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

A series of Al 2p, K 2p, O 1s and N 1s core-level spectra have been used to characterize the interaction between potassium (K) and tris(8-hydroxyquinoline) aluminium (Alq₃) molecules in the K-doped Alq₃ layer. All core-level spectra were tuned to be very surface sensitive in selecting various photon energies provided by the wide-range beamline at the National Synchrotron Radiation Research Center, Taiwan. A critical K concentration (x = 2.4) exists in the K-doped Alq₃ layer, below which the K-doped atoms generate a strained environment near the O and N atoms within 8-quinolinoline ligands. This creates new O 1s and N 1s components on the lower binding-energy side. Above the critical K coverage, the K-doped atoms attach the O atoms in the Al—O—C bonds next to the phenoxide ring and replace Al—O—C bonds by forming K—O—C bonds. An Alq₃ molecule is disassembled into Alq₂ and Kq by bond cutting and bond formation. The Alq₂ molecule can be further dissociated into Alq, or even Al, through subsequent formations of Kq.

1. Introduction

In past years, alkali metals and alkali metal compounds have been used as electron injection materials that are doped into electron transport layers in order to improve the performance of organic light-emitting devices (OLEDs) (Parthasarathy et al., 2001; Lee et al., 2005; Ganzorig et al., 2001; Stößel et al., 2000; Brown et al., 2003). The alkali-metal-doped organic layer is usually formed by depositing an alkali metal layer between the electron transport layer tris(8-hydroxyquinoline) aluminium (Alq₃) and cathode material aluminium (Al). The molecular structure of Alq₃, sketched in Fig. 1, consists of pyridyl and phenoxide rings. The performance of lithium (Li) doped Alq₃ layers near the cathode interface in OLED devices has been intensively studied (Haskal et al., 1997; Kido et al., 1993; Kido & Matsumoto, 1998). Kido & Matsumoto (1998) demonstrated that Li-doped Alq₃ layers are effective in lowering the driving voltage for the electron injection. The Li-doped atoms play a role in donating electrons to the pyridyl ring of Alq₃ such that the conductivity of the Alq₃ layer increases. Recently, another alkali metal, potassium (K), has attracted much attention as it is easier for K to donate electrons compared with Li, owing to the lower ionization energy of K. Fundamental studies on the K-doped Alq₃ layer have been investigated by the valence band and core-level spectra obtained through ultraviolet photoemission spectroscopy and X-ray photoemission spectroscopy (Johansson et al., 1999; Schwieger et al., 2001). Johansson et al. (1999) reported that electron transfer occurred from K to three ligands of the Alq₃, and that a new spectral feature appeared in the valence band. Later, Schwieger et al. (2001) reported that a low intercalation of K led to a rigid energy shift for all electronic levels. A high intercalation of K in Alq₃ resulted in structural modification of the Alq₃ molecules by forming an organometallic complex. However, detailed structural modification by the K-doped atoms was not reported. It is of fundamental interest to investigate how the structure modification occurs at the K/Alq₃ interface. A pristine Alq₃ layer is doped with K from low to high dosage in order to fully understand the interaction between K and Alq₃ at the K/Alq₃ interface.

Figure 1 Schematic drawing of an Alq3 molecule structure consisting of pyridyl and phenoxide rings.

In this article we present a discovery of bond cutting in the K-doped Alq₃ layer. A series of Al 2p, K 2p, O 1s and N 1s core-level spectra were measured in situ during the doping of K onto an Alq₃ layer by synchrotron radiation photoemission. With the benefits of strong intensity and surface sensitivity offered by this technique, the cutting of the Al—O—C bonds in the Alq₃ molecule by the K-doped atoms was observed.

