISC is fast, E-type fluorescence is observed.
The 3MLCT state provides metal character in the emissive state. This leads to large values
of spin orbit coupling, increasing the allowedness of phosphorescence. However, charge transfer
states have low natural transition orbital overlap, which lead to low oscillator strengths for their
transitions.
The 3LLCT state usually arises from a ligand redox mismatch within heteroleptic metal
complexes. That is, a complex contains at least one easy to oxidize ligand and at least one easy to
reduce ligand, such that the lowest lying state is a charge transfer from the former to the latter. In
extreme cases, this state may be lowest lying even in homoleptic complexes, but only in high
dielectric environments and when other states such as 3MLCT are very high in energy (i.e. Zn
dipyrrins). The 3LLCT state is characterized by a broad unstructured emission spectrum, and is
strongly solvatochromic. Due to little metal character as well as poor natural transition orbital
overlap, compounds with a 3LLCT dominant emissive state tend to have low radiative rates, and
modest photoluminescent quantum yields.
9

The 3MC state arises from a transition within the d-orbital manifold of the metal. This is
a deleterious state for OLED phosphors for two reasons: strong vibrational coupling to the ground
state leads to fast non-radiative rates, and popluation of antibonding d-orbitals tends to lead to
ligand dissociation and chemical decomposition. The 3MC states efficient OLED phosphors are
sufficiently inaccessible, either thermodynamically or kinetically, and the structural features
leading to this property will be discussed in section (XX).
The most efficient phosphors are predominantly 3MLCT and 3LC in nature. Low orbital
overlap (and therefore low radiative rate) of the 3MLCT is counteracted by the other contributing
localized state: the pi-pi* 3LC state. The 3LC state has a high degree of orbital overlap, and is
symmetry allowed, as evident from the high extinction coefficients for the singlet transitions of
the free ligands. The 3LC state, however, suffers from small SOC parameters. Thus, a phosphor’s
emissive state with predominantly 3LC character has low radiative rates due to low SOC, while
those with predominantly 3MLCT character suffer from low radiative rates due to low orbital
overlap. Highly efficient phosphors exhibit a balance of the two states to give high allowedness
to the emissive triplet transition.
Another important factor in radiative rate is geometrical in nature. A selection rule for spin
orbit coupling between a particular pair of singlet and triplet states is that they originate from
different orbital parentages. For example, a 3MLCT state involving a platinum dz2 orbital may
not couple to a 1MLCT state involving that same dz2 orbital. Instead, it may couple to a higher
lying 1MLCT involving another Pt d orbital. The energy difference between the states is inversely
proportional to the coupling between them, so pseudo-degeneracy between frontier orbital energies
is beneficial to high SOC. In the case of (pseudo)-square planar Pt(II), the molecular orbital of
dz2 parentage lies well above any other molecular orbital, whereas in the case of (pseudo)-
10

octahedral Ir(III), the formally degenerate t2g manifold of orbitals lead to a high density of 1MLCT
states near in energy to the lowest lying 3MLCT state. This rationalizes the otherwise suprizing
result that iridium complexes have far higher degrees of SOC platinum complexes of the same
ligand system, despite platinum's higher atomic weight and SOC constant.

Knr
Non-radiative rates depend predominantly on two processes: vibrational coupling of the
emissive state with the ground state, and crossing of the emissive state into another, non-emissive,
excited state. Coupling of overtones of the ground state vibrational levels with the zeroth
vibrational level of the excited state are inversely proportional to the order of the overtone, so
redder emitting compounds suffer more vibrational deactivation. This is known as the “energy gap
law”, leading to comparatively low quantum yields for low energy emitting compounds. On the
other hand, bluer emitting compounds tend to cross into a non-emissive 3MC state. For instance,
Irppz3’s 3MLCT state efficiently crosses into a 3MC state at room temperature, which has a
conical intersection with the ground state through ligand dissociation.
The energy of the non-emissive 3MC state is dependant of the ligand field strength of the
cylcometalating ligands. Ligand field strengths depend on geometry and bonding. In general, the
strongest ligand fields are octahedral third-row transition metal complexes with phenyl or carbene
ligands. Weak ligands include azoles such as pyrazoles. It has been shown that under
photoexcitation, Irppz3 undergoes ligand dissociation. In solution, the open coordination site at
iridium can be trapped by isocyanide ligands under photoexcitation. The same sort of mechanism
also occurs in operating devices, where phenylpyrazole is replaced by bathophenanthroline. In
11

square planar 4-coordinate platinum complexes with two bidentate ligands, a common metal
centered deactivation is a D4h to C2V type distortion, where a ligand twists out of plane, leading
to a d-d state. Other ligand based distortions include free rotors or distortions which lead to
deactivation of the emissive triplet state.

- color-
The energy of a transition, and therefore of an excited state, is defined as the difference in
energy between the zeroth vibrational level of the ground state and the zeroth vibrational level of
the excited state. In the case of HOMO to LUMO transitions, it may seem that the excited state
energy is equal to the HOMO-LUMO gap, since that is the vertical transition energy of an electron
from a filled orbital to an unfilled orbital. This is not the case in molecular systems with confined
excitons. The difference in energy between frontier orbital energies and exciton energy is known
in semiconductor physics as exciton binding energy. The proximity of the newly populated orbital
(formerly LUMO) to the nascently vacated orbital (formerly HOMO) leads to both coulombic and
exchange interactions which perturb the energy of the excited state. Coulombic attraction between
the promoted electron to the hole left by the electron’s evacuation stabilizes the excited state by
up to 1 eV with respect to the HOMO-LUMO gap in many organic compounds. Meanwhile,
destabilizing exchange energy due to correlated movement of electrons in the excited state may
counteract the coulombic attraction, resulting in an excited state energy near, but not exactly equal
to, the HOMO-LUMO gap. While coulomb interaction between charges is spin-independent, the
exchange destabilization is only present in the anti-symmetrically spin paired singlet with respect
to the electrochemical gap. In summary, triplet excitons in tightly confined, non-charge transfer
12

molecular compounds are lower in energy than their electrochemical gaps, while similar singlet
excitons are near or slightly below their electrochemical gaps. This is less the case in less confined
or charge transfer excitons. Less overlap between unpaired electron orbital densities leads to lower
exchange interaction, which larger charge transfer distances lead to lower coulomb stabilization.
Consider azulene, with a short charge transfer distance but low HOMO-LUMO overlap. The
singlet excited state of azulene is over 1 eV below the electrochemical gap.
Because frontier orbital energy gaps (and therefore RedOx gaps) are not equal to optical
gaps, their values should be determined from electrochemical or photoelectron experiments alone,
and never from absorption or emission studies. However, despite the absolute discrepancy
between optical gap and redox gap, in they tend to be directly related. That is, widening the redox
gap of a material (opening the HOMO-LUMO gap) generally results in a blue shift in emission.
Tris(2-phenylpyridinato)iridium (III), abbreviated Irppy3, is highly efficient (phi = 0.8)
and has a short excited state lifetime (t = 1.4 us). therefore, Irppy3 provides a case study on the
balance of 3MLCT and 3LC states for an efficient phosphor. Density functional theory calculation
of the frontier orbital wave function contours provides a useful qualitative model of Irppy3. The
highest occupied molecular orbital of Irppy3 is an admixture of an iridium based d-orbital (part of
a nearly triply degenerate ‘t2g-like’ manifold) and pi a bonding Huckel orbital on the phenyl
groups. The lowest unoccupied molecular orbital is predominantly pyridine in parentage, with
some phenyl character. A HOMO to LUMO transition can therefore be imagined as a linear
combination of two states: phenylpyridine 3LC state and a Ir-to-pyridine 3MLCT state, as
discussed in the previous section.
The color shift from phenylpyridine to other ligand systems is usually due to a shift in both
3MLCT and 3LC contributing states. For example, phenylquinoline differs from ppy by a
13

benzannulation of the the pyridine. This both lowers the triplet energy of the ligand as well as
making the nitrogen containing fragment (the locus of LUMO) easier to reduce, thus lowering the
3MLCT energy. Qualitatively, it may be considered that the 3MLCT has a greater contribution,
since the emission line shape if pqir is broad and featureless, while irppy has some vibronic
structure.
In the absence of 3LLCT states, as is the case for homoleptic organometallic phosphors
such as irppy, the triplet excited state is a configurationally mixed state consisting of a metal to
ligand charge transfer state (3MLCT) and a pi-pi* ligand centered state (3LC). If, however a
3LLCT state is low in energy, the emission profile becomes broad and featureless, strongly
solvatochromic, and low in radiative rate.

The following general structure property relationships apply:
1. Replacement of an aromatic CH fragment with N (aza-substitution) makes the containing
heterocycle harder to oxidize and easier to reduce. Therefore, if the aza-substitution occurs under
primarily HOMO density, a blueshift occurs, while under LUMO density a red-shift occurs. [e.g.
Mdq (Red emission), Dbq (green emission), and Bzp (Blue emission)]
2. Ring contraction of an aromatic 6-membered ring to a 5- member ring makes the ring
easier to oxidize and harder to reduce. For instance , pyrrole is more electron rich than pyridine.
Therefore, ring contraction under primarily HOMO density leads to a redshift, while under LUMO
density it leads to a blue shift.
3. Extension of the pi system by benzannulation often leads to a red shift, though not
always. Benzannulation usually lowers the ligand centered triplet energy, while concurrently
14

making the annulated pi system both easier to reduce and easier to oxidize. Such a case should
always lead to a red shift in emission energy. However, in certain cases, benzannulation increases
the ligand centered excited state energies. A more detailed molecular orbital picture is required to
determine when this is the case. Note for example the blue shift in benzanulation of Pt(bzq) to
Pt(dbq).
Ancillary ligands – not so ancillary after all
Not only is the oxidation potential of the metal center governed by the admixed orbitals of
the cyclometallating ligand, but also by formally “ancillary” ligands, bystanders in terms of excited
state orbital contribution, but not so innocent in terms of inductive withdrawal of electron density
from the metal center. Consider the series of acac, pic, pz2H, to pz4B ancillary ligands to bis(2-
phenylpyridinato)iridium. The increasing distance of the anionic charge to the metal center is
directly related to the hypsochromic shift in emission of these complexes, even though these
ligands are not directly involved in the emissive triplet state. Instead, the increasing ionic character
of the LX-type ligand makes the iridium center harder to oxidize, destabilizing the 3MLCT. It is
notable that as the 3MLCT is destabilized along the series, the 3LC remains the same, so the
emission spectrum becomes more vibronically structured due to a greater contribution from the
3LC state.
-Device operation properties-
Transport energy levels
While the dopant is generally responsible for color tuning, and the host is generally
responsible for charge transport, the dopant is almost never innocent of affecting transport of one
or both charges. Furthermore, if host and dopant frontier orbital energies are too strongly
15

mismatched, emission may be quenched by excited state electron transfer. Therefore, the
electrochemical properties of the dopant are critical to device structure and performance. For
instance, a dopant with “shallow” HOMO and LUMO energies close to the vacuum energy
necessitate a similarly “shallow” host material. In this case, electrons are transported at a very
high chemical potential, which may lead to chemical decomposition and device degradation.
Deep trapping regime
If the host-dopant frontier orbital offset is large (> 500 meV), one charge is completely
trapped on dopant molecules and may only move by hopping through percolation pathways of
nearby dopants. In most devices, the dopant carries the holes, since metal complexes tend to be
easier to oxidize than high band gap organic semiconductors. Thus the hole conductivity of a
doped film is strongly dependent on dopant concentration. An example of this deep trapping
regime from Forrest and coworkers is a device with an emissive layer comprising mCBP as host
(HOMO = 6.0 eV), which carries the electrons, doped with Ir(dmp)3 (HOMO = 5.0 eV), which
carries the holes. They demonstrate the ability to modulate charge transport (and therefore exciton
generation) by varying the concentration of dopant throughout the emissive layer, leading to
decreased chemical decomposition and increased device operation lifetime. We suspect, however,
that the device operation lifetime may ultimately be limited by the responsibility of the carbazole
containing host material to carry electrons, inevitably leading to degradation by loss of carbazolate.
Note that while this is an example of the deep trapping regime, the transport levels are shallow.
Reversal of the charge trapping roles of host and dopant, where the dopant is a deep trap for
electrons, relieves the host of its duty to carry electrons at high chemical potential. Many examples
of this regime make use of dopants such as Firpic in carbazole containing hosts. Unfortunately,
any potential benefits to device operational lifetime are masked by the presence of fluorine in blue-
16

emitting dopants with deep transport levels. Carbon-fluorine bonds are unstable to reduction,
leading to very short operational lifetimes. Design of fluorine free dopants with similar
electrochemical properties to Firpic is a strategy for improved operational lifetime that we are
pursuing.
Shallow trapping regime
On the contrary, if one or both frontier orbital offsets is less than about 300 meV, charges
are only partially trapped on the dopant and may be transported by the host through endothermic
electron transfer. The benefit of this regime is that charge transport is less reliant on percolation
pathways, as well as removing the responsibility of either host or dopant to be the sole transporter
of a given charge. It has been suggested by Coehoorn et. al. that the ideal OLED architecture is
within this shallow trapping regime. In that particular case, a co-host structure is suggested, with
two host materials transporting one charge each. Regardless, a host-dopant frontier orbital offset
of 200 meV is predicted by their modelling study to be optimal for device performance.
Chemical stability
Aside from the effect charge transport has on device stability mentioned above, the dopant
also affects operational lifetime through its own structural integrity. Some functional groups and
coordination modes seem to add to or detract from the chemical stability of the phosphor. For
example, fluorine containing phosphors are categorically unstable in operating devices. Upon
reduction, fluoroaromatics decompose into fluoride and the corresponding aryl radical. During
device operation, electrons flowing through or near a fluorinated compound promote this
decomposition, leading to device degradation. In another example, an iridium complex of
phenylpyrazole decomposes by ligand loss in its excited state, due to the high energy of the excited
17

state, the weak coordination offered by pyrazole, as well as the sub-optimal coordination bite angle
of a 5-membered metallacycle fused to a 5-membered heterocycle. Presumably, the excited state
energy is high enough to overcome a barrier to Ir-N bond dissociation at room temperature leading
to a 5 coordinate complex which can be trapped by coordinating solvent such as tBuNC. Further
excitation leads to loss of the ppz ligand entirely, and the same process has been shown to occur
in an operating device. Design principles for stronger coordination includes more favorable bite
angles, rigid ligand frameworks, stronger field ligands such as carbenes, and increased denticity
of ligands.
A quantitative measurement of efficiency for electrically driven white lighting is not
straightforward. The ratio of emitted photons to injected charge carrier pairs is defined as the
external quantum effieciency (EQE) in unites of γ/e
-
. While EQE is a useful parameter in a
research setting, it is insufficient to meaningfully describe the efficiency of a light source because
it does not take thermalization losses into account. That is to say that the energy difference
between the charge carrier pair and the resulting photon is not expressed in the EQE. Power
conversion efficiency (PCE) is the ratio of input electrical power to output optical power, a unitless
parameter expressed as a percentage out of an ideal 100 %. PCE is a very natural efficiency metric
for devices such as photovoltaics, but fails to capture a key aspect of lighting applications: the
human photopic response. While PCE sufficiently represents the conversion of electrical energy
to electromagnetic radiation, it does not describe the portion of that radiation which is useful for
lighting, namely, the visible portion. The incandescent lamp, for instance, has a high PCE, but
most of the energy is converted into infrared radiation that is not visible to the human eye. The
human photopic response is an approximately gaussian power function of observed intensity vs.
wavelength centered on 556 nm.
18

