Bond cutting in Cs-doped tris­(8-hydroxyquinoline) aluminium

A series of Cs 4d and Al 2p spectra associated with valence-band and cut-off spectra have been used to characterize the interaction between caesium and tris(8-hydroxyquinoline) aluminium (Alq₃) molecules in a Cs-doped Alq₃ layer. The Cs 4d and Al 2p spectra were tuned to be very surface sensitive by selecting a photon energy of 120 eV at the National Synchrotron Radiation Research Center, Taiwan. A critical Cs concentration exists, above which a new Al 2p signal appears next to the Al 2p peak of Alq₃ in the lower binding-energy side. The Al 2p signal was analyzed and assigned as being contributed from a mixture of Alq₂, Alq and Al. Experimental data supported the observation that bond cutting of Alq₃ by the doped Cs atoms occurred at high Cs doping concentration.

1. Introduction

Doping in organic materials has attracted attention in the past years owing to the improvement in the performance of organic light-emitting devices (OLEDs) (Tang et al., 1989; Shi & Tang, 1997). A metal-doped organic layer is usually formed by depositing a metal layer between the electron transport layer [tris(8-hydroxyquinoline) aluminium (Alq₃)] and a cathode material Al. A common metal-doped organic layer used in an OLED device is the Li-doped Alq₃ layer near the cathode interface (Haskal et al., 1997; Kido et al., 1993; Kido & Matsumoto, 1998). Kido & Matsumoto (1998) demonstrated that a Li-doped Alq₃ layer is effective in lowering the driving voltage by reducing the barrier height for the electron injection. In that study, Li donates electrons to the pyridyl ring of Alq₃, increasing the conductivity of the Alq₃ layer (Kido & Matsumoto, 1998). Recently, Cs-doped Alq₃ films have attracted attention because it is easier for Cs to donate electrons than Li owing to its lower ionization energy. Most work on Cs-doped Alq₃ has been focused on fundamental issues based on the results from the valence band and core-level measurements (Lai et al., 2003; Fung et al., 2004; Ding & Gao, 2005). The major finding relating to the donation of electrons from Cs is the generation of a gap state, reported by Lai et al. (2003). Later, Fung et al. (2004) and Ding & Gao (2005) further justified the formation of a Cs-induced gap state contributed by electron donation from Cs to nitrogen in the pyridyl ring.

The role of Cs atoms in Cs-doped Alq₃ does not simply result in a chemical shift of core levels. In this article we will present an experimental observation of the bond-cutting of Alq₃ induced by electron donation of Cs in the Cs-doped Alq₃ layer. A series of Cs 4d and Al 2p core-level spectra, along with valence-band and cut-off spectra, were recorded in situ by synchrotron radiation photoemission. With the benefits of strong intensity and surface sensitivity of the technique, bond cutting of Alq₃ by Cs was observed in the Cs-doped Alq₃ layer at high Cs doping concentration.

2. Experimental

Mirror-polished n-type Si(100) wafers (ρ = 1–10 Ω cm) were pre-oxidized using the method of Ishizaka & Shiraki (1986) in order to remove possible carbon contamination. The Si(100) substrates were then mounted on a heating holder inside an ultra-high-vacuum (UHV) photoemission chamber with a base pressure of less than 7 × 10⁻¹¹ torr. In order to remove the protected oxide layer the Si(100) substrates were annealed stepwise up to 1153 K to obtain a clean Si(001) 2 × 1 surface on which purified Alq₃ powder was vacuum-evaporated. Cs was evaporated from a SEAS getter onto the pre-evaporated Alq₃ layer. The operation pressures for depositing Alq₃ and Cs were maintained below 1.5 × 10⁻⁹ and 5 × 10⁻¹⁰ torr, respectively. Photoelectrons were collected with an OMICRON EA125 hemispherical analyzer in the UHV chamber located at the low-energy spherical-grating monochromator beamline at the National Synchrotron Radiation Research Center in Taiwan.