2. Experimental

Mirror-polished n-type Si(100) samples (ρ = 1–10 Ω cm) were pre-oxidized using the Ishizaka and Shiraki method in order to remove possible carbon contamination (Ishizaka & Shiraki, 1986). The Si(100) samples were then fixed on a rotatable molybdenum sample holder with two pairs of tantalum plates in a photoemission chamber pumping down to a base pressure of ∼7 × 10⁻¹¹ torr. In order to remove the protected oxide layer, the Si(100) samples were annealed stepwise up to 1153 K to obtain a clean Si(001) 2 × 1 surface on which Alq₃ was vacuum-evaporated using an Omicron EFM3 evaporator. The deposition of Alq₃ was recorded by the evaporation time and the evaporation flux of EFM3. The deposition rate of Alq₃ was estimated to be 0.14 ML min⁻¹. K atoms were evaporated from a SEAS getter onto the pre-evaporated Alq₃ layer. The deposition of K was recorded by the evaporation time and the power used to heat the SEAS getter. The K flux out of the SEAS was divergent so it was not possible to estimate the deposition rate of K onto Alq₃. The K concentration in the Alq₃ layer was determined by comparing the relative peak areas of K 2p and C 1s under various deposition conditions, taking into account their relative cross-section factors. The cross-section factors of K 2p and C 1s at a photon energy of 330 eV are 2.4 and 0.35, respectively (Yeh & Lindau, 1985). The concentration of K in the Alq₃ layer was expressed as x in all the core-level spectra.

The operation pressures during the deposition of Alq₃ and K were maintained below 1.5 × 10⁻⁹ and 6 × 10⁻¹⁰ torr, respectively. Normal emission photoelectrons were collected with an Omicron EA125 hemispherical analyzer in the photoemission chamber using the photon energies provided by the wide-range (WR) spherical-grating-monochromator beamline at National Synchrotron Radiation Research Center in Taiwan. Al 2p, K 2p, O 1s and N 1s core levels were taken at an incident angle of 70° with their kinetic energies tuned to 40–60 eV, in order to obtain surface-sensitive spectra according to the universal curve (Brundle, 1975). The pass energy of the Omicron EA125 hemispherical analyzer for all the photoelectrons was set to 10 eV. The corresponding electron energy resolution was 0.12 eV. However, the total energy resolution was limited by the photon energy resolution. The photon energy resolution was affected by the selection of photon energy and the grooves of the spherical grating at the WR beamline. The photon energy resolutions were determined to be 0.16, 0.44, 0.44 and 0.66 eV for photon energies of 136, 330, 440 and 580 eV, respectively. In the analysis of the photoelectron spectra, core-level spectra were normalized by the light intensity. The light intensity was characterized by measuring the mesh current at the entrance of the photoemission chamber. No charging problem was observed during the photoemission measurements. A sample of Alq₃ intercalated with Ag was used to determine the Fermi edge since no Fermi edge was observed even at a high intercalation of K into the Alq₃ layer. All the core-level spectra were de-convoluted in a Voigt function and the backgrounds were subtracted in a power-law function by using the curve-fit program UNIFIT 2006. The branching ratios of Al 2p and K 2p were both fixed at 0.5. The spin-orbit splittings of Al 2p and K 2p were fixed at 0.4 eV and 0.75 eV, respectively. The evolution of the curve-fit in the convolution was plotted separately for all core-level spectra.