CHAPTER 2 - BENZOPHENANTHROLINE
Blue emitting phosphorescent organic light emitting diodes (OLEDs) suffer from shorter
operational lifetimes than their green to red emitting counterparts, but new materials will enable
robust device architectures for more resilient blue luminescence. Chemical decomposition of the
active layer of an OLED, consisting of a conductive organic host material doped with a
luminescent organometallic complex of Ir or Pt, is the primary source of luminance loss in state of
the art devices. The instability particular to blue OLEDs stems from the molecular design principle
for blue shifting the luminescence of known green dopant phosphors: widening the HOMO-
LUMO gap. Destabilization of the LUMO by any means necessitates conduction band energies
well shallow of -2 eV, limiting the choice of suitable host materials to inherently unstable structure
classes such as phosphine oxides or to materials with suboptimal band alignment for high power
efficiency. Alternatively, stabilization of the HOMO by fluorine substitution leads to electron
promoted decomposition of the dopant phosphor by fluoride loss. Rather than substitution of CH
with more electronegative CF, isoelectronic transmutation of CH with more electronegative N is
a robust strategy for tuning frontier orbital energies, particularly for HOMO stabilization.
The triphenylene framework offers a unique combination of extended pi- conjugation and
high triplet state energy. Triphenylene’s triplet state energy (E T = 2.90 eV) is surprisingly more
than double that of its structural isomer, tetracene (ET = 1.25 eV). In fact, this is slightly higher
than biphenyl (ET = 2.85 eV), the framework for the phenylpyridine based ligands discussed in
chapter 1. This high ligand centered triplet energy makes triphenylene suitable for blue emitting
metal complexes. Isoelectronic transmutation of CH to N in triphenylene leads to an
aza-triphenylene capable of phenylpyridine-like cyclometallation. However, aza-triphenylenes
19

are not as well explored as phenylpyridines, due at least in part to their relative synthetic
inaccessibility.
The regiochemisty of aza-substitution
on the triphenylene frame is critical to the
emission wavelength of the resulting metal
complex. This influence is due almost
entirely to perturbation of the triplet metal to
ligand charge transfer state (
3
MLCT) energy
and not to the ligand centered state (
3
LC)
energy. Figure 2.1a shows the common
phenylpyridine-like cyclometallation motif
and the structures of triphenylene
(compound 1) and three azatriphenylenes.
Figure 2.1b lists calculated triplet state, HOMO, and LUMO energies for those four compounds.
The HOMO and LUMO values were calculated using an adiabatic solvent model implemented in
the Schrodinger Materials Science Suite. The triplet state energies were calculated as the
difference in gas phase energies between the optimized ground state geometry and optimized
triplet state geometry for each compound. As a result of this approximate method, the calculated
triplet state energy for compound 1 is 220 meV higher than the experimental value (ET = 2.9 eV).
It stands to reason that this is a standard error across the series of azatriphenylenes. It is important
to note that aza-substitution of triphenylene has little to no effect on its predicted triplet state
energy. The frontier orbital energies, however, depend on both the extent and location of
aza-substitution. Any incorporation of nitrogen into triphenylene lowers both HOMO and LUMO

Figure 2-1. Phenylpyridine-like cyclometallation and
representative triphenylenes

Table 2-1. Calculateda frontier orbital and triplet state
energies of candidate triphenylenes calculated at
B3LYP/631G** level
Structure 1 2 3 4
Triplet (eV) 3.12 3.10 3.12 3.13
LUMO (eV) -1.34 -1.63 -2.19 -1.79
HOMO (eV) -5.70 -5.83 -5.94 -5.98

20

energies, but for the doubly substituted compounds 3 and 4, the substitution pattern has differing
effects. The pyriazine moiety in compound 3 strongly stabilizes its LUMO energy, providing a
localized electron deficient aromatic ring which is easy to reduce. This stabilizes the LUMO by
Koopmann’s Theorem. The two pyridines in compound 4, however, are each harder to reduce
than pyrazine, and so its LUMO is less stabilized than compound 3 with respect to compound 2.
Iridium complexes with compounds 2 and 3 as cyclometallating ligands are known. a bis
cyclometalated iridium complex using 2, (2)2Ir(acac), phosphoresces green (λ em = 540 nm)
whereas the analogous derivative using 3 emits red (λ em = 640 nm). In the case of 3, stabilizing the
LUMO relative to 2 lowers the energy of the
3
MLCT state. On the other hand, the nitrogen in 4
that is para to the carbon metal bond should stabilize the metal d orbitals, thus stabilizing the
HOMO and making the complex harder to oxidize. Since the nitrogen in this position has a
minimal stabilizing effect on the LUMO, the net result is an increase in the HOMO-LUMO gap
and destabilization of the
3
MLCT state. This effect has been demonstrated in Ir complexes
cyclometallated with 2,3’ bipyridine ligands. Lastly, because 4 has two inequivalent sites for
cyclometallation, the coordination regiochemistry of 4 must be controlled with a substituent that
blocks the nitrogen that is not intended to participate in cyclometallation. The target structure, Bzp,
is therefore a derivative of 4 with an R group in the 12 possition Unfortunately only unsubstituted
compound 4 is known, synthesized from diaza-naphthalene by a double Skraup quinoline synthesis
in 15% yield. It should be noted that the reported compound is characterized only by combustion
elemental analysis and as bright yellow in appearance, whereas the pure compound is in fact
colorless (vide infra). This leaves some doubt as to the isomeric specificity of the double Skraup
approach.
21

The initial synthetic route to Bzp followed the same methodology as the known synthesis
of compound 4. Starting from readily available 1,8-naphthalic anhydride, nitration in the 3-
position occurred under standard conditions. Subsequent decarbonylative mercuration by HgO in
refluxing acetic acid yields a highly insoluble mercury compound, which, upon treatment with HI,
yields 3-nitro-8-iodo-1-naphthoic acid. Reduction of the nitro group to amine, and Curtius
rearrangement of the carboxylic acid to the
amine yields 1,3-dinitro-8-iodonaphthaline.
This compound is analogous to the reported
synthetic intermediate to compound 4, with an
iodine as place-holder to subsequent blocking
R group. Unfortunately, heating this
intermediate diamine in glycerol in the
presence of sulfuric acid and either arsenic
oxide or 3-nitrobenzoic acid as oxidant lead to
a complex mixture of products, from which
the desired product could not be identified by mass spectrometry or NMR spectroscopy.
A seemingly easier route to a 9,12- substituted derivative was then explored, employing a
4+2 cycloaddition as the key step. In this case, we attempted to trap transiently generated 1,7-
phenanthryne with 2,4-dimethylfuran. This route is incapable of regiospecifically installing the R
group (in this case, methyl) and so two methyl groups would be installed. Commercially available,
though not readily available, 1,7-phenanthroline was brominated in the 6- position under
electrophilic halogenation conditions. Generally, ortho-deprotonation of halo aromatics with
lithium bases leads to expulsion of lithium halide and the highly reactive aryne. Regrettably,

Figure 2-2. Retrosynthetic analysis of Bzp

22

treatment of this bromophenanthroline with lithium diisoproylamide resulted in decomposition of
the compound. NMR analysis of the complex product mixture indicated the presence of many
aliphatic proton resonances, suggestive of pyridine ring opening reactions.
A new route comprised the synthesis of substituted benzoquinoline from monocyclic
starting materials. This relied on the desymmetrization of methyl groups on a meta-xylyl group,
with on methyl group participating in a Dieckmann condensation to produce the b- ring, and the
other acting as blocking R group in the final product. The commerciallly available
2-bromonicotinic acid can be converted into the corresponding diethylamide by refluxing in
thionyl chloride to activate the electron poor carboxylic acid, followed by treatment with
diethylamide. A palladium catalyzed Suzuki-Maurya reaction then installed a 2,6-dimethylphenyl
group to yield a highly substituted phenylpyridine. Exposure to one equivalent of lithium
diisopropylamide in tetrahydrofuran cleanly provides the Dieckmann product,
5-hydroxy-10-methyl-benzo[h]quinoline. The hydroxy group in this intermediate is situated
where a nitrogen shall be as part of a pyridine moiety in the desired product. Thus, a Smiles
rearrangement was attempted to convert the hydroxy group to an amine. Treatment with
2-chloroacetamide in DMSO cleanly alkylated the phenolic position. However, subsequent
treatment with strong base such as sodium hydride was unproductive. Presumably, the 5- position
of benzo[h]quinoline is too electron rich to stabilize the intermediate Meisenheimer “ate-complex”
in the Smiles rearrangement.
A synthesis from an intermediate orthoquinone was then attempted. The retrosynthetic
analysis is shown in figure 2.2. The formation of the d- ring by a Boger pyridine synthesis
represents the key retrosynthetic disconnection. Because the Skraup quinoline synthesis had
previously been unsuccessful in generating the d- ring pyridine, we sought an easier-to-synthesize
23

heterocycle to then transmute into pyridine. This approach employs the 4+2 cycloaddition of an
electron rich alkene to an electron poor 1,2,4- triazine in an inverse electron demand Diels-Alder
reaction. Dinitrogen is then expelled to rearomatize the six membered ring. In the case of
transmuting triazine to pyridine, the dienophile is ostensibly acetylene. However, acetylene is too
electron poor to participate in an inverse electron demand Diels Alder reaction, not to mention its
volatility. Therefore, norbornadiene, the cycloaddition product of acetylene and cyclopentadiene,
was used as an electron rich acetylene precursor, proceeding to the desired pyridine product via
one inverse electron demand Diels-Alder reaction and two retro-Diels-Alder reactions in one pot.
Several oxidation conditions exist for the oxidation of phenols such as the previously
synthesized 5-hydroxy-10-methyl-benzo[h]quinoline to orthoquinones. Peroxides are generally
not compatible with the presence of amines or pyridines, since they produce the corresponding N-
oxides. In this case, however, the methyl group in the 12- position sterically blocks reactivity at
the pyridinic nitrogen, protecting it from undesired oxidative side reactions. Thus, a peroxide-like
oxidation proved most effective, employing dimethyldioxirane, nascently generated from acetone
and potassium persulfate. The orthroquinone was condensed with formamidhydazone to afford a
regiochemical mixture of two phenanthrotriazines as a brick red precipitate. This mixture was not
easily separated and therefore carried through without further purification. Treatment with excess
norbornadiene in triethyleneglycol cleanly afforded a mixture of me-Bzp and its structural isomer,
iso-me-Bzp, which were separable by chromatography.
Because the Boger pyridine synthesis lacks regiospecificity, lowering synthetic yield and
hampering product purification, a regiospecfic modification was explored where
5-amino-10-methyl-benzo[h]quinoline would be produced directly from the Dieckmann
cyclization rather than the previously unsuccessful Smiles rearrangement. In this case, the starting
24

2-bromo-nicotinamide was replaced by 2-chloro-nicotinnitrile. Unfortunately, the chloro
derivative was unreactive under Suzuki-Miyura conditions, presumably due to steric demands and
the less reactive halogen. Therefore, the commercially available 2-chloro-nicotinnitrile was
halogen exchanged to 2-iodo-nicotinnitrile via an aromatic Finkelstein reaction in the presence of
potassium iodide promoted by anhydrous hydrochloric acid. With the iodonicotinnitrile in hand,
a palladium catalyzed Suzuki-Maurya reaction with 2,6-dimethylphenylboronic acid yielded the
highly substituted phenylpyridine analogous to above. Unfortunately, many attempts to promote
the desired Dieckmann cyclization under basic condtions were not productive.
This initial route relies on the de-symmetrization of m-xylene to install a methyl group as
the blocking substituent, but is not appropriate for other blocking substituents such as phenyl which
are useful for conferring solubility to the product. Thus, a new route to Bzp was devised based on
a directed metalation strategy. The necessity of the blocking group to prevent unwanted metalation
regiocehmistry in the final product inspired a desired metalation reaction to install such a group.
Palladium, ruthenium, and cobalt catalyzed methods are known for 10-position functionalization
of benzo[h]quinoline. This would proceed from 5-hydroxybenzo[h]quinoline, which is not a
previously known compound. In order to produce this intermediate, 1-methoxynaphthalene was
nitrated to yield 1-methoxy-4-nitro-naphthalene. This reaction does not occur under standard
conditions with sulfuric acid as solvent, due to decomposition. Indeed, treatment of
methoxynaphthalene with sulfuric acid, even at low temperature, produces intractable black
polymerization products. Instead, treatment of 1-methoxynaphthalene with one equivalent of
nitric acid in a solvent of acetic anhydride yields the desired 1-methoxy-4-nitro-naphthalene. Care
must be taken to maintain a temperature below 20C during aqueous quench to prevent
demethylation to the highly water soluble 1-hydroxy-4-nitro-naphthalene, which, although
25

potentially competent to be carried forward in subsequent reactions, is difficult to handle and
purify. Reduction of 1-methoxy-4-nitro-naphthalene with one atmosphere of hydrogen gas in
methanol in the presence of catalytic palladium supported on carbon cleanly affords
1-methoxy-4-amino-naphthalene, which, unlike the hydroxy derivative, can be purified by
precipitation from methanol/ether with hydrochloric acid to form its corresponding hydrochloride
salt. This naphthalamine hydrochloride is poised for a Skraup quinoline synthesis of
5-hydroxybenzo[h]quinoline. However, the strongly acidic conditions of sulfiric acid in hot
glycerol lead to intractable decomposition of the starting material, so a milder quinoline synthesis
was sought. The Doebner-Miller modification of the Skraup quinoline synthesis was found to be
effective. Treatment of 1-methoxy-4-amino-naphthalene hydrochloride with three equivalents of
acrolein in tetrahydrofuran and 0.5 equivalents of concentrated hydrochloric acid gave a modest
yield of 5-hydroxybenzo[h]quinoline after silica column chromatography.
In this case, Sanford’s palladium acetate catalyzed halogenation conditions were chosen to
install a halogen as a place holder for installation of the blocking substituent as the last step of the
ligand synthesis. Treatment of 5-hydroxybenzo[h]quinoline with one equivalent of
n-bromosuccinimide and 0.1 equivalents of palladium acetate in acetonitrile at 50C did produce
some of the desired 5-hydroxy-12-bromo-benzo[h]quinoline, but also included significant
amounts of the 6-bromo product expected from the uncatalyzed reaction. After attempted
optimization of conditions failed to achieve regiospecificity for 12- position bromination, a one
pot oxidation to 12-bromo-benzo[h]quinoline-5,6-dione was attempted with excess
n-bromosuccinimide. Unacceptably low yields of the desired product could be observed by NMR,
IR, and Mass Spectrometry, so the route was reluctantly abandoned.
26

Figure 2.3 shows the improved
synthesis of Bzp. The starting material,
10-chlorobenzo[h]quinoline is prepared from
benzo[h]quinoline by the literature reported
method on up to 25g scale in excellent yield.
This compound can then be oxidized to the
corresponding orthoquinone in modest yield
using iodic anhydride as oxidant. Other
oxidants were explored, including
dimethyldioxirane,
tertbutylhydrogenperoxide, sodium perchlorate, bromine, nitric acid, and chromic acid, all of
which lead to either very poor conversion or decomposition. Condensation of this diketone with
formamidhydrazone proceeds in the same fashion as the original route, leading to a 3:2 mixture of
the desired triazine intermediate to the undesired isomer. The isomeric mixture is carried through
in the next transformation without separation. The cycloaddition Boger pyridine synthesis
conditions of the original route in triethyleneglycol as solvent efficiently produces
benzo-1,7-phenanthroline, unfortunately with complete dehalogenation of the 12- position. Use
of the non-protic solvent orthodichlorobenzene, however, produces the desired
12-chloro-benzo-1,7-phenanthroline with only traces of hydrodehalogenation products.