3. Results and discussion

The interaction between Cs and Alq₃ was monitored in situ using surface-sensitive Al 2p and Cs 4d core-level spectra (Fig. 1) recorded at a photon energy of 120 eV during deposition of Cs onto a thick Alq₃ layer. The probing depth is estimated to be ∼0.5 nm according to the universal curve (Brundle, 1975). The Al 2p signal from the pristine Alq₃ is located at a binding energy of 74.6 eV, which appears as a single peak in Alq₃. As can be seen in Fig. 1, the Al 2p signal reduces in intensity with Cs deposition time. No Cs signal was observed at a deposition time of less than 65 s within the detection limit. A very weak Cs 4d doublet (4d_5/2 and 4d_3/2) begins to appear upon further deposition and gradually becomes the dominant signal above 420 s. In addition to the above-mentioned Cs 4d and Al 2p signals, a weak signal (`M') also appears at a binding energy of ∼71.6 eV at 220 s. The reported binding energy of Al 2p in a pure Al metal is 72.7 eV (Gredelj et al., 2001). Based on the energy position, the Cs-induced M state is assigned as the Al 2p peak from the metallic Al or a molecule containing Al (Le et al., 2000). The full width at half-maximum of the M state is about 2 eV, which is larger than that of the Al 2p in pristine Alq₃ (∼1 eV). This suggests that more than one Al-related chemical state is generated owing to the interaction between Cs and Alq₃. In other words, new types of Al bonds are induced by the doping of Cs into Alq₃. Examining the molecular structure, it is probable that Cs atoms attach to the O atom in an 8-quinolinoline ligand in Alq₃, cut the Al—O bonds and form Csq and Alq₂ molecules. The probable chemical reaction is described as Cs + Alq₃ = Csq + Alq₂. However, the subsequent reaction between Cs and Alq₂ is also possible, which will be discussed later. Note that the later appearance of the M state (220 s) than that of the Cs 4d state (65 s) is not due to the cross-section effect since the Al 2p core exhibits a higher cross section than the Cs 4d core at the present excitation energy (Yeh & Lindau, 1985). The generation of new Al-related chemical states seems to require a critical Cs doping concentration. This implies that a critical Cs concentration is required for the chemical reaction Cs + Alq₃ = Csq + Alq₂ to occur.

Figure 1 Cs 4d and Al 2p core-level spectra from the Cs/Alq3 interface at different Cs coverages. The spectra were recorded at a photon energy of 120 eV.

Chemical information about the new Al-related chemical states can be extracted through de-convolution of the Cs 4d and Al 2p core-level spectra in Fig. 1. As shown in Fig. 2, two Cs 4d doublets (Cs-a and Cs-b) and four Al 2p peaks (Al-a, Al-b, Al-c and Al-d) are required to obtain a consistent fit for all the Cs 4d and Al 2p spectra. The binding energies of the Cs-a, Cs-b, Al-a, Al-b, Al-c and Al-d components and their corresponding Gaussian widths for all the Cs 4d and Al 2p spectra are summarized in Table 1. The Al-a component at 74.6 ± 0.1 eV arises from the pristine Alq₃. The binding energy values of Al-b, Al-c and Al-d are 73.5 ± 0.1, 72.1 ± 0.1 and 71.0 ± 0.1 eV, respectively, indicating the existence of three different bonding environments of Al. From a chemical view, Cs reacts rather strongly with the 8-quinolinoline ligands in Alq₃. Thus, it is possible to dissemble the Alq₃ molecule into Csq and Alq₂ by formation of Csq. Alq₂ is likely to be further dissociated into Alq or even Al in subsequent formations of Csq. Based on the charge transfer between an Al atom and an 8-quinolinoline ligand, the Al atom may donate valence charge towards the O atom in Alq, Alq₂ or Alq₃. The Al 2p component of pure Al is expected to exhibit a binding energy lower than that of Alq, Alq₂ and Alq₃. Al-b, Al-c and Al-d are thus assigned as the Al 2p signals from Alq₂, Alq and Al, respectively, since the binding energies of Al-b, Al-c and Al-d are in a reduction sequence. Note that the Al 2p component of the pure Al atom (71.0 eV) cut out of the Alq_x molecule, i.e. the Al-d component, exhibits a binding energy lower than that of pure Al metal (72.7 eV). The energy difference of Al 2p is probably due to the different bonding environment between free Al atoms and Al metallic bonds.

Table 1
Relative binding energies of the Cs 4d_5/2 and Al 2p_3/2 components and their corresponding Gaussian widths

Average values of binding energy and Gaussian width with error margin for each component are also listed.

	Coverage (s)	Binding energy (eV)	Gaussian width (eV)	Area percentage (%)
Cs-a	65	76.3	1.32	100
	160	76.4	1.35	100
	220	76.4	1.45	63.2
	300	76.3	1.45	49.9
	420	76.4	1.51	41.7
	660	76.4	1.53	39.2
	900	76.5	1.54	36.8
		Average 76.4 ± 0.1	Average 1.45 ± 0.13
Cs-b	220	76.7	1.45	36.8
	300	76.8	1.45	50.1
	420	76.8	1.50	58.3
	660	76.9	1.53	60.8
	900	77.0	1.54	63.2
		Average 76.8 ± 0.2	Average 1.49 ± 0.05
Al-a	0	74.5	1.05	100
	65	74.5	1.29	100
	160	74.5	1.39	100
	220	74.6	1.44	92.0
	300	74.6	1.42	85.2
	420	74.6	1.45	79.5
	660	74.6	1.43	69.4
	900	74.6	1.43	62.8
		Average 74.6 ± 0.1	Average 1.36 ± 0.31
Al-b	220	73.5	1.46	4.6
	300	73.6	1.44	7.5
	420	73.5	1.45	10.0
	660	73.4	1.42	13.3
	900	73.4	1.42	16.6
		Average 73.5 ± 0.1	Average 1.44 ± 0.02
Al-c	220	72.0	1.46	2.1
	300	72.0	1.47	5.0
	420	72.1	1.45	7.1
	660	72.1	1.46	11.3
	900	72.1	1.43	13.0
		Average 72.1 ± 0.1	Average 1.45 ± 0.02
Al-d	220	71.0	1.43	1.2
	300	71.0	1.47	2.2
	420	70.9	1.45	3.4
	660	70.9	1.42	6.0
	900	71.1	1.44	7.6
		Average 71.0 ± 0.1	Average 1.44 ± 0.03

Figure 2 Curve fit of the Cs 4d and Al 2p spectra from the Cs/Alq3 interface at different Cs coverages.