3. Results and discussion

Fig. 2(a) shows the surface-sensitive Al 2p core-level spectra in situ taken at a photon energy of 136 eV during deposition of K onto a thick Alq₃ layer. The Al 2p signal from the pristine Alq₃ is located at a binding energy of 74.6 eV with a peak width of 1.2 eV. Upon the deposition of K, the Al 2p signal slightly shifts to higher binding energy and its peak width becomes broader (∼1.4 eV). This is attributed to the band bending resulting from charge donation of K into the lowest unoccupied molecular orbital (LUMO) of the Alq₃ layer (Ishii et al., 1999). Above the critical K concentration (at x = 2.4), a new peak (A) begins to appear and becomes an apparent shoulder at a binding energy of ∼73.6 eV at x = 3.5. The shoulder A is about 2 eV in width higher than the Al 2p width in pristine Alq₃ (∼1.2 eV). This indicates that the shoulder A consists of at least two Al 2p signals of Al-related oxidation states in organometallic complex at x = 3.5. In other words, at least two new Al bonding environments are induced by the doping of K into Alq₃. The new types of Al bonding environments can be obtained from the deconvoluted data of the Al 2p spectra shown in Fig. 2(b), where four Al 2p components (Al-a, Al-b, Al-c and Al-d) are required to obtain a consistent fit for all the Al 2p spectra. The binding energies and Gaussian widths of the four Al 2p components are summarized in Table 1. The curve-fit with three Al 2p components is excluded since their Gaussian widths scatter too much. The Al-a peak at a binding energy of 74.6 eV arises from the pristine Alq₃. The Al-b, Al-c and Al-d peaks are the induced components, which begin to appear at x = 2.4, 2.9 and 3.5, respectively. The four Al 2p components are strongly correlated to three Al oxidation states (Al³⁺, Al²⁺, Al⁺) and metallic Al according to their relative energy positions reported in a study of Fe₃Al-type alloys (López et al., 1997). From a chemical perspective the Al-b, Al-c and Al-d components indicate the existence of three different bonding environments for the Al atoms. This supports the bond cutting and bond formation occurring at the Al—O—C bonds next to the phenoxide rings in Alq₃ at x > 2.4 (critical K concentration) (Burrows et al., 1996). Very probably the Alq₃ molecule is disassembled into Kq and Alq₂ through the formation of Kq at x = 2.4. Alq₂ is further dissociated into Alq or even Al through subsequent formations of Kq. Note that the Al 2p_3/2 of the Al-d component is located at a binding energy of 72.6 eV that is consistent with the pure Al metal data (Gredelj et al., 2001). Thus Al-a, Al-b, Al-c and Al-d are assigned as the Al 2p signals from Alq₃, Alq₂, Alq and Al, respectively. In other words, bond cutting occurs during the doping of K atoms into the Alq₃ layer.

Table 1
Binding energies and Gaussian widths of Al 2p_3/2 components and their corresponding area percentages

Components	Evaporationtime (min)	K concentration(x)	Binding energy(eV) Al 2p3/2	Gaussianwidth (eV)	Area percentage(%)
Al-a	0	0	74.49	1.20	100
	3	0.2	74.59	1.39	100
	9	0.7	74.66	1.38	100
	16	1.3	74.78	1.30	100
	21	1.7	74.77	1.37	100
	27	2.1	74.72	1.34	100
	31	2.4	74.64	1.37	90.4
	37	2.9	74.88	1.36	88.8
	45	3.5	74.86	1.35	52.3
Al-b	31	2.4	73.91	1.36	9.6
	37	2.9	74.16	1.36	9.3
	45	3.5	74.11	1.34	23.5
Al-c	37	2.9	73.36	1.37	1.9
	45	3.5	73.34	1.34	20.5
Al-d	45	3.5	72.61	1.36	3.7

Figure 2 (a) Surface-sensitive Al 2p core-level spectra taken in situ at a photon energy of 136 eV during deposition of K onto a thick Alq3 layer. (b) Curve fit of the Al 2p spectra in (a).

Fig. 3(a) shows the surface-sensitive K 2p core-level spectra in situ taken at a photon energy of 330 eV during deposition of K onto a thick Alq₃ layer. The K 2p signal increases in intensity with deposition time, which agrees well with the amount of K doped into the Alq₃ layer. The energy of K 2p also shifts to higher binding energy owing to band bending as observed for the Al 2p signals in Fig. 2(a). Similar to the Al 2p spectra, the K 2p signal becomes broader at x = 3.5, indicating the existence of new K bonding environments. The new types of K bonding environments can be obtained by the deconvoluted data of the K 2p spectra, where two K 2p doublets (K-a and K-b) are required to obtain a consistent fit for all the K 2p spectra. The binding-energy values and Gaussian widths of K-a and K-b are listed in Table 2. The bond cutting of Alq₃ by K is also supported by the deconvoluted K 2p spectra in Fig. 3(b), where a new K 2p doublet (K-b) begins to appear at x = 2.4, concurrent with the appearance of the Al-b chemical state in Fig. 2(b). The newly formed doublet K-b is thus assigned as the K 2p from Kq since one of the Al—q bonds in Alq₃ is cut via the reaction Alq₃ + K = Alq₂ + Kq. The doublet K-a is assigned to the K 2p signal from the K atoms doped in Alq₃ before bond cutting (at x < 2.4).