Figure 2-3. Synthesis of Bzp (5)

27

Standard Suzuki-Miyura cross coupling conditions with phenylboronic acid, catalyzed by
tetrakistriphenylphosphine palladium(0), do not produce the desired product, presumably by to
arrest of the catalytic cycle due to the stability of the intermediate benzophenanthroline
palladacycle. This unusual behavior was noted by Weimar et. al. and remedied by the use of a
palladium(II) catalyst in the presence of
oxidant to turn over the catalytic cycle.
Indeed, a good yield of cross coupling product
is obtained when palladium acetate and
triphenylphosphine oxide are employed as
catalyst. Interestingly, the yield of this
reaction is higher when the halogen is chlorine
rather than bromine, unlike standard
Suzuki-Miyura conditions.
In order to assess its use as a ligand, 5
was then platinated similarly to the previously
reported 2 to form its analogous platinum
dipivaloylmethane complex, 5Pt(dpm).

Figure 2-4. Synthesis of 5Pt(dpm) and crystal structures of 5
and 5Pt(dpm). ORTEP structures shown at 50 %
probablility level

28

The x-ray crystal structure of 5 reveals a 69° dihedral angle between the phenyl group and the
a-ring to which it is bound. This angle is 65° in 5Pt(dpm). Furthermore, the triphenylene core of
5 is slightly twisted, with a 10° angle between the mean plane of the a-ring and that of the c- and
d- rings [9° in 5Pt(dpm)]. The photoluminescence spectrum of 5 at 77 K (figure 2.5) shows both
fluorescence (E00 = 340 nm) as well as phosphorescence (E00 = 425 nm), the latter corresponding
to a triplet energy of 2.9 eV. As expected from our earlier analysis, iso-5 has a similar triplet
energy to 5 but lacks the specific nitrogen substitution pattern leading to sky-blue phosphoresce.
Triplet energies for 5Pt(dpm) and iso-5Pt(dpm) are predicted to be 2.75 eV and 2.48 eV,
respectively, so iso-5Pt(dpm) was not pursued. At 77 K in frozen MeTHF, 5Pt(dpm) exhibits sky
blue phosphorescence (E00 = 458 nm, 2.71 eV,  = 12.1 s), 70 meV higher in energy than its
isoelecronic parent, 2PtDpm (E00 = 470 nm, 2.64 eV). Unfortunately, 5Pt(dpm) is non-emissive
in fluid solution at room temperature, likely
due to structural distortions in the luminescent
excited state leading to other, non-emissive
states. Indeed, the complex becomes emissive
when dispersed (0.5 wt %) in rigid
poly(methylmethacrylate) (PMMA) matrix
( = 0.11). By inspection of the crystal
structures of 5 and 5Pt(dpm), we hypothesized
that partial rotation of the phenyl group may
constitute a mode of deactivation which is not
completely restricted in PMMA. Therefore,
we expected an increased quantum yield from

Figure 2-5. Absorption and emission spectra of (a) 5 and iso-
5; (b) 5PtDpm; (c) 10PtDpm. All absorption spectra were
collected in methylcyclohexane at 298 K. MeCy =
methylcyclohexane; MeTHF = 2-methyltetrahydrofuran;
PMMA = polymethylmethacrylate

0.0
0.5
1.0
300 350 400 450 500 550 600
(c)
(b)

Normalized intensity (arbitrary units)
iso-5 77 K (MeCy)
iso-5 absorption (MeCy)
5 77 K (MeCy)
5 absorption (MeCy)
(a)
0.0
0.5
1.0
absorption (MeCy)
77 K MeTHF
300 K PMMA
300 350 400 450 500 550 600
0.0
0.5
1.0
absorption (MeCy)
Wavelength (nm)
77 K MeTHF
300 K PMMA
300 K MeCy
29

a rotationally restricted 2,6-dimethylphenyl substituted derivative, 10Pt(dpm). Synthesis of the
ligand, 10, required only a trivial variation of the last step in the production of 5, illustrating the
versatility of our synthetic approach. Gratifyingly, this simple modification resulted in a four-fold
increase in quantum yield in PMMA ( = 0.46). Furthermore, luminescence from 10Pt(dpm) can
be observed in fluid solution ( = 0.01,  = 118 ns). It is notable that 10Pt(dpm) shares the same
triplet energy as the analogous difluorophenylpyridine complex, (F2ppy)Pt(acac) (2.71 eV), with
similar radiative and non-radiative rate constants in solution.
The promising properties from the platinum complex confirmed the design strategy for the
Bzp ligand system, so syntheses of various iridium complexes were then sought. By far the most
common strategy for cyclometallation of phenylpyridines with iridium occurs by the Nonoyama
reaction. The reaction conditions consist of two equivalents of ligand, one equivalent of iridium
(III) chloride hydrate, and a 4:1 solvent mixture of 2-ethoxyethanol and water heated to 140° C
overnight. In most cases, a poorly soluble precipitate forms which can be collected by filtration
and washed to yield a product with empirical formula L2IrCl, where L is the cyclometallating
ligand. The by-product of this reaction is two equivalents of hydrogen chloride. In uncommon
cases where L is sterically demanding, this mononuclear species may exist as a 5-coordinate
iridium complex, or as a 6-coordinate solvento species in the presence of coordinating solvent or
exogenous strongly coordinating ligands such as isocyanides. Otherwise, this composition exists
as a chloride bridged dimer, namely u-dichloro-bis(L) diiridium (III), also known as the Nonoyama
dimer. The kinetic stability of the Nonoyama dimer drives its formation to occur in high yields
with exquisite coordination isomeric purity. The strong trans-influence of carbon ligands (i.e.
phenyl) causes them to be cis- to each other and trans- to the chlorides. As a result, the nitrogen
ligands (i.e. pyridine) are trans- to each other and the coordination environment around each metal
30

center is C2 symmetric (there is a diastereomeric mixture of delta- and lambda- metal centers in
the dimer). Therefore, this compound is a valuable and often utilized synthetic intermediate.
The Nonoyama dimer tends to be weakly photoluminescent, if at all, due to the weak ligand
field strength of chloride leading to non-emissive metal centered excited states. Replacement of
chloride with stronger ligand field strength ligands can produce highly emissive heteroleptic
complexes of the formula L2IrX, where L is the cyclometallating ligand. For neutral complexes,
X is a monoanionic bidentate ligand, often acetylacetonate or picolinate. Ligand X is often referred
to as ancillary, though as discussed in chapter 1, it may have a profound effect on the photophysical
properties of the resulting heteroleptic complex. For homoleptic complexes of the composition
L3Ir, the Nonoyama dimer may be treated with one equivalent of L. Due to the kinetic stability
of the bis-cyclometallated iridium chloride intermediate, a soluble silver salt is added to abstract
chloride, thereby activating the compound to a third cyclometallation event. The usual conditions
consist of Nonoyama dimer, one equivalent of L, and silver triflate in refluxing dichloroethane
(67° C). The trans- arrangement of pyridines in the Nonoyama dimer is retained at this relatively
low temperature, resulting in a meridonal heteroleptic compound, which is the kinetically but not
thermodynamically favored product. Meridonal heteroleptic complexes are generally thermally
stable, even surviving sublimation in high yield with only trace isomerization to the
thermodynamic facial isomer. Generally, meridonal homoleptic complexes are susceptible to
photoisomerization, and irradiation of these compounds with long wave ultraviolet light (ca.
350 nm) in a coordinating solvent such as acetonitrile quantitatively converts the compound to the
facial isomer.
Subjecting Bzp to standard Nonoyama conditions resulted in a bright yellow
precipitate as expected, however it was completely insoluble in common organic solvents. No
31

proton NMR resonances could be observed for samples prepared in CDCl3, DMSO-d6, acetone-d6,
or methanol-d4. Furthermore, any attempts to elaborate this intermediate with ancillary ligands in
the presence of silver triflate resulted in no reaction, leading us to believe that the product was not
the Nonoyama dimer and rather another species henceforth referred to as the abnormal Nonoyama
product. It was presumed that the non-coordinated (distal) nitrogen, though sterically blocked by
an aryl group, could be protonated by the hydrogen chloride generated in the Nonoyama reaction.
The resulting salt may either arrest the reaction after the first cyclometallation event, or render the
Nonoyama dimer kinetically inert. Disappointingly, addition of an acid scavenger such as
potassium carbonate or pyridine to the Nonoyama reaction did not produce a different result, nor
if added to subsequent reactions using the abnormal Nonoyama product of Bzp.
Different Ir metalation conditions were then attempted which did not include the formation
of HCl. Following Crabtree’s work with oxidative additions of Ir (I) complexes to form
cyclometallated compounds, [(COD)IrCl]2 (COD = 1,5-cyclooctadiene) was selected as a starting
material. Due to the precious nature of Bzp, benzo[h]quinoline (Bzq) was selected as a model
compound. Although Bzq does in fact react in the normal way under Nonoyama conditions, a
non-Nonoyama type route to a similar product was sought. Activation of [(COD)IrCl] 2 by silver
hexafluorophosphate in acetonitrile yielded the bright yellow (COD)Ir(NCMe)2 PF6. Treatment
of this iridium compound with Bzq in room temperature deuterated acetonitrile rapidly produced
the cyclometallated (Bzq)(COD)(H)(NCMe-d3)Ir. The presence of the hydride was confirmed by
its characteristically upfield NMR shift (delta = -15.5 ppm). Treatment of this compound with a
second equivalent of Bzq does not lead to further cyclometallation reactions, and the solution
darkens over the course of hours with the formation of unidentified decomposition products.
Hoping to affect a reductive deprotonation to be followed by a second oxidative addition of Bzq,
32

(Bzq)(COD)(H)(NCMe-d3)Ir was treated with one equivalent of triethylamine. Unfortunately,
this resulted in immediate formation of a black precipitate which could not be identified.
Rather than rely on oxidative addition to a CH bond, producing a hydride, oxidative
addition to a CBr bond was explored. Due to its use as a synthetic intermediate in the production
of Bzp, 10-bromobenzo[h]quinoline (BrBzq) was readily available. Treatment of [(COD)IrCl]2
with two equivalents of BrBzq yielded two equivalents of (Bzq)(COD)(Cl)(Br)Ir as an air stable
yellow solid. An attempt was made to exchange (COD) and both halogens with two
acetylacetonate (Acac) ligands.
Treatment of (Bzq)(COD)(Cl)(Br)Ir with
excess NaAcac in refluxing
dichloroethane yielded
(Bzq)(COD)(Acac)2Ir, whose NMR
spectrum displays 25 unique proton
resonances. Figure 2.6 shows the
1
HNMR spectrum, a proposed structure
for the compound, and tentative assignment of the proton resonances. It was assumed that heating
the material under vacuum would expel COD, leading to (Bzq)(Acac)2Ir, a useful synthetic
intermediate and potentially interesting compound in of itself. Surprisingly, at 250 C and 10
-6
torr,
(Bzq)(COD)(Acac)2Ir sublimed cleanly without a change in its NMR spectrum. At this point,
further attempts to elaborate the oxidative addition products of cyclometallating ligands to Ir(I)
precursors was abandoned in favor of a modified Nonoyama procedure.
Combustion elemental analysis (CHNS) of the abnormal Nonoyama product was
consistent with a composition of BzpIrCl2∙HCl∙H2O. Several structures were hypothesized for

Figure 2-6. Proton NMR spectrum of (Bzq)(COD)(Acac)2Ir

33

this composition, and although none were directly observed, all presumed a protonated,
mono-cyclometallated BzpIr complex. The rest of the coordination environment around iridium
could either be trichloro aquo, the insolubility and inertness of which would be somewhat
surprising, or a monohydrate of a protonated linear u-chloro coordination polymer, accounting for
the insolubility. Gratifyingly, treatment with triethylamine in acetonitrile-d3 produced a proton
NMR spectrum consistent with a cyclometallated Bzp complex, though the ligand stoichiometry
with respect to iridium was not readily discernable. LCMS analysis of this triethylamine treated
sample clearly indicated the presence of BzpIrCl2(NCMe)2 as its M+H ion.
Rather than recover and elaborate the abnormal Nonoyama product towards Bzp2IrX and
Bzp3Ir complexes, a one pot modification of the Nonoyama conditions was attempted in the
presence of various bases. The reaction was monitored visually, and assessed by liquid
chromatography tandem mass spectrometry after 16 hours. Without added base or in the presence
of potassium carbonate, the reaction proceeded as shown photographically in figure 2.7. The initial
olive-green color due to IrCl3
changed to pale yellow within
minutes. The homogeneous
solution then deepened in
color to orange before
precipitating a bright yellow
solid. Liquid chromatography tandem mass spectrometry analysis of the solution indicated the
presence of excess Bzp ligand and traces of the abnormal Nonoyama product as its acetonitrile
(chromatography mobile phase) adduct. The synthetic procedure was then repeated in the presence
of triethylamine. After the initial color change in minutes from green to yellow, the solution