The existence of four Al 2p components implies that Alq₃ can be bond-cut by the Cs-doped atoms in the Alq₃ layer. Bond cutting of Alq₃ by Cs is also supported by the deconvoluted Cs spectra in Fig. 2, where a new Cs 4d doublet begins to appear at a deposition time longer than 220 s, concurrently with the appearance of the new Al-related chemical states. Cs-a is a state without chemical shift (no charge transfer), which is assigned as the Cs 4d component from the neutral Cs atoms in Alq₃. Cs-b is a state with chemical shift, which is assigned from Csq.

The assignments of the Cs 4d and the Al 2p components in the Cs-doped Alq₃ layer are further justified by their relative amounts at different Cs coverages, as shown in Figs. 3(a) and 3(b). The relative amount of each Cs 4d (or Al 2p) component is expressed as the relative peak area %. The Al-b, Al-c and Al-d components increase in intensity accompanied by an intensity reduction of the Al-a component at deposition times longer than 220 s. This implies that the Al-a component is from the pristine Alq₃ and that the Al-b, Al-c and Al-d components are from the induced Al-related chemical states. Note that the closer the component is to the Al-a component the greater the increase in intensity. The Al-b component is thus assigned as the decomposed Alq₂ molecule since it exhibits a highest peak area among the induced components. As for the Al-c and Al-d components, we assign them as the Alq molecule and the free Al atom, respectively. Note that Cs-a decreases and Cs-b increases at deposition times longer than 220 s. This supports the assignment of Cs-b as Csq since the generation of Csq is expected to increase with Cs doping concentration. Cs-a is assigned from the neutral Cs atoms in Alq₃, since some Cs atoms are consumed to react with Alq₃ in favor of Csq formation and result in the peak area reduction of Cs-a.

Figure 3 Relative amounts of each component of Cs and Al core levels for different Cs coverages: (a) Cs 4d and (b) Al 2p. The relative amount of each component is expressed as peak area %.

Fig. 4(a) shows the surface-sensitive valence band spectra of Alq₃ doped with different Cs coverages, recorded in situ at a photon energy of 46 eV. The peaks A–G are the characteristic peaks for Alq₃, which reduce in intensity with Cs doping (Ishii & Seki, 1997; Sugiyama et al., 1998). No Fermi edge was observed, even at 900 s (the highest Cs doping in our study), indicating that no metallic Cs exists at the Cs/Alq₃ interface. The highest occupied molecular orbital (HOMO) peak `A' stays at ∼3 eV at a deposition time less than 65 s, but gradually shifts towards higher binding energy and reaches ∼4 eV at 220 s. The HOMO peak continues to shift towards higher binding energy at 65–220 s, which is before the bond cutting at 220 s. Similar to surface band bending, some extent of charge may be donated from Cs to Alq₃ near the surface to form a dipole layer, which causes the downward HOMO peak shift. Note that the HOMO peak remains at the same position at deposition times longer than 400 s.

Figure 4 (a) Surface-sensitive valence-band spectra from the Cs/Alq3 interface at different Cs coverages. (b) Cut-off spectra from the Cs/Alq3 interface at different Cs coverages. The spectra were recorded at a photon energy of 46 eV.

The shift of the HOMO peak is consistent with the energy shift of the cut-off at different Cs coverages, as shown in Fig. 4(b). The cut-off shifts to higher binding energy, which indicates a downward bending of the vacuum level owing to the doping of Cs. The downward bending is attributed to the `donor' character of Cs atoms in Alq₃. The charge donation from Cs may result in an effective dipole layer near the Cs/Alq₃ interface and cause the downward bending and the downward HOMO peak shift.

In summary, Cs is very reactive with Alq₃; however, a critical concentration exists below which no chemical reaction occurs. At a concentration lower than the critical one, Cs atoms may donate electrons to Alq₃ near the surface and cause downward bending of the cut-off and the HOMO peak. At a concentration higher than the critical one, bond cutting of Alq₃ by Cs occurs. An Alq₃ molecule may be dissembled into Csq and Alq₂ by formation of Csq through a bond-cutting mechanism. Subsequent formations of Alq and Al are also possible by bond cutting of Alq₂ and Alq by Cs.