Table 2
Binding energies and Gaussian widths of K 2p_3/2 components and their corresponding area percentages

Components	Evaporationtime (min)	K concentration(x)	Binding energy (eV) K 2p3/2	Gaussianwidth (eV)	Area percentage(%)
K-a	3	0.2	294.16	1.72	100
	9	0.7	294.31	1.61	100
	16	1.3	294.71	1.63	100
	21	1.7	294.83	1.64	100
	27	2.1	295.07	1.64	100
	31	2.4	294.70	1.62	88.3
	37	2.9	294.90	1.65	78.2
	45	3.5	294.68	1.59	65.6
K-b	31	2.4	295.29	1.73	11.7
	37	2.9	295.50	1.73	21.8
	45	3.5	295.28	1.79	34.4

Figure 3 (a) Surface-sensitive K 2p core-level spectra taken in situ at a photon energy of 330 eV during deposition of K onto a thick Alq3 layer. (b) Curve fit of the K 2p spectra in (a).

Fig. 4(a) shows the surface-sensitive O 1s in situ core-level spectra taken at a photon energy of 580 eV during deposition of K onto a thick Alq₃ layer. The prominent peak at a binding energy of 531.6 eV is the O 1s peak from the pristine Alq₃. Upon the deposition of K, the O 1s signal exhibits an energy shift to higher binding energy, owing to band bending. The line shape of the O 1s peak becomes broad and asymmetric at x > 1.7. Two new signals B1 and B2 clearly appear at a binding energy of ∼531 eV (x = 2.9) and ∼529 eV (x = 3.5), respectively. The new O 1s signals can be further characterized through the deconvoluted data in Fig. 4(b) where three components O-a, O-b and O-c are required to obtain a consistent fit for all the O 1s spectra. The binding-energy values and Gaussian widths of the three components are listed in Table 3. The component O-a is assigned from the pristine Alq₃, which is emitted from the O atoms bonded with Al in Alq₃. The component O-b starts to appear (Fig. 4b) before bond cutting (x < 2.4). This suggests that the O-b component is not from the K—O bonds that form after bond cutting (x > 2.4) and will be described later. Very probably the K-doped atoms generate a strained environment near the O atoms in Alq₃ such that the Al—O—C bond angle changes and results in the component O-b. Note that the component O-c starts to appear (Fig. 4b) at x = 2.4, the same as the beginning of the bond cutting (Alq₃ + K = Alq₂ + Kq). The component O-c is therefore from the O atom in either Kq or Alq₂. The component O-c is assigned to the O atom in Kq rather than that in Alq₂, since the O 1s in Alq₂ exhibits the same energy as that in Alq₃ based on the following arguments. First, the O 1s signal is emitted from the Al—O—C bonds next to the phenoxide rings in either Alq₃ or Alq₂. Note that there exist two further reactions, Alq₂ + K = Alq + Kq and Alq + K = Al + Kq, at deposition times longer than x = 2.4, according to the Al 2p spectra in Fig. 2. If the energy position of O 1s in Alq₂ was different from that in Alq₃, three O 1s components from Alq₃, Alq₂ and Alq, respectively, should exist. However, no new O 1s component appears in subsequent bond cutting occurring at x = 2.9 and at x = 3.5. This suggests that the induced O 1s signals from Alq₂ and Alq may overlap with that from Alq₃ (the O-b component). According to the fitted data in Table 3, the amount of O-b increases at x > 2.4. Additionally, the amount of the pristine Alq₃ (O-a) decreases with deposition time. This supports the conclusion that the component O-b from the Al—O—C bond next to the phenoxide ring is the same in Alq₃, Alq₂ and Alq.

Table 3
Binding energies and Gaussian widths of O 1s components and their corresponding area percentages

Components	Evaporationtime (min)	K concentration(x)	Binding energy (eV) O 1s	Gaussianwidth (eV)	Area percentage(%)
O-a	0	0	531.56	1.54	100
	3	0.2	531.84	1.75	100
	9	0.7	532.00	1.70	96.7
	16	1.3	531.86	1.69	90.9
	21	1.7	532.03	1.67	79.4
	27	2.1	532.25	1.64	64.9
	31	2.4	532.52	1.64	57.0
	37	2.9	532.29	1.68	31.4
	45	3.5	532.55	1.62	30.3
O-b	9	0.7	530.86	1.71	3.3
	16	1.3	530.70	1.67	9.1
	21	1.7	530.86	1.66	20.6
	27	2.1	531.11	1.66	35.1
	31	2.4	531.38	1.62	38.7
	37	2.9	531.15	1.71	55.9
	45	3.5	531.42	1.62	53.3
O-c	31	2.4	529.45	1.61	4.3
	37	2.9	529.14	1.65	12.7
	45	3.5	529.48	1.60	16.4

Figure 4 (a) Surface-sensitive O 1s core-level spectra taken in situ at a photon energy of 580 eV during deposition of K onto a thick Alq3 layer. (b) Curve fit of the O 1s spectra in (a).