Figure 2-7. Visual progression of the modified Nonoyama reaction

2 M NaOH
34

darkened to a turbid black suspension, presumably due to precipitation of either iridium metal or
iridium oxide. LCMS analysis indicated only the presence of Bzp ligand in solution. Two
mechanisms could lead to this undesired reaction. Triethylamine may have acted as reducing agent
to convert iridium chloride to iridium black. Alternately, triethylamine in protic solvent was in
equilibrium with triethylammonium hydroxide which in turn converted iridium chloride to iridium
hydroxide, ultimately leading to insoluble iridium oxide. In order to test the reducing agent
hypothesis, a less reducing (though more basic) amine without alpha protons was selected: tertiary
butyl amine. Repetition of the reaction conditions with t-BuNH2 resulted in a similar darkening
of the reaction without product formation, although more rapidly than in the case with
triethylamine. This suggests that the presence of hydroxide, rather than reducing agent, leads to
the precipitation of the iridium precursor before the reaction can take place.
Because a protic solvent is required to solubilize IrCl3 and a base as strong or stronger than
triethylamine is required to deprotonate the abnormal Nonoyama product of Bzp, the presence of
hydroxide cannot be avoided. Therefore, a one-pot stepwise procedure was employed in which
the abnormal Nonoyama product was allowed to form as a heterogeneous suspension, followed by
the addition of base to promote the completion of the reaction. Addition of two equivalents of 1 M
aqueous sodium hydroxide to the Nonoyama reaction after 16 hours immediately produced a dark
black suspension, presumably due to the formation of iridium oxide from unreacted iridium
chloride. Gratifyingly, after being allowed to react at reflux overnight, the solution was observed
by liquid chromatography tandem mass spectrometry to contain the usual Nonoyama product
(Bzp2IrCl) as the major component. Aside from unreacted ligand, the heteroleptic complex Bzp3Ir
was detected as a minor side product. This reactivity is observed regardless of blocking
substituent, be it phenyl- (phBzp) or xylyl- (xyBzp).
35

With phBzp2IrCl and xyBzp2IrCl in hand, the heteroleptic picolinate complexes analogous
to Firpic were synthesized in good yield by treatment with picolinic acid, potassium carbonate,
and silver triflate. Both phBzp2IrPic and xyBzp2IrPic were purified by silica column
chromatography with a mobile phase of acetone. Figure 2.8 shows the emission spectra of the two
compounds in de-aerated 2-methyltetrahydrofuran at room temperature and in a matrix of frozen
2-methyltetrahydrofuran glass at 77 K. The absorption and low temperature emission spectra are
nearly identical both compounds. The emission spectra are featured with a clear vibronic
progression, somewhat similar to firpic, with an emission maximum at 460 nm, corresponding to
a triplet energy (ET = 2.7 eV).

The compounds differ significantly in room temperature fluid solution. In the case of the
phenyl derivative, the second vibronic peak is higher than the first, suggesting a high degree of
structural distortion in the excited state. Furthermore, its photoluminescent quantum yield of 5 %
is unexpectedly low for this type of complex. Its excited state lifetime is biexponential in fluid
solution (158 ns, 36 %; 861 ns, 64 %), which increases dramatically in frozen solvent matrix
350 375 400 425 450 475 500 525 550 575 600
0.0
0.5
1.0

Intensity (A.U.)
wavelength (nm)
Abs
77K
RT

Figure 2-8. Absorption and emission spectra of phBzp2IrPic (left) and xyBzp2IrPic (right)

350 375 400 425 450 475 500 525 550 575 600
0.0
0.5
1.0
Intensity (A.U.)
wavelength (nm)
Abs
77K
RT
36

(3.9 us, 61 %; 9.6 us, 39 %). This is consistent with the structural distortion of phBzp observed
for its platinum complex, phBzpPt(dpm).
The conformationally restricted derivative, xyBzp2IrPic, behaves as desired. The bright
yellow compound exhibits efficient sky blue photoluminescence in solution with a
photoluminescent quantum yield of 80 %. The room temperature fluid solution emission spectrum
is somewhat broadened from the featured low
temperature spectrum, but appears to have features
which correspond to the same vibronic
progression. The CIE coordinates of the room
temperature emission are (.14, .33), representing
an acceptable sky blue component for a white or
warm white OLED. Furthermore, its excited state
lifetime is 2.9 us, corresponding to a radiative rate
constant of 2.8x10
5
s
-1
. At 77 K, this lifetime increases to 4.8 us, corresponding to a radiative rate
constant equal to or less than 1.7x10
5
s
-1
. This two-fold decrease in radiative rate upon cooling is
consistent with high spin orbit coupling and a triplet manifold zero field splitting on the order of
100 cm
-1
.
Electrochemical oxidation and reduction potentials were measured by cyclic voltammetry
in dimethylformamide with 0.1 M nBu4N PF6 against an internal ferrocene reference. A reversible
oxidation wave was observed at 0.82 V vs. Fc/Fc
+
, corresponding to a HOMO value of -5.75 eV
by an empirical correlation. Similarly, a quasi-reversible reduction wave was observed at -2.23 V
vs. Fc/Fc
+
, corresponding to a HOMO value of -2.13.
Table 2-2. Comparison of relevant properties of
xyBzp 2IrPic and Firpic

xyBzp 2IrPic Firpic
E red (V vs. Fc) -2.23 -2.29
E ox (V vs. Fc) 0.82 0.92
HOMO (eV) -5.75 -5.89
LUMO (eV) -2.13 -2.05
CIE (.14, .33) (.16, .29)
PLQY 80 % 80 %
t (RT) 2.9 us 1.9 us
T sublime Decomp. 300C

37

Figure 2.9 compares xyBzp2IrPic to Firpic across several parameters relevant to OLEDs.
By inspection, the two compounds are very similar. A notable exception is that xyBzp2IrPic
decomposes at temperatures above 350C without sublimation. This precludes vacuum deposition
of devices for operational lifetime testing of xyBzp2IrPic doped OLEDs.
Facial (fac-) and meridional (mer-)
isomers of the homoleptic xyBzp3Ir
complexes were recovered from the
synthesis of xyBzp2IrPic as minor
impurities, carried through from the
modified Nonoyama conditions used to
produce xyBzp2IrCl. The meridonal isomer
makes up about 10 % of the material, while
the facial isomer is a trace component.
Figure 2.10 shows the emission spectra of
both compounds in room temperature methyltetrahydrofuran solution and 77 K frozen solvent
matrix. The meridional isomer exhibits a bathochromic shift from the facial isomer. This is
consistent with those two isomers of homoleptic Irppy3. The effect is less pronounced at low
temperature and more pronounced in fluid solution. Unlike meridional Irppy3, the meridional
isomer of xyBzp3Ir is highly luminescent, with a quantum yield of 80 %. Notably, the excited
state lifetime at room temperature of 3.1 us rises to 12 us at 77K. This marked decrease in radiative
rate, between three and four fold, is too large to be accounted for by the thermal Boltzmann
population of a single triplet manifold with reasonable zero-field splitting. Instead, this change in
radiative rate as well as line shape is more consistent with a thermal equilibrium between more
450 475 500 525 550 575 600 625 650 675 700
0.0
0.5
1.0

normalized intensity
wavelength (nm)
fac- 77K
fac- 300K
mer- 77K
mer- 300K

Figure 2-9. Emission spectra of facial (black) and meridonal
(red) xyBzp3Ir in room temperature methyltetrahydrofuran
solution (solid trace) and 77 K frozen solvent matrix (dashed
trace)

38

than one closely lying triplet manifold. Because the three Bzp ligands of the meridional isomer
are each distinct, as seen by their unique proton NMR resonances, it stands to reason that triplet
spin densities on each represent non-degenerate triplet states. The observed line shapes and excited
state decay times at each temperature are therefore assigned as follows: At low temperature, the
lowest lying of three triplet manifolds emits as a sharp featured band, with red-shifted onset (it
being the lowest energy state) but blue shifted maximum (it being localized and therefore narrowed
with vibronic features. One the other hand, at room temperature, a thermally activated process
populates higher lying triplet states corresponding to spin density on the other ligands, leading to
increased radiative rates, a blue shifted onset (populating higher energy states) and broadening
(being delocalized over all of the ligands). The facial isomer, however, does not have inequivalent
ligands, and does not exhibit a dramatic change in lifetime or lineshape upon cooling, futher
supporting the assignment of the nature of the meridional isomer.
In conclusion, an aza-substitution strategy for blue shifting phenylpyridine based ligands
was successfully employed. Benzophenanthroline, a fluorine free tetracyclic heterotriphenylene
ligand, exhibits efficient sky blue phosphorescence with similar properties to the ubiquitous
difluorophenylpyridine ligand (Firpic), without the potential for dehalogenative degradation
mechanisms. A synthetic route was devised which allowed for late stage variation of a necessary
blocking substituent. The versatility of the synthetic route was validated in exploration of a
platinum complex using the ligand. A blocking phenyl substituent was observed to have an
unexpected and undesired electronic effect due to excited state structural distortions. A xylyl
blocking substituent avoided the undesired distortion and produced an efficient sky blue emitting
platinum complex. Attention was then turned to iridium complexes, in particular the complex
analogous to the ubiquitous sky blue emitting phosphor Firpic. A synthetic challenge was
39

encountered during the metalation step in which standard Nonoyama conditions produced an
abnormal monocyclometallated product. Modified conditions allowed for the synthesis of the
standard Nonoyama product and further elaboration to the bis-ligand iridium picolinate complex.
By comparison to Firpic, the aza-substitution para to the site of metalation, along with
rigidification of the ligand frame with triphenylene, was observed to have a similar electronic
effect to fluorination. Bzp2IrPic is similar to Firpic in triplet energy, quantum yield, and
electrochemical potentials. While its thermal instability renders Bzp2IrPic incompetent for
vacuum processing of a doped OLED, the instability is most likely due to the picolinate ancillary
ligand. Moving forward, a more thermally stable electron withdrawing ancillary ligand may yield
a useful dopant for OLEDs with long operational lifetimes.

40

CHAPTER 3 - COMPUTATIONAL METHODS
Note: Much of the work in this chapter was conducted with coworkers Daniel Sylvinson M.
R., Hsiao Fan Chen, Lauren Martin, and Mark E. Thompson. See Sylvinson et. al. ACS Appl.
Mater. Interfaces. 2019 11 (5), 5276-5288. DOI: 10.1021/acsami.8b16225
Elucidating structure-property relationships empirically requires a series of related
compounds, or a structure class, which has been characterized. The time and resources required
in the synthesis of new compounds, however, represents a bottle neck in the process.
Computational methods offer an alternative, often complimentary approach for examining a set of
compounds. In particular, high-throughput combinatorial screening is standard practice in
medicinal chemistry. Lipinski’s ‘Rule of Five’ has canonized qualitative molecular design
principles in drug discovery, but is there a 'Rule of Five' for organic semiconductors? Organic
materials chemists still struggle to find axioms to define 'semiconductor-like' compounds to the
extent that medicinal chemists define 'drug-like' compounds. Computational methods offer
synthetic chemists an insight into structure-property relationships which are not feasibly arrived at
experimentally, but we lack an agreed upon set of best practices in computational materials
discovery. Rather than searching for a needle in a haystack by calculating properties in a large
chemical space, we can inform our chemical intuition by thoughtfully screening a smaller chemical
space. This chapter covers the use of computational methods to inform the chemist's intuition in
organic semiconductor materials discovery.
A large-scale computational screen known as the “mega-run” that was undertaken to search
for organic photovoltaic materials is an informative case study for the successes and failures of
high throughput computing. In that study, six properties were predicted for each of over 200
thousand compounds with the goal of finding promising candidates for electron acceptor materials.
41

A record amount of CPU resources was used, amounting to 240 compute years, generating a
massive data set.
The mega-run follows a previous
small scale study by Halls, et. al. involving a
screen of aza-substituted pentacenes with the
goal of finding potential electron accepting
materials. The search criterion was to
identify those compounds with calculated
LUMO energies near or below that
calculated for fullerene C60, a ubiquitous
electron acceptor. The strategy of isoelectronic transmutation of sp
2
hybridized CH fragments for
more electronegative sp
2
hybridized N fragments discussed in Chapter 2 was employed here to
attempt to stabilize the LUMO energy of pentacene. Only those structures with four nitrogen
atoms were considered, and all 135 tetra-aza-pentacene isomers were enumerated. Figure 3.1
shows the predicted LUMO energies plotted versus an arbitrary structure index and color coded
by dipole moment. The four tetra-aza-pentacene structures shown are those with LUMO energies
predicted to lie below -3.66 eV, the calculated value for fullerene C60. By inspection, no obvious
structural motif is common to these compounds, other than a pyrazine in one ring, though this
motif is present in several other compound which do not meet the search criterion. The megarun
sought to dramatically increase the size of the library in order to uncover emergent patterns.
With the goal of enumerating a library with a size of approximately one million members,
acepentacene was selected as a parent structure. Aceannulation of acenes is known to stabilize
LUMO energies, and in the case of pentacene, will desymmetrize the structure from D2d to CS.

Figure 3-1. Predicted LUMO energies of 135
tetraazapentacenes

-3.66 eV
-2.65 eV
1
2
3
4
42

This desymmetrization is operationally useful to prevent duplicate structures from being
enumerated which are related by a symmetry operation. Next, between zero and five of the
fourteen unique CH fragments were transmuted to N, resulting in 3,473 structures. Each was then
decorated with between zero and three phenyl groups, producing a 553,855 member structure
library. The following six properties were predicted for a random subset of 201,021 these
compounds: oxidation potential, reduction potential, triplet energy, electron reoganization energy,
hold reoganization energy, and static dipole moment. Figure 3.2 shows a plot of the predicted
oxidation potentials versus reduction potentials, color coded by triplet energy (increasing from
blue to red). The easier to oxidize, harder to reduce compounds lie is the lower left quadrant of
the plot, representing the electron donor type materials. Conversely, the upper right quadrant
contains the electron acceptor type materials. As expected, the parent structure lies in the donor
region, and the compounds trend towards to acceptor region with increasing nitrogen content. The
lower right quadrant of the plot contains compounds with decreased electrochemical gaps. The
most extreme few examples have predicted electrochemical gaps below 0.5 eV, and are considered

Figure 3-2. Predicted oxidation potential vs. reduction potential of all 201,021 'mega-run' azapentacenes

43

to be erroneous predictions. As expected, the compounds towards the lower right quadrant have
lower predicted triplet energies than those towards the upper left, consistent with the correlation
between triplet energy and electrochemical gap. The upper right quadrant is sparsely populated,
suggesting that aza-substitution does not significantly widen electrochemical gaps or, by
extension, blue-shift absorbtion in these compounds.
While these structure-property relationships can be read directly from the plot, they
come as no surprise and are well known for heteroaromatic compounds in general, not just this
structure class. Ultimately, the structure-property relationships which could be determined from
the data set were largely obvious even before the study was carried out, while others may remain
hidden to a human observer’s eyes. Machine learning offers a way to search large data sets by
pattern recognition, but the neural networks involved are somewhat of a “black box” and offer
little in the way of insight to chemists beyond blind predictive power.
High throughput searches for OLED materials are faced with similar problems. For
instance, a recent large-scale computational screening of thermally activated delayed fluorescence
(TADF) emissive OLED dopants by the Aspuru-Guzik group identified 20 high efficiency emitters
from a library of 1.6 million. While the search was successful in finding desirable compounds, it
can be noted that no structure-property relationships for TADF were established. Furthermore,
single molecule (gas phase) calculations are well applied only to dopant molecules, where
intermolecular interactions are minimized. Our approach seeks to generalize combinatorial
materials screening to include solid state interactions applicable to host materials. Our strategy is
the explore a diverse set of candidate structures the a tiered, multi-scale workflow by
understanding structure-property relationships at each tier.

44

As an example of our approach, we set out to find candidate structures for a chemically
robust host material capable of producing a high-performance sky-blue OLED. Among the
important design criteria for blue OLED host materials are high triplet energy, appropriate
HOMO/LUMO energy level alignment with respect to the emissive dopant, balanced carrier
transport, and robust molecular structure. Finding host materials that satisfy all of these criteria
for a given phosphorescent dopant is complex, because these properties are not usually
independently tunable with molecular structure. For our study, we chose to search for host
materials for the well-studied sky blue dopant FIrpic. The choice of this phosphorescent dopant
sets limits on the electronic properties for the host material. To achieve a high luminance
efficiency for a FIrpic dopant, the triplet energy of the host must be > 2.65 eV. In order to use the
FIrpic doped host material in an efficient OLED stack the LUMO of the host needs to be near or
above that of FIrpic, so the desired LUMO energies of the host are < -2.1 eV.
The structural starting point of our search for high triplet energy host materials is
triphenylene, which has a triplet energy in the deep blue range (ET = 425 nm, 2.9 eV) as well as a
large aromatic system, leading to efficient charge transport. Additionally, the rigid tetracyclic
structure inhibits one potential OLED degradation pathway involving exocyclic bond cleavage of
the host material, as is widely suspected for the standard phenylcarbazole derivatives. Due to the
synthetic challenges and structural limitations inherent to the synthesis of regiospecifically
substituted heterotriphenylenes, as discussed in Chapter 2, this report explores a related but more
synthetically accessible class of molecular structures in which one of the 6 membered rings of
triphenylene is replaced with a 5 membered heterocyclic ring, which can readily be formed from
orthoquinone precursors. This phenanthroheterole based structure is referred to as H2P, due to its
two peripheral hexagonal “H rings” and one pentagonal “P ring”.
45

Isoelectronic, heteroatomic transmutations of CH fragments of the parent
cyclopentaphenanthrene structure (Scheme 1) have qualitatively different effects whether they be
in the P- or H- rings. For neutral aromatic structures, the H-rings may include pnictogens (only N
was considered), while the P-ring may also include chalcogens (O and S were considered).
Because the chalcogens, which provide a large chemical space with potentially desirable
properties, are only available to the P- ring, our structure search was broken into two Tiers: Tier
1 selection focused on the heteroatoms of the P-ring heterocycle, while Tier 2 focused on aza
substitution in the H-ring, using the best candidates identified in Tier 1.
Tier 1 identified the candidates with triplet energies over 2.7 eV, based on a FIrpic dopant.
Furthermore, the selection strategy in Tier 1 involved maximizing the LUMO energy, rather than
optimizing it based on the LUMO of FIrpic. This strategy anticipates the aza substitution in Tier
2 which will categorically stabilize LUMO energies, so a destabilized LUMO energy from Tier 1
allows room for tunability through the desired range of LUMO energies in Tier 2
4a
.
The Tier 1 selection involved a survey of 15 P-ring heterocycles and the Tier 2 selection
involved 37 unique H-ring aza substitution patterns. By carrying out the selection processes
serially, choosing the best candidate in Tier 1 to carry into Tier 2, the number of structures that
needed to modeled dropped from 555 (15 * 37), corresponding to all H-ring substitutions for each
P-ring structure, to 52 (15 + 37) calculations necessary to survey the optimal chemical space of
H2P structures.
Tier 1 Selection
The first Tier selection involved incorporation of oxygen, sulfur and nitrogen for X into
the P-ring of the H2P framework to produce the library of chemically relevant structures shown in
46

Figure 3.3. DFT calculations were performed
at the B3LYP/MIDIX level for screening,
which is a low level of theory suited for rapid
screening. The triplet energies of S and O
substituted P-rings were lower than that of
N-substituted compounds, with non-chalcogen
containing compounds 10-15 having the
highest triplet energies. The highest triplet
energies were predicted for triazoles 12 and
13, which, while are promising candidates, fail
to meet the Tier 1 requirement of maximized
LUMO energy. Tier 2 N-substitutions of 12
and 13 would likely result in LUMO energies
below the desired range. The target structures
were therefore 5, 6, and 10, since they present
both high triplet and less negative LUMO energies. Compound 10 was selected because it and
substituted versions of it can be readily prepared from phenanthrene-9,10-dione or its
aza-substituted analogs, as illustrated in Scheme 2. It is noteworthy that while 6,6,6,5-membered
tetracyclic imidazo[4,5-f]-1,10-phenanthroline derivatives have been investigated as ligands in
phosphorescent emitters for OLEDs,
19
they have never been used as host materials for either
fluorescent or phosphorescent OLEDs. Anticipating a crystalline rather than glassy morphology
of vapor deposited films of planar 10, as well as potentially recalcitrant synthetic preparations,
substituent functional groups were chosen for the imidazole before Tier 2 selection. When R1 =

Figure 3-3. (top) Elemental enumeration of O, S, NH and CH
in the “P ring” of the H2P structure. (bottom) Lowest
unoccupied molecular orbital energy vs. triplet energy for
iterative cyclopentaphenanthrene structures.

2.4 2.6 2.8 3.0
-1.4
-1.2
-1.0
-0.8
-0.6 P-ring substitution
O and N,O
S and N,S
N only

LUMO (eV)
Triplet Energy (eV)
4
8
3
2
9
1
7
11
15
14
12
13
10
5
6
47

phenyl, the singlet and triplet energies are markedly red shifted, due to conjugation of the phenyl
group with the imidazole ring. For example, when R1 = R2 = H, the S1/T1 energies are predicted
by DFT to be 3.54/2.96 eV, while when R1 = Ph, R2 = H the S1/T1 energies are predicted to be
3.33/2.56 eV. Thus, R1 = H or alkyl is preferred for high triplet energy, however, in synthetic trials
we found that when R1 is methyl, ethyl, or isopropyl, the reaction does not give 10, but the
corresponding alkylidine 2H-imidazole compound instead (Scheme 2 far right). The derivative
with R2 = tert-butyl derivative is well behaved due to the lack of α-protons, giving exclusively the
desired imidazo-phenanthrene and is therefore the best choice in terms of S1/T1 energies as well
as stability and solubility of the compound. In contrast to R1 = phenyl, when R2 is phenyl steric
interactions force the aromatic ring out of conjugation, so the orbital and excited state energies are
largely unaffected, e.g. R1 = H, R2 = Ph give S1/T1 energies of 3.30/2.89 eV (based on DFT
calculations).

Tier 2 Selection
The Second Tier selection explored nitrogen substitution in the H-rings of 10.
Incorporating 0, 1 or 2 nitrogens into the H-rings of the parent phenanthro[4,5-f]imidazole gives a
total of 37 unique aza-substituted compounds (see the Electronic Supporting Information, ESI).

Figure 3-4. General scheme of H2P synthesis

48

For the Tier 2 screening, we chose to carryout DFT calculations on the 37 aza-substitutions of 10
with R1 = tert-butyl and R2 = phenyl using the B3LYP hybrid functional and LACV3P** basis set.
In Tier 2 we chose a markedly larger basis set than was used in Tier 1, so direct comparisons of
theoretically predicted properties to experimentally determined values will be meaningful. In these
calculations, we predicted a number of parameters, including singlet and triplet energies, as well
as molecular orbital energies and surfaces for each of the 37 molecules. The number of nitrogens
and the site of nitrogen substitution in the H-rings markedly affects the electronic properties of the
H2P molecules. For a complete listing of the parameters predicted in the Tier 2 calculations see
the ESI. Figure 3.3 plots two of the calculated Tier 2 properties: LUMO energy vs. triplet energy.
As expected, the LUMO is stabilized with nitrogen substitution in the H-ring, shifting from -1.0
eV for 10a to a range of -1.2 to -1.4 for a single nitrogen to -1.4 to -1.9 for two nitrogens, with the
site of nitrogen substitution markedly affecting the degree of stabilization of the LUMO. The
greatest stabilization is seen for substitution in the 3- and 8-postions and the least for substitution
at the 2- and 9-positions. The larger range of LUMO energies seen for the molecules with two
nitrogens is due to an additive effect when both nitrogens are in a single H-ring. The ortho, meta,
and para derivatives together give LUMO energies within a range of roughly 0.2 eV and an
average of -1.7 eV, while the materials with a single nitrogen in each H-ring give the same range
of LUMO energies, but with an average LUMO energy of -1.55 eV. The site of nitrogen
substitution also effects the triplet energy, but not in the same manner as it does the LUMO
49

energies. Substitution of a single nitrogen into
the two H-rings, two nitrogens meta in a single
H-ring or a single nitrogen in both H-rings gives
a minimal change in the triplet energy of the
molecule, relative to the unsubstituted
compound 10a. In contrast, substitution of two
nitrogens into a single H-ring in either an ortho
or para disposition lowers the triplet energy
substantially below 2.8 eV in nearly every case
(the exception has N in the 2,3-positions). The
reason for the marked red shift in the ortho- and
para-substituted derivatives is a filled-filled
interaction of the two nitrogen lone pairs in
these compounds, leading to a marked
destabilization of the out of phase combination
of the lone pair orbitals and a resultant narrowing of the HOMO-LUMO gap. This interaction of
filled non-bonding orbitals is not seen for the meta-substituted derivative or those with a single
nitrogen in each H-ring.
The Tier 2 screen also included estimation of the electron and hole reorganization energies
for each of the H2P molecules. Reorganization energies are useful parameters to evaluate the
kinetics of intermolecular hole and electron hopping and assess charge carrier conduction. A lower
reorganization energy reduces the barrier to carrier hopping between molecules in the thin film
and efficient carrier conduction is more favorable for a given material. The compounds in the Tier

Figure 3-5. LUMO vs. triplet energy for second tier
iteration of aza substitution in phenanthrene section the
H rings of the parent phenantho[4,5 f]imidazole.
Compounds 10a 10d¬ are illustrated by colored circles.
The identities of the other compounds in the scr

2.0 2.2 2.4 2.6 2.8 3.0
-2.0
-1.8
-1.6
-1.4
-1.2
-1.0
-0.8
10c
10b
10d

parent 10a
one N
Two N:
One N in each ring
ortho
meta
para
LUMO (eV)
Triplet Energy (eV)
10a
50

2 screen ranged from low to moderate reorganization energies for holes (0.24-0.30 eV) and
electrons (0.18-0.40 eV), suggestive of efficient carrier transport in the H2P family (see the ESI
for a full listing).
We chose to focus on three imidazo[4,5-f]phenanthroline derivatives 10b, 10c and 10d for
characterization and study in blue OLEDs due to their high triplet energies and favorable reduction
potentials. Additionally, these three materials have symmetric nitrogen substitution patterns in the
H-rings, so only a single regioisomer will be formed for each derivative by our synthetic approach.
Figure 3.4 shows the DFT optimized ground state geometries along with the HOMO and
LUMO orbitals for 10a-10d. The orbital density distribution of the HOMOs and LUMOs for
10b-10d are very similar. The LUMOs are localized on the phenanthroline while the HOMOs
involve both the phenanthroline and imidazole fragments. The addition of tert-butyl and phenyl
groups have a negligible effect on the composition of the HOMO or LUMO of these compounds.
For comparison, the HOMO and LUMO are shown for 10a as well. The HOMO matches that seen
for 10b-10d, but the LUMO of 10a is localized on the phenyl group, rather than the phenanthroline,
due to the relative difficulty of reducing phenanthrene compared to phenanthroline. The LUMO+1
orbital of 10a is a good match to the LUMO orbital of 10b-10d. Nitrogen substitution in the
H-rings stabilizes the phenanthrene system, such that the phenanthroline orbitals fall below the
phenyl based p-orbitals for 10b-10d.
Synthesis and Physical Characterization of H2P Host Materials
Compounds 10b-10d were synthesized in 60-80 % yields as illustrated in Scheme 2, where
the phenanthrene-dione precursor is replaced with the appropriate phenanthroline-dione. The H2P
compounds exhibit intense π-π* absorptions between 250 and 290 nm. A weak tail from 320 to
51

400 nm was observed for 10b and 10d, whereas a more intense and distinct absorption is seen for
10c in the same spectral region. The absorption features at longer wavelengths suggest a degree
of charge transfer character, which is strongest for 10c. Solutions of 10b, 10c, and 10d have
emission maxima at 392, 402, and 396 nm, respectively, at room temperature. Singlet excited state
energies corresponding to the energy of the short wavelength edge of the emission band (intensity
= 0.1×lmax) for the three compounds are 3.47, 3.20 and 3.42 eV, respectively. The triplet energies

Figure 3-6. Highest occupied molecular orbital and lowest unoccupied molecular orbital diagrams calculated at the
B3LYP/LACV3P** level of theory. The permanent dipole moment for each molecule is illustrated in the images at the top.

52

(ET) for 10b, 10c, and 10d at 2.91, 2.76,
and 2.83 eV, respectively, were
determined from the high energy edge of
the phosphorescence spectra measured in
2-MeTHF at 77 K. Triplet energies
obtained from neat solids are lower than
those in solution. Triplets in the solid state
were found to be ET = 2.65 eV for 10b,
2.47 eV for 10c and 2.62 eV for 10d
(Figure 3.6). The triplet energy
depression in the solid state, of
approximately 250 meV, for all three
materials could not have been predicted
from the modeling in Tiers 1 and 2,
prompting us to continue with a third Tier
(vide infra).
The photoluminescent quantum
yields (FPL) of 10% FIrpic-doped films are
0.61, 0.40 and 0.20 for 10d, 10b, and 10c,
respectively. The photoluminescence
efficiency of a FIrpic doped film of
10b-10d will limit the internal quantum
efficiency for OLEDs made with each host

Figure 3-7. Absorption (black line), fluorescence (blue line), and
cryogenic (77 K) spectra (red line) of (a) 10b, (b) 10d, and (c) 10c.
In each subfigure, the bottom axis refers to wavelength and the
top axis denotes energy in eV. (d) Solid state phosphorescence

250 300 350 400 450 500 550 600 650
0.0
0.4
0.8
(a)
Wavelength (nm)
(b)
Absorbance (a.u.)
10d
Absorption
(c)
0.0
0.4
0.8
Emission (RT)
Emission (77K)
Emission Intensity (a.u.)
0.0
0.4
0.8
10c
Absorption
Absorbance (a.u.)
0.0
0.4
0.8
Emission (RT)
Emission (77K)
Emission Intensity (a.u.)
0.0
0.4
0.8

10b
Absorption
Absorbance (a.u.)
0.0
0.4
0.8
Emission (RT)
Emission (77K)
Emission Intensity (a.u.)