Fig. 5(a) shows the surface-sensitive N 1s core-level spectra taken at a photon energy of 440 eV during deposition of K onto a thick Alq₃ layer. The N 1s signal from pristine Alq₃ is located at ∼400.2 eV, which shifts towards higher binding energy upon K deposition owing to band bending. A broad peak C clearly appears on the lower binding-energy side at x = 0.7. The broad peak C increases in intensity and becomes the dominant signal at higher K concentration. The observation of N 1s core-level spectra is consistent with previously published results (Greczynski et al., 2000; Osada et al., 1999). The bonding environments of N atoms can be further characterized via the fitted data in Fig. 5(b). Two N 1s components, N-a and N-b, are required to obtain a consistent fitted result for all the N 1s spectra. The binding-energy values and Gaussian widths of the two components are listed in Table 4. The component N-a is from pristine Alq₃, which is assigned from the N atom in the C=N—C bond in the pyridyl ring (Burrows et al., 1996). The component N-b appears at x = 0.2, which increases in intensity with K concentration. Very probably the K-doped atoms generate a strained environment near the N atoms in the pyridyl ring in Alq₃ such that the C=N—C bond angle changes and results in the component N-b. At x = 3.5, N-b becomes the dominant signal. Note that no other N 1s-induced components appear at x = 2.4, 2.9 or 3.5, as was observed for the Al 2p components. This indicates that the cutting of C=N—C bonds in the pyridyl ring by the doped K atoms does not occur.

Table 4
Binding energies and Gaussian widths of N 1s components and their corresponding area percentages

Components	Evaporationtime (min)	K concentration(x)	Binding energy (eV) N 1s	Gaussianwidth (eV)	Area percentage(%)
N-a	0	0	400.18	1.18	100
	3	0.2	400.21	1.40	92.2
	9	0.7	400.35	1.42	75.2
	16	1.3	400.31	1.43	53.5
	21	1.7	400.30	1.43	43.3
	27	2.1	400.56	1.43	23.9
	31	2.4	400.68	1.45	17.6
	37	2.9	400.40	1.40	9.8
	45	3.5	400.53	1.42	9.0
N-b	3	0.2	398.58	1.56	7.8
	9	0.7	398.69	1.53	24.8
	16	1.3	398.68	1.65	46.5
	21	1.7	398.70	1.54	56.7
	27	2.1	398.96	1.58	76.1
	31	2.4	399.06	1.64	82.4
	37	2.9	398.77	1.65	90.2
	45	3.5	398.89	1.69	91.0

Figure 5 (a) Surface-sensitive N 1s core-level spectra taken in situ at a photon energy of 440 eV during deposition of K onto a thick Alq3 layer. (b) Curve fit of the N 1s spectra in (a).

Note that the core-level spectra of K 2p, Al 2p, O 1s and N 1s shift towards higher binding energy upon K deposition, which is opposite to the photoemission screening effect. Similar to the work of Xu et al. (2002), the energy shift to higher binding energy is attributed to charge transfer in the initial state. The charge donated by K is considered to be delocalized in the valence band of Alq₃ since the K 2p core-level is also shifted to higher binding energy.

4. Conclusion

K is very reactive with Alq₃. A critical K concentration exists below which the K-doped atoms are near the O and N atoms within 8-quinolinoline ligands without bond cutting. Above the critical K concentration the Al—O—C bonds next to the phenoxide ring are cut. An Alq₃ molecule may be dis­assembled into Kq and Alq₂ by bond cutting and bond formation. Alq₂ is likely to be further dissociated into Alq or even Al through subsequent formations of Kq.