5 4.5 4 3.5 3 2.5 2
Energy in eV
400 450 500 550 600 650 700
0.0
0.2
0.4
0.6
0.8
1.0
2.47 eV
2.65 eV

Intensity (a. u.)
Wavelength (nm)
10b
10c
10d
2.62 eV
3 2.8 2.6 2.4 2.2 2 1.8
Energy (eV)
53

material. The low FPL for doped films of 10c is due to the low triplet energy of that host, such that
significant quenching of FIrpic emission is expected. However, 10b with the highest triplet energy
did not provide the highest FPL, likely due to a lower average triplet energy in the solid state than
that of 10d, leading to partial quenching of FIrpic emission in 10b.
Cyclic voltamograms of 10a-10d showed quasi-reversible reduction waves, however, only
10d gives a quasi-reversible oxidation, whereas oxidation of 10b and 10c are irreversible (see
Table 1 for potentials). DFT calculations suggest that reduction occurs predominantly on the
phenanthroline fragment and oxidation involves the imidazole ring (Figure 3.5). The
electrochemical potentials are consistent with these predictions. The imidazole groups are not
significantly affected by the nitrogen position in the H-rings, leading to a closely spaced set of
oxidation potentials for 10b-10d (1.11 to 1.17 V). In contrast, the phenanthroline fragment is
significantly affected by the nitrogen position, leading a more disparate set of reduction potentials
(-2.25 V to -2.54 V). Placement of nitrogens in the 3-, 8- positions of the phenanthroline gives
10c the lowest reduction potential. HOMO and LUMO levels in Table 1 were calculated from
redox potentials using previously published correlations.

Table 3-1. Summary of calculated and experimental physical properties for 10b, 10c and 10c.

E
ox
/E
red

HOMO (eV)

LUMO (eV)
(+)
e

(eV)
(-)
f
(eV)

ET (eV)
 PL
g

(V/V)
a
Calc. Exp.
d
Calc. Exp.
d
Calc. Solution Solid

(sol-sol)
10a 0.89
b
/-2.97
b
-5.45 -5.85 -1.54 -1.27 0.24 0.21 2.90 2.86 2.70 0.16 --
10b 1.15
c
/-2.54
b
-5.63 -6.21 -1.50 -1.76 0.27 0.40 2.91 2.92 2.65 0.27 0.40
10c 1.11
c
/-2.25
b
-5.54 -6.15 -2.10 -2.10 0.25 0.30 2.84 2.71 2.47 0.24 0.20
10d 1.17
b
/-2.51
b
-5.58 -6.24 -1.83 -1.79 0.28 0.22 2.90 2.85 2.62 0.23 0.61
a
Oxidation and reduction potentials vs. ferrocene/ferrocenium redox couple.
b
Quasi-reversible.
c
irreversible.
d
HOMO and LUMO
energy levels were calculated from the redox potential with published correlation.
51, 52

e
Calculated hole reorganization energy.
f

Calculated electron reorganization energy.
g
Photoluminescent quantum yield of vacuum deposited films containing 10% FIrpic-doped
H2P.

54

The HOMO and LUMO values for 10a-10d were computed using the adiabatic scheme
(implemented in the Materials Science Suite) at the B3LYP/LACV3P** level where single-point
energies were computed for the cation, anion and neutral species using the finite-element Poisson-
Boltzmann solver (PBF) continuum solvent model with DMF as the solvent on the corresponding
gas-phase optimized structures. The HOMO/LUMO energies calculated using this procedure are
in good agreement with experimental values as seen in Table 1.
Tier 3 selection
The gas phase calculations in Tiers 1 and 2 agree well with the properties of the
three host compounds in fluid solution. However, these single-molecule calculations cannot
predict the solid-state properties of the bulk materials. The decrease in triplet energy from solution
to neat solid (vide supra) prompted us to include a third Tier of modeling to address solid–state
properties. Because modeling excited state energies in amorphous solids is challenging, and
because triplet energy depression were similar for the three materials, attention was turned to
charge transport, the critical role of a host material. In order to screen for the performance of
10b-10d as host materials, we carried out theoretical studies of the three materials as amorphous
solids using MD and electron coupling DFT calculations. The search criterion for this Tier 3
screen is a maximization of charge mobility. MD simulations were performed on a cell of 125
molecules of each host material with periodic boundary conditions in order to model solid state
morphologies. The final density for each simulation was between 1.11-1.14 g/cm
3
. As a first
approximation of charge mobility, the number of π-π interactions for 10b, 10c and 10d were
counted, with more interactions presumably being favorable for charge transport. Here, the -
interactions are classified into two types: 1) face-face interactions where two aromatic rings are
within a distance of 4.4 Å and a maximum angle of 30

with respect to each other, and 2) edge-
55

face interactions where two aromatic rings
are within a distance of 5.5 Å with a
minimum subtended angle of 60

between
them. The number of - contacts includes
the total number of face-face and edge-face
interactions, giving a total of 324, 409, and
364 for 10b, 10c, and 10d, respectively.
Compound 10c has substantially more face-
to-face -contacts than either 10b or 10d,
while 10d has more face-to-edge -contacts than the other two compounds. This preliminary
analysis suggests a trend in charge transport as: 10c > 10d > 10b. A similar trend is seen for the
center of mass (COM) radial distribution functions (RDF), as shown in Figure 3.7. The COM
RDFs reflect differences in solid state center-to-center distances, which could have an effect on
charge transport in the host material. The 10b RDF clearly shows differences in distances in the
range of 4-6 Ǻ. There is approximately a 35% and 50% difference in height between the
nearest-neighbor peak maximum of the 10b RDF and the nearest-neighbor peak maximum of 10d
and 10c RDFs, respectively. This indicates a lower proportion of short center-to-center distances
in bulk 10b with respect to the other H2Ps.
A plausible explanation for the dissimilar morphologies of 10b from 10c and 10d involves
the dipole moments of each molecule. In all three cases the permanent dipole moment for each
molecule lies in the H2P plane, however, in 10c and 10d the dipole extends from the imidazole
ring into the phenanthroline, while in 10b the dipole is largely within the imidazole ring
(figure 3.5). In the amorphous solid, the molecules tend to form dimer pairs with adjacent dipoles
3 4 5 6 7 8
0.0
0.5
1.0
1.5

g(r)
r (Å)
10b
10c
10d

Figure 3-8. Center of mass radial distributions [g(r)] from MD
simulations of three host materials, obtained by averaging
over 30ns.

56

in a roughly antiparallel orientation. For 10c and 10d, closely spaced dimers can be formed with
antiparallel dipoles, however, for 10b overlapping the dipoles of adjacent molecules are sterically
hindered by the tert-butyl and phenyl groups on the imidazoles.
Complimentary to structural analysis of MD simulations, a quantum mechanical analysis
was performed. Dimer frontier orbital splitting electron coupling calculations (see methods

Figure 3-9. Histogram plots showing the distribution of hole and electron hopping rates extracted from the frontier dimer
orbital splitting coupling calculations of both the exhaustive dimer set and the smaller 20 dimer subset for the three host
materials
(top: 10b, middle: 10c, bottom: 10d)
0
5
10
15

20 Dimers
Exhaustive
0
200
400
600
800
0 5 10 15 20 25
0
5
10

Hole Hopping Rate (10
13
/s)
0
200
400
0
5
10
15

20 Dimers
Exhaustive
0
100
200
300
400
0
5
10

Count
0
200
400
600

0
5
10

Count
0
200
400
600
0 5 10 15 20 25
0
5
10

Electron Hopping Rates (10
13
/s)
0
100
200
300
10b
10d
10c
10b
10d
10c
57

section) were performed on the MD equilibrated structure for all dimer pairs within a contact
distance of 4Å from each other amounting to a total of 838, 857, and 880 dimer pairs for 10b, 10c,
and 10d respectively). These calculations were used to estimate the charge carrier hopping rates
according to equation (3) to assess variations in charge transport between neat host materials.
Figure 3.8 shows histograms of the calculated hole and electron hopping rates. Notably, the
distribution of electron hopping rates for 10b is significantly narrower and slower than those of
10c and 10d. There were no significant differences in hole hopping rates between the three host
materials. While feasible for this study, transport calculations carried out for a large number of
dimer pairs of molecules is too time consuming and impractical to be repeated for a large number
of different materials. In order to develop a more rapid, high-throughput Tier 3 materials screen,
we sought to test the validity of a smaller scale estimate of the hopping rates for each of the
materials using a truncated random subset of dimer pairs. Calculations were repeated for twenty
random dimers from the exhaustive set of dimer pairs with a spacing of 4 Ǻ or less. The
distribution of hopping rates for the smaller sets mirrors what we observed for the exhaustive
calculations (Figure 3.8), suggesting that a less rigorous calculation could be used in the future to
compare the range of hoping rates expected for different materials.
The two coupling methods predict similar hole mobilities for the three host materials, but the

Table 3-2. Number of - contacts and calculated mobilities for 10b, 10c and 10c.

-contacts
Isotropic mobility (x10
-4
m/Vs)
Dimer-splitting CDFT
Total
-ff
(face-
face)
-ef
(edge-
face)
-ff/-ef

µ h µ e µ h µ e

10b 324 77 247 0.31 4.45 0.67 20.3 23.4
10c 409 142 267 0.53 6.47 3.46 27.5 53.8
10d 364 72 292 0.25 6.16 5.34 14.9 41.7

58

electron mobility calculated for 10b is
significantly lower than that of either 10c or
10d. The low electron mobility for 10b is
consistent with its greater center-of-mass
intermolecular spacing leading to a lower
average hoping rate for the 500 dimer pairs
examined. Tier 3, performed at various
levels of theory, predicts that 10b may
display poor electron mobility in the solid
state and therefore fails to meet the selection
criterion. After 3 Tiers of screening on a
555 membered structure space, 2 materials
were selected: 10c and 10d. For the benefit
of validating the above theoretical methods,
10b was also investigated experimentally as
an example of a poorly performing material.

OLED Fabrication and Testing
We employed a relatively simple
configuration for OLED devices with H2P
host materials: ITO/NPD (20 nm)/mCP (5
nm)/H2P: FIrpic (10%, 30 nm)/BCP (50
-7
-6
-5
-4
-3
-2
-7
-6
-5
-4
-3
-2
10d 10c
Energy (eV)
ITO
NPD
10b
mCP
FIrpic
BCP
LiF:Al
(a)
0 2 4 6 8 10 12
0.01
0.1
1
10
100
1000
10b
10c
10d

Brightness (cd/m
2
)
Voltage (V)
(b)
0
50
100
150
200
250
300
Current Density (mA/cm
2
)
0.1 1 10 100
0.1
1
10
10b
10c
10d

External Quantum Efficiency (%)
Current Density (mA/cm
2
)
(c)

Figure 3-10. OLED characterization: (a) Energy level diagram of
the OLEDs (blue line is FIrpic), (b) J V L characteristics, (c) plot
of EQE versus current density.

59

nm)/LiF (1 nm)/Al (NPD = N ,N ′-Di(1-naphthyl)- N ,N ′-diphenyl-(1,1′-biphenyl)-4,4′-diamine,
mCP = 1,3-di(9H-carbazol-9-yl)benzene , BCP = 2,9-Dimethyl-4,7-diphenyl-1,10-
phenanthroline). NPD/mCP functioned as the hole-transporting layers, and BCP was introduced
as both hole-blocking and electron-transporting layer (energy level diagram is given in Figure
3.9a). Figure 3.9b shows the current density-voltage-brightness (J-V-L) characteristics of these
devices. The highest current is observed for the devices based on 10c, with comparable light
output at a given voltage observed for 10c and 10d, leading to higher external quantum efficiency
(EQE) for 10d than 10c. The maximum EQE values are 0.7 %, 5.0 % and 9.0 %, for devices with
10b, 10c and 10d, respectively.
In order to understand the source of the differences in device efficiencies, we roughly
deconvoluted the EQE into four limiting factors as shown in Equation 7.
(7)
The photoluminescent quantum yield (ΦPL) was measured for phosphor doped host films
and is assumed to be unchanged in the device. The usable exciton fraction (χ) takes into account
the statistical branching ratio of electro-generated singlets to triplets with respect to the emissive
species, which is unity for a phosphorescent
dopant such as Firpic. The outcoupling factor
(ηe) accounts for losses due to wave-guiding
and plasmon absorption, and is usually 0.2-0.3.
Lastly, the charge recombination factor (ηr) is
the ratio of excitons formed to injected charge
carrier pairs, sometimes referred to as charge
F
EL
=F
PL
ch
r
h
e

Table 3-3. Efficiency parameters of the OLEDs for 10b, 10c
and 10d. (η r was computed assuming x =1 and n e = 0.2)
Host
EQE
(%)
 PL
Charge
recombination
factor ( r)
10b 0.7 0.4 0.09
10c 4 0.2 1
10d 9 0.6 0.75

60

balance. Charge recombination factors that are less than unity are due to differential charge
injection or transport between electrons and holes.
Table 3 lists the parameters from Equation 7 for each host. The EQE and ΦPL were
measured from devices and thin films, respectively. The χ and η e are assumed to be 1 and 0.2,
respectively. The ηr is calculated from the other four parameters.
The highest efficiency is seen for the device with a 10d based emissive layer. The device
performance with 10c is noteworthy since the device exhibited an EQE at its theoretical limit.
Considering a 20 % FPL and out-coupling of 0.2, the maximum achievable efficiency for the
OLEDs is no greater than 0.04, assuming that the dopant is isotropically dispersed in the 10c host.
This suggests near unit efficiency for carrier recombination in these devices. The 10b-based
device exhibited the lowest device efficiency among the three hosts despite a 40% QY in a 10%
FIrpic-doped film, due to a very low charge recombination factor. The trend in J-V characteristics
shows current densities of 10c > 10d > 10b at a given voltage, agreeing well with the Tier 3
screening. Furthermore, OLEDs with 10b have a markedly higher turn on voltage than the other
two materials, suggesting that charge carrier mobilities within the devices varies with the trend
10c > 10d > 10b. The turn-on voltage (voltage at a brightness of 0.1 cd/m
2
) for devices utilizing
10b is ca. 2 V higher than those for devices with 10c and 10d.
61

By inspection of the energy level diagram (Figure 3.9a) and the nature of charge transport
in other blue OLEDs, the dopant (FIrpic) mediates hole transport, while the host is expected to
mediate electron transport. To test the predictions of the MD study with regards to electron
transport characteristics of the three hosts, we fabricated electron-only devices with a structure of
ITO/BCP (10 nm)/H2P (40 nm)/LiF:Al. In this device architecture, holes are blocked so the current
passing through the organic layers is purely an electron current. Figure 3.10 shows the current
versus voltage characteristics of the
electron-only devices for the three
materials. As anticipated, the trend in
the J-V characteristics shows current
densities of 10c > 10d > 10b at a given
voltage, and the device with 10b gave a
current that is roughly 3 orders of
magnitude lower than those of devices
with either 10c or 10d.
Using computational methods to
accelerate the materials discovery process, we identified the best candidates from a large library
of different host structures before preparing them. Synthesizing and physically testing only those
candidate materials with a high likelihood of success minimizes the time taken to discover useful
materials. The screening of the materials was based on DFT and TDDFT calculations for the
candidate molecules in the gas phase and was used to identify the materials that gave the most
promising triplet and LUMO energies for blue PHOLEDs. The criteria used to choose the host
material also depends on the other materials used to fabricate the blue OLEDs. Here we chose to

Figure 3-11. Current-voltage plots for electron only devices, i.e.
ITO/BCP (10 nm)/H2P (40 nm)/LiF:Al.

0 2 4 6 8 10 12
1E-8
1E-7
1E-6
1E-5
1E-4
1E-3
0.01
10b
10c
10d
Current (A)
Voltage (V)
62

search for host materials that are optimally designed for a FIrpic dopant, an NPD/mCP hole
transporting stack and a BCP electron transporting layer. If it were desirable to change the other
materials in the PHOLED, i.e. the emissive dopant or transport materials, the screening process
could be reassessed to find the best host material for the new set of PHOLED materials. It is
important to stress that the DFT/TDDFT calculations would not need to be run again, as the new
set of PHOLED materials would simply change the parameters used to screen the database of
calculated molecular properties. One could take this process further and consider the materials in
this library for another application all together. If one were interested in finding materials for
photovoltaic applications for example, a different set of search criteria could be used to screen the
same database of calculated properties to find the best candidate materials for that application. In
the context of the Tier 1 and Tier 2 screening approach reported here, the Tier 2 screen might need
to be repeated with a different candidate or candidates advanced from Tier 1.
The screening methods based on gas phase computational modeling allowed us to identify
the materials with suitable triplet and LUMO energies to serve as host materials for blue
PHOLEDs. The calculations gave excellent agreement with the experimental parameters
measured in fluid solution. Note that Tiers 1 and 2 were not sufficient to predict device
performance. Extending our modeling studies of the three synthesized materials into the solid
state in Tier 3, using MD simulations and modeling of the resultant equilibrated amorphous solids,
predicted the differences in the carrier transport we observed in the materials. While the two host
materials identified after Tier 3 screening (10c and 10d) appeared to be well suited as PHOLED
host materials, the gas phase modeling studies did not include condensed phase polarization or
packing effects. We observed a marked red shift in the triplet energy between the gas/solution
phase triplet energies to those in the solid state, to the extent that emission from Firpic was either
63

partially (10d) or mostly (10c) quenched. We are currently exploring further multi-scale modeling
methods to develop an expanded Tier 3 screen that will predict triplet energies of organic materials
in the solid state. With such a multi scale “solid-state” method in hand we can use a computational
approach to accelerate materials discovery of compounds with triplet energies high enough to act
as host materials in deep blue PHOLEDs.
The total chemical space of materials considered in our Tier 1/Tier 2 screening process was
555 structures, however, it could be expanded to much larger chemical space. Recent reports from
other groups have shown a similar computational screening approach used with libraries that are
several orders of magnitude larger than the one considered here. In a study from the Aspuru-Guzik
group, the researchers identified 20 candidate TADF emitters for PHOLEDs from a 1.6 million
member library, prepared them and demonstrated external efficiencies as high as 20% for one of
them, comparable to some of the best Ir-phosphor based PHOLEDs. It is important to stress that
this study involved dopants, which will be present in the OLED at low concentration and thus have
no interactions between dopants. This is an important distinction from the thin film host materials
considered here, where intermolecular interactions mediate energy and electron transport. Gas
phase modeling studies are sufficient to predict PHOLED performance for an emissive dopant,
such as a TADF material, but condensed phase studies are needed for materials that will transport
charge or excitons in the device. In the present study, we used a multi-scale approach involving
MD in conjunction with DFT calculations to model the solid state properties of the three materials
that we prepared and studied, however, this would not be practical for libraries with thousands of
members or more. We are currently working to develop methods to streamline the condensed
phase modeling studies, with the goal to rapidly identify new host and transport materials for
optoelectronic applications.
64

CHAPTER 4 - BEYOND PHENYLPYRIDINE: NOVEL PLATINUM PHOSPHORS
The previous three chapters lay out several strategies for blue shifting emission from
phenylpyridine based metal complexes. Decoration with electron accepting or donating
substituents has the benefit of being synthetically accessible, and fluorination in particular has
been successful, but suffers from the chemical instability of those substituents. Azasubstitution is
a more robust strategy for electronic modulation without the introduction of unstable substituents.
Chapter 2 demonstrates an application of this strategy to the blue shifting of a phenylpyridine like
ligand system. Chapter 3 demonstrates an azasubstitution based computational approach to finding
stable host materials. This chapter explores possible future directions beyond phenylpyridines
using other strategies for blue shifting emission. This includes he effect of multiple metal centers,
ring expansion of the metallocycle to a 6 membered ring system, and the design, synthesis, and
characterization of a cycloplatinated aryl phosphine imide.

1. Multiply platinated phosphor
It is desirable to further blue shift the Bzp ligand system described in chapter 2. Following the
color tuning logic from chapter 1, either the HOMO should be stabilized or the LUMO should be
destabilized. Stabilization of the HOMO by isoelectronic transmutation of CH to more
electronegative N proved to be an effective strategy in the design of Bzp. In that case, the phenyl
group of phenylpyridine was transmuted to pyridine. The logical continuation of this approach is
to further transmute pyridine to pyrimidine. The regiochemistry of this transmutation is such that
the two pyrimidine nitrogen atoms are ortho- and para- to the site of metalation. This aza-Bzp
ligand system would be expected to reliably blue shift phosphorescence from Pt or Ir metal
complexes with respect to Bzp. A proposed synthetic route analogous to Bzp requires
65

benzo[h]quinoxaline as a starting material. Unfortunately, this compound is not readily available,
adding complexity to the overall synthesis.
Electron donating or withdrawing substituents can destabilize or stabilize frontier orbitals,
respectively. Those substituent effects are quantified by their Hammett parameters, with respect
to the spacial wavefunction of the frontier orbitals. For instance, a methoxy group para to the locus
of HOMO density is destabilizing to the HOMO due to its negative para Hammett parameter.
Conversely, a methoxy group meta- to HOMO density is stabilizing, due to its positive meta-
Hammett parameter. Thus, a particular substituent may act as both electron donating and
withdrawing depending on its regiochemistry. The Hammett analysis is limited to main group
organic substituents. This section aims to propose a Hammett-like effect for transition metal
substituents, and differentiate between coordination modes as fundamentally different electronic
perturbations.
A transition metal
based strategy for blue shifting
the BzpPt(acac) framework is
shown in figure 4.1. Two
additional transition metal
centers are introduced. One is
coordinated by the nitrogen
para- to the cycloplatinated
carbon, which carries a large HOMO density. Sigma donation of the L-type (two electron)
pyridine ligand to the metal results in inductive withdrawing of electron density from that ring.
One of the limiting resonance structure representations of the metal nitrogen dative bond places a

Figure 4-1. Metal donating/withdrawing strategy for blue shifting
phosphorescence
66

partial positive formal charge on nitrogen, thereby stabilizing the HOMO. The second additional
metal center is bound to the carbon para to the platinum coordinated nitrogen. In this case, the
carbon acts as an X-type (one electron) ligand to the metal, with strong pi- back donation. One of
the limiting resonance structure representations of the metal carbon sigma bond places a partial
negative formal charge on carbon, thereby destabilizing the LUMO. The salient destiction
between the electron donating metal and the electron accepting metal is not the identity of the
metal but the coordination mode. The metal that is L-type is electron withdrawing and the metal
that is X-type is electron donating. In fact, both metal atoms may be of the same element and serve
opposite electronic purposes depending on their coordination mode.
A natural example of such a compound is that in which the two perturbing metal atoms are
platinum. In order to allow for a stable square planar geometry of the X-type carbon bound
platinum atom, a third nitrogen was introduced to coordinate it as shown in figure 4.1. Since this
ligand is a triazatriphenylene, it is henceforth referred to as tripyridylene, or Trippy. Furthermore,
the L-type nitrogen bound platinum is also bound to the adjacent carbon to form a five member
cyclometallated ring. Beta-diketonate ligands fill out the coordination sphere of these metal atoms.
It is clear by inspection that the resulting trinuclear complex is three-fold symmetric in the point
group C3h, and the metal atoms are chemically identical.
The three fold symmetric ligand, Trippy, does not include a blocking group like Bzp and
is therefore significantly less chemically complex. This symmetry prompted a seemingly natural
cyclotrimerization based retrosynthetic approach. It is well known that triphenylene can be
synthesized in high yields from the fluoride promoted generation of transient benzyne from
orthotrimethylsilyltriflylbenzene in the presence of palladium (0) as cyclotrimerizing template.
Heterocyclic cyclotrimerizations were recently reported by Garg and co-workers, though those did
67

not contain pyridinic nitrogens. Pyridine is a known reactive intermediate, generated from
2-triflyl-3-trimethylsilylpyridine, and has been trapped as its 4+2 cycloadduct with furan. A range
of conditions were screened for the cyclotrimerization of 2-triflyl-3-trimethylsilylpyridine.
Unfortunately, no detection of trimerization products were observed.
A triple Skraup synthesis from triaminobenzene was then attempted. Triaminobenzene is
a hydroscopic air sensitive colorless compound which darkens to deep purple upon exposure to
oxygen. As such, it is not commercially available and difficult to handle. It can be reliably
produced from the reduction of 3,5-dinitroaniline by hydrogen gas in tetrahydrofuran in the
presence of carbon supported palladium metal, since it can be collected by filtration of the
heterogeneous catalyst and evaporation of the solvent and subsequently used without further
manipulation or purification. Treatment of crude, nascently generated 1,3,5-triaminobenzene with
excess glycerol and concentrated sulfuric acid in the presence of sodium 3-nitrobenzenesulfonate
at 150C resulted in a myriad of products, including a 1.4 % yield of the desired Trippy.
A synthesis of Trippy analogous to Bzp was then undertaken, proceeding through an
orthoquinone. Commercially available 1,7-phenanthroline was oxidized to
1,7-phenanthroline-5,6-dione by nitric acid in the presence of potassium bromide at 150C. The
resulting diketone was condensed with formamidhydrazone in methanol at room temperature to
yield aan isomeric mixture of triazines, similarly to the synthesis of Bzp. Regrettably, the
regiomeric ratio of triazines was 7:3 in favor of the undesired regioisomer. The synthesis was
completed by refluxing in o-dichlorobenzene in the presence of excess norbornadiene and
purification by liquid chromatography to yield Trippy in 10 % overall yield.
With Trippy in hand, platination conditions were carried out similarly to those used
to generate BzpPt(dpm), with the modification of an excess amount of K2PtCl4 being employed to
68

promote polymetallation of the ligand. Unfortunately, only the mono-platinated species
[Pt]1Trippy was observed, presumably due to the insolubility of the monoplatinated Trippy
chloride intermediate. Gratifyingly, conditions developed by the Wang group using an
organoplatinum reagent were successful in generating the triply platinated [Pt]3Trippy.
Figure 4.2 shows
the emission spectrum of at 77 K
in frozen methylcyclohexane and
at room temperature dispersed in
polymethylmethacrylate
(1 wt%). The compound is not
measurably emissive in fluid
solution, and weakly emissive in
PMMA (PLQY < 1 %,
t = 14 ns). The low temperature
spectrum shows clear features,
though they are notably unusual in progressive intensities. The high energy emission feature,
tentatively assigned to the zeroth vibrational emission band, is at 450 nm. This is blue shifted from
BzpPt(dpm) as expected. The excited state decay time at 77 K is 8 us, which is indicative of a
phosphorescent transition with strong spin orbit coupling.

425 450 475 500 525 550 575 600 625 650
0.0
0.2
0.4
0.6
0.8
1.0

Normalized Intensity
Wavelength (nm)
r.t. PMMA
77K MeCH

Figure 4-2. Emission spectra of [Pt]3Trippy in PMMA (red) and frozen
methylcyclohexane (black)

69

2. Metallocycle expansion
Another strategy for blue shifting
emission from ligands beyond
phenyl pyridine is to expand the
ring size of the metallocycle from
five membered to six membered.
This effectively breaks
conjugation between phenyl and
pyridine. This strategy has been
successfully employed in
tetradentate hemimacrocyclic
platinum complexes, where the
triplet spin density is localized on a pyridyl carbazole metallocycle. Unlike the highly luminescent
tetradentate platinum complexes, iridium complexes with bidentate pyridyl carbazole are weakly
luminescent at room temperature, with fast non-radiative rates. Figure 4.3 shows the emission
spectra of Fac Czpy3Ir in at 77 K in frozen MeTHF glass and at room temperature in a
polymethylmethacrylate matrix. In room temperature fluid solution, the compound is not
emissive, and in the polymer matrix it has a quantum yield of 4 %. The quenching mechanism is
hypothesized to be similar to that of tris-phenylpyrazolyo iridium (III), which involves Ir-N bond
rupture in the excited state, leading to a non-emissive metal centered ligand field state.
In this study, a conformationally restricted derivative of pyridylcarbazole was synthesized
and examined for its photophysical properties. The design strategy is shown in figure 4.4. The
bay position of pyridiylcarbazole opposite the metallocycle is “tied back” with a methylene bridge.
400 425 450 475 500 525 550 575 600 625 650
0.0
0.2
0.4
0.6
0.8
1.0

normalized intensity (arb. units)
wavelength (nm)
excit. (77 K)
77 K (MeTHF)
300 K (PMMA)

Figure 4-3. Excitation and emission spectra of Fac Czpy3Ir

70

The methylene bridge may be decorated with two R
group substituents, or be part of a ring system as a spiro
center. Retrosynthetically, bond formation between the
methylene and the carbazole moiety is most sensibly
formed by the electrophyllic substitution of a methylene
carbocation with a carbazole CH fragment. As such, the
methylene ought to be highly substituted with alkyl or aryl groups to stabilize the carbocation.
Therefore dimethyl, diphenyl, and spirofluorenyl methylene groups were selected as synthetic
targets. The dimethyl target was approached first, due to its simplicity and the use of readily
available acetone as starting material. Figure 4.5 shows the synthetic route to Me2Snap and its
unsuccessful iridium cyclometalation conditions. The isopropylene group was installed in the
3- position of 2-bromopyridine by deprotonation with strong base (LDA) and quench with acetone.
While the most acidic protons of pyridine are in the 2- and 4- positions, the bromide promotes
ortho-deprotonation in the 3- position by stabilizing the lithiated species. It is important to note
that the resulting 2-bromo-3-lithiopyridine is unstable via two decomposition pathways. First,
ortho-lithiobromoaromatics may decompose to arynes with loss of lithium bromide. To prevent
this side reaction, the reaction was kept at -78C during the course of the reaction. Second, if any
2-bromopyridine is present with the
2-bromo-3-lithiopyridine, a bimolecular
“halogen dance” reaction may occur in
which a lithium-halogen exchange
cascade results in rearrangement to
3-bromo-2-lithiopyridine. In order to

Figure 4-5. Syntheisis of Snap3Ir

Figure 4-4. Rotationally restricted design
strategy

71

prevent this side reaction, the starting material, 2-bromopyridine, was added dropwise to a solution
of LDA so that the base was in excess and 2-bromopyridine did not exist in significant
concentration in the presence of the nascent 2-bromo-3-lithiopyridine. After quenching the
reaction with anhydrous acetone, followed by water, the corresponding tertiary alcohol was
obtained as a yellow solid by chromatography. A copper catalyzed Ullmann arylation with
carbazole efficiently yielded the 2-carbazolylpyridine tertiary alcohol. The tertiary alcohol was
then promoted to ring closure through dehydration by heating in polyphosphoric acid, presumably
through either a dimethylaryl carbocation or a tertiary alkylphosphate intermediate. Me2Snap was
recovered in high yield by chromatography.
Treatment of Me2Snap by the Nonoyama conditions with IrCl3 in ethoxyethanol/water
resulted in no reaction, with nearly complete recovery of the starting ligand. The steric demand of
the ligand likely precludes the formation of cyclometallation intermediates. Therefore, a platinum
425 450 475 500 525 550 575 600
0.0
0.5
1.0
425 450 475 500 525 550 575 600

Norm. Intensity
wavelength (nm)
CzPy 300 K
Snap 300 K
Snap 77 K
CzPy 77 K

Figure 4-6. Emisssion spectra of Pt complexes of CzPy and Snap in MeTHF

72

complex was synthesized from a reactive Pt source. Two equivalents of the ligand were reacted
with one equivalent of [Me2Pt(SMe2)]2 , followed by an excess of dipivaloylmethane and
potassium carbonate to yield SnapPtDpm. Figure 4.6 shows the emission spectra of SnapPtDpm
and analogous CzPyPtDpm in PMMA matrix. Both compounds are non-emissive in fluid solution
and weakly luminescent in polymer matrix. SnapPtDpm is red shifted from CzPyPtDpm, likely
due to electronic donation of the quaternary bridinging carbon to the pyridine ring, destabilizing
the LUMO. Surprisingly, the measured photoluminescent quantum yields are within experimental
error of each other. This indicated that the hypothesized mechanism for non-radiative deactivation
of the excited state of the CzPy ligand system may not in fact be due to CN bond rotation.
Derivatives of hemimacrocyclic tetradentate Pt complexes like PtNON with similar methylene
bridges also have unexpectedly decreased photoluminescent quantum yields.

3. Hemimacrocyclic ligand systems
Highly efficient tetradentate hemimacrocyclic platinum complexes are generally composed
of two bidentate ligands linked at the HOMO position. That is, in a phenyl pyridine type ligand
system, the two phenylpyridines are linked at the phenyl, usually by an oxygen atom. An under
explored type of tetradentate hemimacrocycle is one in which the chemical linkage is made through
the locus of the LUMO, for instance, in the case of phenylpyridine, through the pyridine moiety.
Linkage between two pyridines via an oxygen atom (a dipyridyl ether) would significantly
destabilize the LUMO due to the electron donating effect of oxygen. Direct linkage of the sp2
hybridized pyridine units produces unfavorable bond angles for the resulting hemimacrocyle, as
well as lowering the ligand centered triplet energy, which is not desireable for blue emitting
73

complexes. Contraction of the pyridine to an imine allows the bridging carbons to be sp3
hybridized, reducing their bond angles to circa 109 degrees rather than 120 degrees, ultimately
leading to a favorable coordination geometry for platinum. Such a compound is known: ethylene
bis (imido anthracene), a low energy absorbing compound. This compound is red in appearance,
due both to the low ligand centered excited state energy of anthracene and the relative ease of
reducing an imine with respect to pyridine. A potentially blue emitting analogue was therefore
designed by transmuting anthracene to carbazole. Not only is carbazole higher in ligand centered
excited state energy, but the carbazole nitrogen is strongly pi basic to the imine fragment. Indeed,
a nitrogen bound to an imine is
electronicallty distinct enough to have a
different functional group name: amidine.
The chemistry of amidines is different
from imines synthetically as well as
electronically. While imines can be synthesized by condensation of amines and aldehydes, amines
react with formamides by transamidation, effectively transferring a formyl group to the amine.
This precludes condensation of N-formylcarbazole with ethylenediamine as shown in figure 4.7 to
form the desired bisamidine ligand. Several routes to amidine formation were identified. One
involves activation of the formamide with a strongly electrophilic chlorinating agent such as PCl5,
similar to Vilsmeier-Haack conditions. A second route involves nucleophilic attack of the
carbazole nitrogen on an isothiocyanide to form a thiourea. The thiourea could then be reduced
by Ranney nickel to the desired product. Thirdly, a recent report demonstrates the direct
conversion of isocyanides to amidines with amines in the presence of N-heterocyclic carbenes.
None of these routes are particularly promising, so an alternate ligand design was sought.

Figure 4-7. Undesired side reaction resulting from attempted
amidine condensation

74

Rather than destabilize the LUMO of
an imine by transmutation to an amidine,
transmutation of the sp2 hydridized amidine
carbon to sp3 hybridized phosphine is
expected to be both harder to reduce and insert
a break in conjugation between phenyl and
imine. The contracted bond angle of the four
coordinate phosphorus (V) center is
insufficient for formation of a six membered metallocycle, but well suited to form a five membered
metallocycle. Thus, a bis phosphine imide ligand system was designed. Metal complexes of
phosphine imides, in particular of the designed ligand, are known. Both nickel (II) and palladium
(II) complexes are known, though they are not carbon cyclometallated and are only bound through
nitrogen.
A common route to phosphine imides is an arrested Staudinger reduction, in which an alkyl
azide is reacted with a phosphine to produce the desired phosphine imide with expulsion of
dinitrogen. The usual Staudinger reduction conditions include aqueous workup to hydrolyze the
product to phosphine oxide and amine via a mechanism similar to the Wittig reaction. As such,
the ligand is highly moisture sensitive. The Staudinger route was not pursued due to the
inaccessibility of ethylene diazide, not to mention the safety hazards associated with handling this
starting material at gram scale. Instead, a Kirsonov reaction was pursued. Triphenylphosphine
was oxidized by treatment with elemental bromine. The resulting triphenylphosphinedibromide
was then treated with ethylene diamine to yield the ligand as its dihydrobromide salt. This
compound is highly insoluble is all common organic solvents with the exception of chloroform

Figure 4-8. Generic scheme of phosphine imide formation

75

and dimethylformamide. Unfortunately, neither of those solvents is compatible with the strong
base required to deprotonate the ligand to its neutral form. After an extensive optimization of
reaction conditions, a heterogeneous deprotonation protocol was found. A suspension of the
dihydrobromide salt in THF was stirred rapidly at -78C while two equivalents of nBuLi were added
dropwise over a period of one hour. Addition of the base at any faster rate resulted in a color
change to orange and intractable decomposition of the ligand. After addition of the base under
anaerobic conditions the reaction was allowed to warm to room temperature. Dimethyl platinum
dimethylsufide dimer was added to the colorless solution, immediately producing a bright yellow
solution. Upon stirring overnight, a bright yellow precipitate formed, which was collected by
filtration. Analysis by NMR and LCMS indicated the presence of two components which were

Figure 4-9. X-ray crystal structure of PtPie

76

tentatively assigned to the desired
compound as a minor component, and
the mono cyclometallated
methylplatinum intermediate as the
major component. Heating the crude
product mixture in refluxing toluene
lead to an increase of the presence of
the desired product, but also a myiad
of unidentified decomposition
products as judged by NMR. Instead,
the crude product mixture was directly subjected to vacuum sublimation. At 275C and 10
-6
torr, a
bright yellow solid sublimed as needle like crystals. The X-ray crystal structure is shown in figure
4.9. The compound crystallizes in the polar space group Fdd2. Both enantiomers of the compound
(point group: C2) are present in the crystal as a racemic pair. Furthermore, the platinum centers
are more than 7 angstroms apart, preventing Pt-Pt interactions which lead to a redshifted dimer
emission. As such, the solid is emits green with a moderately high quantum yield of 10 %. Figure
4.10 shows the excitation and emission spectra of solid PtPie suspended in DMSO. The emission
is Gaussian, with an onset of 425 nm and a maximum of 510 nm, which is consistent with the
calculated frozen (430 nm) and relaxed (493 nm) triplet state energies calculated by DFT. The
excited state decay time as measured by TCSPC is 420 ns at room temperature, and 3.8 us at 77K.
The extreme insolubility of this compound is common organic solvents precludes examination of
its solution state photophysical parameters, or those in doped polymer matrix. Work is ongoing to
300 350 400 450 500 550 600 650 700

Normalized Intensity
Wavelength (nm)
ex. (560 nm em)
em (350 nm ex)

Figure 4-10. Excitation and emission spectra of solid PtPie suspended
in DMSO

77

investigate the photophysical parameters of coevaporated doped films in wide band gap hosts en
route to vapor deposited OLED devices.

78

Bibliography

1. Patrick J. G. Saris and Mark E. Thompson, Gram Scale Synthesis of Benzophenanthroline
and Its Blue Phosphorescent Platinum Complex. Organic Letters 2016 18 (16), 3960-3963.
DOI: 10.1021/acs.orglett.6b01693

2. Daniel Sylvinson M. R., Hsiao-Fan Chen, Lauren M. Martin, Patrick J. G. Saris, and Mark
E. Thompson, Rapid Multiscale Computational Screening for OLED Host Materials. ACS
Applied Materials & Interfaces 2019 11 (5), 5276-5288. DOI: 10.1021/acsami.8b16225

Abstract (if available)

Abstract Blue emitting phosphorescent organic light emitting diodes (OLEDs) suffer from shorter operational lifetimes than their green to red emitting counterparts, but new materials will enable robust device architectures for more resilient blue luminescence. Chemical decomposition of the active layer of an OLED, consisting of a conductive organic host material doped with a luminescent organometallic complex of Ir or Pt, is the primary source of luminance loss in state of the art devices. The instability particular to blue OLEDs stems from the molecular design principle for blue shifting the luminescence of known green dopant phosphors: widening the HOMO-LUMO gap. Destabilization of the LUMO by any means necessitates conduction band energies well shallow of -2 eV, limiting the choice of suitable host materials to inherently unstable structure classes such as phosphine oxides or to materials with suboptimal band alignment for high power efficiency. Alternatively, stabilization of the HOMO by fluorine substitution leads to electron promoted decomposition of the dopant phosphor by fluoride loss. Rather than substitution of CH with more electronegative CF, isoelectronic transmutation of CH with more electronegative N is a robust strategy for tuning frontier orbital energies, particularly for HOMO stabilization. For instance, benzophenanthroline, a fluorine free tetracyclic heterotriphenylene ligand, exhibits efficient sky blue phosphorescence with similar properties to the ubiquitous difluorophenylpyridine ligand (e.g. Firpic), without the potential for dehalogenative degradation mechanisms.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Simple complexes: synthesis and photophysical studies of luminescent, monovalent, 2-coordinate carbene-coinage metal complexes and higher coordination geometries

PDF

Molecular approaches to solve the blue problem in organic light emitting diode display and lighting applications

PDF

Synthesis and photophysical study of phosphorescent hetero-cyclometalated organometallic complexes involving phosphino-carbon ligands

PDF

High energy hosts and blue emitters for phosphorescent organic light emitting diodes

PDF

Novel pyrimidine-based hole blocking materials for long-lived and highly efficient organic light emitting diodes

PDF

Photophysical properties of luminescent iridium and coinage metal complexes

PDF

Studies of molecular orientation using iridium phosphors and integration of corannulene into organic light emitting diodes (OLEDs)

PDF

Molecular design of pyridine and quinoline based compounds for organic optoelectronics

PDF

Development of N-type chromophores for organic photovoltaics, and thermally activated delayed fluorescence NHC complexes for organic light-emitting diodes

PDF

Ir(III) and Pt(II) phosphorescent emitters in organic light emitting diodes: from materials development to light out-coupling

PDF

Molecular design strategies for blue organic light emitting diodes

PDF

Synthesis, structural and photophysical characterization of phosphorescent three-coordinate Cu(I)-N-heterocyclic carbene complexes

PDF

Synthesis, photo- and electroluminescence of three- and two-coordinate coinage metal complexes featuring non-N-heterocyclic carbene and non-conventional N-heterocyclic carbene ligands

PDF

Molecular and morphological effects on the operational parameters of organic solar cells

PDF

Controlling structure and photophysical properties of bi triple perovskite through substitutional series Cs₃Bi₂Br₍₉₋ₓ₎Iₓ

PDF

Improving the sustainability of conjugated polymer synthesis via direct arylation polymerization

Asset Metadata

Creator Saris, Patrick James Gerhardus (author)

Core Title Qualitative quantum chemical description of the structure-color relationships in phosphorescent organometallic complexes

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Chemistry

Publication Date 03/05/2019

Defense Date 01/17/2019

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,organic light emitting diode,organometallic,phosphorescence

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Thompson, Mark (committee chair), El-Naggar, Moh (committee member), Melot, Brent (committee member)

Creator Email psaris@gmail.com,psaris@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-129401

Unique identifier UC11675791

Identifier etd-SarisPatri-7131.pdf (filename),usctheses-c89-129401 (legacy record id)

Legacy Identifier etd-SarisPatri-7131.pdf

Dmrecord 129401

Document Type Dissertation

Format application/pdf (imt)

Rights Saris, Patrick James Gerhardus

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA