18
1. Additional data 1.1 – Formation of oxides and their identification using XPS The first major challenge of this study was the preparation and characterization of the oxide substrates. Oxide films were grown on metal substrates by means of controlled oxidation in-situ. The goal was to be able to grow oxides of various stoichiometries and metal oxidation states, determine their compositions using XPS, and draw correlations between their properties and energy-level alignment by eliminating extraneous factors related to ex-situ sample preparation. By analyzing the high-resolution XPS spectra of the main core level emission features of the transition metals we could determine the composition of each oxide. High- resolution XPS spectra of the main transition metal photoemission peaks for all the oxides tested in this work are shown in Fig. 1.1.1. Figure 1.1.1 – XPS spectra of transition metals and transition metal oxides Oxide identification is usually accomplished by measuring the chemical shift of high resolution core-level photoemission peaks of the oxidized transition metal. In the majority of cases the core-level binding energy is positively correlated with the oxidation state of the element being measured. By comparing the binding energy of the oxidized metal core level with that of the non-oxidized metallic standard, and comparing with several consistent literature reports of the same, one can identify the oxide present in the sample. The accepted literature values are generally taken from single crystal or polycrystalline samples whose compositions have been determined by x-ray diffraction. Very often however there is a significant spread in reported values of binding energies, which can be a result of differing sample preparation procedures, sample conditions, and calibration procedures. Sometimes the spread in reported values for one oxide is larger than the difference in binding 1 SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159 NATURE MATERIALS | www.nature.com/naturematerials 1 © 2011 Macmillan Publishers Limited. All rights reserved.

Supplementary Information 1. Additional data 1.1 – … · Supplementary Information 1. Additional data 1.1 – Formation of oxides and their identification using XPS The first major

  • Upload
    halien

  • View
    226

  • Download
    0

Embed Size (px)

Citation preview

Supplementary Information 1. Additional data 1.1 – Formation of oxides and their identification using XPS

The first major challenge of this study was the preparation and characterization of the oxide substrates. Oxide films were grown on metal substrates by means of controlled oxidation in-situ. The goal was to be able to grow oxides of various stoichiometries and metal oxidation states, determine their compositions using XPS, and draw correlations between their properties and energy-level alignment by eliminating extraneous factors related to ex-situ sample preparation. By analyzing the high-resolution XPS spectra of the main core level emission features of the transition metals we could determine the composition of each oxide. High-resolution XPS spectra of the main transition metal photoemission peaks for all the oxides tested in this work are shown in Fig. 1.1.1.

Figure 1.1.1 – XPS spectra of transition metals and transition metal oxides

Oxide identification is usually accomplished by measuring the chemical shift of high resolution core-level photoemission peaks of the oxidized transition metal. In the majority of cases the core-level binding energy is positively correlated with the oxidation state of the element being measured. By comparing the binding energy of the oxidized metal core level with that of the non-oxidized metallic standard, and comparing with several consistent literature reports of the same, one can identify the oxide present in the sample.

The accepted literature values are generally taken from single crystal or polycrystalline samples whose compositions have been determined by x-ray diffraction. Very often however there is a significant spread in reported values of binding energies, which can be a result of differing sample preparation procedures, sample conditions, and calibration procedures. Sometimes the spread in reported values for one oxide is larger than the difference in binding

1

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

NATURE MATERIALS | www.nature.com/naturematerials 1

© 2011 Macmillan Publishers Limited. All rights reserved.

energies between two different oxidation states. However this can generally be reconciled by retrieving the original literature reports and closely analyzing their spectra, calibration procedures and sample conditions.

In the cases where the binding energy values for different oxidation states are very similar, qualitative spectral features can sometimes be used to assist in identification of the oxide. For example, a number of oxides such as CuO, NiO, CoO and Cr2O3 exhibit satellite features caused by electron-correlation and relaxation effects. Satellite features may be present for one oxidation state of an element and absent for another, in which case it is easy to identify the oxide. Sometimes satellite features may be present for more than one oxidation state; however, the satellite features for each material have unique shapes, and binding energies which can allow one to distinguish oxides.

The O 1s photoemission peaks are not nearly as useful in identifying oxides as the main metal photoemission peaks are. The O 1s binding energies for all transition metal oxides are generally between 529.5 and 530.5 eV. Given this small range of binding energies oxides cannot usually be identified from the O 1s spectra. The reason for the small range of O 1s binding energies is that oxygen is invariably in the -2 oxidation state in all oxides.

A summary of the oxides observed in this study is shown in table 1, along with the binding energy of the main XPS metal peak, the oxidation state of the metal cation, and the valence electron configuration of the metal cation.

The compositions of the highly oxidized oxides are fairly straight forward and well-defined. Under the oxidizing conditions used, most metals oxidized into their thermodynamically most stable forms. As for the reduced oxides, a few cases form single oxidation state oxides (Cu, Co), other oxides contain multiple oxidation states (Mo, Ti), or cannot be reduced to a lower oxide at all and become reduced to metal instead (Ni). Table 1 – Summary of Binding energies, transition metal oxidation states and valence electron configurations for select transition metals and their oxides.

Binding Energy (eV)

Oxidation State

Valence Configuration

Reference

Ti (2p3/2) 454.1 0 d4 1 TiO2 458.8 +4 d0 1

TiO1+x 455.0 457.6

+2 +3

d2 d1 2

V (2p3/2) 512.2 0 d5 1 V2O5 517.4 +5 d0 1

Cr (2p3/2) 574.4 0 d6 1 Cr2O3 577.4 +3 d3 3

CrO3-x 578.3 576.3

+6 +4

d0 d2 4

Co (2p3/2) 778.3 0 d9 1 Co3O4 779.4 +2, +3 d7, d6 5 CoO 780.4 +2 d7 5

2

Ni (2p3/2) 852.7 0 d10 1 NiO 853.9 +2 d8 1

Cu (2p3/2) 932.7 0 d10s1 1 Cu2O 932.5 +1 d10 6 CuO 933.3 +2 d9 6

Mo (3d5/2) 228.0 0 d6 1 MoO3 232.4 +6 d0 7

MoO2.5 229.3 231.0

+4 +5

d2 d1 8

Ta (4f7/2) 21.9 0 d5 1 Ta2O5 26.6 +5 d0 9

W (4f7/2) 31.4 0 d6 1 WO3 35.6 +6 d0 10

Ag (3d5/2) 368.3 0 d10s1 1 Ag2O 368.0 +6 d10 1

1.2 – Substrate morphology The metal substrates were prepared by radio-frequency magnetron sputtering of 99.95 - 99.99 % pure metal targets, to produce 200 nm thick metal films on polished silicon wafers. The wafers were positioned 50 cm away from the deposition sources, and were constantly rotated to ensure uniform film coverage. The substrates were kept at room temperature during deposition. This procedure gives rise to metal films composed of grains that are between 20 – 50 nm in diameter, with an average surface roughness of 0.6 nm, as illustrated in the AFM images of Cu in Fig. 1.2.1. Note that when the AFM profile is plotted with a 1:1 scale (x:y) it is clear that the surface is very smooth, with an average vertical-to-horizontal feature ratio of ~ 60:1.

Figure 1.2.1 – (a) AFM height image of a sputter-deposited Cu film. (b) Cross section from (a) using an x:y scale of 1:56. (c) Cross section from (a) using an x:y scale of 1:1.

After oxidation, the gross features of the substrates remain the same, preserving the underlying metal grain structure. Oxide films were always grown to be between 2 – 10 nm thick, as determined using angle-resolved XPS measurements. Transmission electron microscopy

3

2 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

energies between two different oxidation states. However this can generally be reconciled by retrieving the original literature reports and closely analyzing their spectra, calibration procedures and sample conditions.

In the cases where the binding energy values for different oxidation states are very similar, qualitative spectral features can sometimes be used to assist in identification of the oxide. For example, a number of oxides such as CuO, NiO, CoO and Cr2O3 exhibit satellite features caused by electron-correlation and relaxation effects. Satellite features may be present for one oxidation state of an element and absent for another, in which case it is easy to identify the oxide. Sometimes satellite features may be present for more than one oxidation state; however, the satellite features for each material have unique shapes, and binding energies which can allow one to distinguish oxides.

The O 1s photoemission peaks are not nearly as useful in identifying oxides as the main metal photoemission peaks are. The O 1s binding energies for all transition metal oxides are generally between 529.5 and 530.5 eV. Given this small range of binding energies oxides cannot usually be identified from the O 1s spectra. The reason for the small range of O 1s binding energies is that oxygen is invariably in the -2 oxidation state in all oxides.

A summary of the oxides observed in this study is shown in table 1, along with the binding energy of the main XPS metal peak, the oxidation state of the metal cation, and the valence electron configuration of the metal cation.

The compositions of the highly oxidized oxides are fairly straight forward and well-defined. Under the oxidizing conditions used, most metals oxidized into their thermodynamically most stable forms. As for the reduced oxides, a few cases form single oxidation state oxides (Cu, Co), other oxides contain multiple oxidation states (Mo, Ti), or cannot be reduced to a lower oxide at all and become reduced to metal instead (Ni). Table 1 – Summary of Binding energies, transition metal oxidation states and valence electron configurations for select transition metals and their oxides.

Binding Energy (eV)

Oxidation State

Valence Configuration

Reference

Ti (2p3/2) 454.1 0 d4 1 TiO2 458.8 +4 d0 1

TiO1+x 455.0 457.6

+2 +3

d2 d1 2

V (2p3/2) 512.2 0 d5 1 V2O5 517.4 +5 d0 1

Cr (2p3/2) 574.4 0 d6 1 Cr2O3 577.4 +3 d3 3

CrO3-x 578.3 576.3

+6 +4

d0 d2 4

Co (2p3/2) 778.3 0 d9 1 Co3O4 779.4 +2, +3 d7, d6 5 CoO 780.4 +2 d7 5

2

Ni (2p3/2) 852.7 0 d10 1 NiO 853.9 +2 d8 1

Cu (2p3/2) 932.7 0 d10s1 1 Cu2O 932.5 +1 d10 6 CuO 933.3 +2 d9 6

Mo (3d5/2) 228.0 0 d6 1 MoO3 232.4 +6 d0 7

MoO2.5 229.3 231.0

+4 +5

d2 d1 8

Ta (4f7/2) 21.9 0 d5 1 Ta2O5 26.6 +5 d0 9

W (4f7/2) 31.4 0 d6 1 WO3 35.6 +6 d0 10

Ag (3d5/2) 368.3 0 d10s1 1 Ag2O 368.0 +6 d10 1

1.2 – Substrate morphology The metal substrates were prepared by radio-frequency magnetron sputtering of 99.95 - 99.99 % pure metal targets, to produce 200 nm thick metal films on polished silicon wafers. The wafers were positioned 50 cm away from the deposition sources, and were constantly rotated to ensure uniform film coverage. The substrates were kept at room temperature during deposition. This procedure gives rise to metal films composed of grains that are between 20 – 50 nm in diameter, with an average surface roughness of 0.6 nm, as illustrated in the AFM images of Cu in Fig. 1.2.1. Note that when the AFM profile is plotted with a 1:1 scale (x:y) it is clear that the surface is very smooth, with an average vertical-to-horizontal feature ratio of ~ 60:1.

Figure 1.2.1 – (a) AFM height image of a sputter-deposited Cu film. (b) Cross section from (a) using an x:y scale of 1:56. (c) Cross section from (a) using an x:y scale of 1:1.

After oxidation, the gross features of the substrates remain the same, preserving the underlying metal grain structure. Oxide films were always grown to be between 2 – 10 nm thick, as determined using angle-resolved XPS measurements. Transmission electron microscopy

3

NATURE MATERIALS | www.nature.com/naturematerials 3

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

profiles of thin cross sectional samples would be necessary to accurately determine the morphology of such samples. While determining the film morphologies using this method is a rather large-scale and extremely expensive undertaking, and furthermore, outside the main scope of this work, we suffice to draw some logical conclusions regarding our oxide morphologies, using combination of AFM, XPS and previous TEM work on oxide thin films. Previous studies of films grown in a similar manner have been shown oxides to grow as nano-sized grains (~ 3 nm) as shown in Fig. 1.2.2 (a),11-13 and in some cases amorphous, as shown in Fig. 1.2.2 (b).14-15 Thus, based on the knowledge of roughness and metal grain size as determined from AFM measurements, and from oxide film thickness determination from XPS, we propose our samples generally had morphologies as illustrated in Fig. 1.2.2 (c).

Figure 1.2.2 – (a) TEM image of 5 nm thick oxide film on Fe nanoparticle (from reference 13). (b) TEM image of a 5 nm thick oxide film on iron foil (from reference 15). (c) Proposed model of the typical substrate surface structure for the oxide films used in this study.

While there is essentially no preferred oxide orientation, we can assume that in all cases the most thermodynamically stable oxide face was the exposed face (with the inclusion of kinks and step edges). Therefore, the measured energy-level alignment in this study represents an average molecule-oxide interaction over all available oxide faces and molecular orientations. While admittedly, this approach washes over some details regarding molecule-substrate geometries, it is done so out of necessity of addressing a large number of oxides. However, by taking this approach we can provide a broader perspective than one could obtain by examining an infinite number of small details. 1.3 – Band gaps of oxides The figure in the manuscript showing the valence bands and conduction bands of the oxides (Figure 4) was constructed using valence band maxima and work functions determined from ultraviolet photoemission measurements, and band gaps from the literature, taken from combined photoemission and inverse photoemission measurements. The values used to construct Figure 4 are shown in table 2.

4

Table 2 – Summary of oxide ionization energies, work functions, and band gaps used to construct Fig. 4 in the main manuscript. The references given in the final column are for the oxide band gaps, which were determined using combined UPS and IPES.

Ionization Energy (eV)

Work Function (eV)

Band Gap (eV)

Reference

TiO2 8.0 5.2 3.4 16

V2O5 9.26 6.85 2.8 17

Cr2O3 6.5 5.6 3.2 3 CrO3 9.0 6.9 2.25 18

Co3O4 6.4 6.1 1.6 19 CoO 5.2 4.5 2.4 19

NiO 6.7 6.3 3.2 3

Cu2O 5.5 4.9 2.3 3 CuO 5.7 5.6 1.4 3

MoO3 9.4 6.9 3.0 20 MoO2 6.0 6.0 0.0 21

Ta2O5 8.9 5.2 3.9 22

WO3 9.8 6.65 3.4 23

Ag2O 6.4 5.51 1.3 24

α-NPD 5.35 n/a 3.95 20

CBP 6.00 n/a 4.0 20 1.4 - Work functions of metal oxides The work functions of transition metal oxides can be difficult to reproduce because there are so many factors that alter an oxide’s work function. For example, oxygen vacancy defects decrease an oxide’s work function because they act as n-type dopants, shifting the Fermi level closer to the conduction band and closer to the vacuum level than the stoichiometric oxide. For example, in Fig. 1.4.1 we show the effects introducing oxygen vacancies into a TiO2 sample by annealing in vacuum. The vacancies leave occupied states within the band gap of the oxide, as can be seen in the valence spectra. These states cause the Fermi level to be shifted upwards, thus reducing the work function.

Figure 1.4.1 – He I UPS spectra of a TiO2 thin film sample during annealing in vacuum at 200°C. (a) Secondary electron cut-off, (b) valence band (c) expanded valence to enhance the defect feature. As the sample is annealed, the work function

5

4 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

profiles of thin cross sectional samples would be necessary to accurately determine the morphology of such samples. While determining the film morphologies using this method is a rather large-scale and extremely expensive undertaking, and furthermore, outside the main scope of this work, we suffice to draw some logical conclusions regarding our oxide morphologies, using combination of AFM, XPS and previous TEM work on oxide thin films. Previous studies of films grown in a similar manner have been shown oxides to grow as nano-sized grains (~ 3 nm) as shown in Fig. 1.2.2 (a),11-13 and in some cases amorphous, as shown in Fig. 1.2.2 (b).14-15 Thus, based on the knowledge of roughness and metal grain size as determined from AFM measurements, and from oxide film thickness determination from XPS, we propose our samples generally had morphologies as illustrated in Fig. 1.2.2 (c).

Figure 1.2.2 – (a) TEM image of 5 nm thick oxide film on Fe nanoparticle (from reference 13). (b) TEM image of a 5 nm thick oxide film on iron foil (from reference 15). (c) Proposed model of the typical substrate surface structure for the oxide films used in this study.

While there is essentially no preferred oxide orientation, we can assume that in all cases the most thermodynamically stable oxide face was the exposed face (with the inclusion of kinks and step edges). Therefore, the measured energy-level alignment in this study represents an average molecule-oxide interaction over all available oxide faces and molecular orientations. While admittedly, this approach washes over some details regarding molecule-substrate geometries, it is done so out of necessity of addressing a large number of oxides. However, by taking this approach we can provide a broader perspective than one could obtain by examining an infinite number of small details. 1.3 – Band gaps of oxides The figure in the manuscript showing the valence bands and conduction bands of the oxides (Figure 4) was constructed using valence band maxima and work functions determined from ultraviolet photoemission measurements, and band gaps from the literature, taken from combined photoemission and inverse photoemission measurements. The values used to construct Figure 4 are shown in table 2.

4

Table 2 – Summary of oxide ionization energies, work functions, and band gaps used to construct Fig. 4 in the main manuscript. The references given in the final column are for the oxide band gaps, which were determined using combined UPS and IPES.

Ionization Energy (eV)

Work Function (eV)

Band Gap (eV)

Reference

TiO2 8.0 5.2 3.4 16

V2O5 9.26 6.85 2.8 17

Cr2O3 6.5 5.6 3.2 3 CrO3 9.0 6.9 2.25 18

Co3O4 6.4 6.1 1.6 19 CoO 5.2 4.5 2.4 19

NiO 6.7 6.3 3.2 3

Cu2O 5.5 4.9 2.3 3 CuO 5.7 5.6 1.4 3

MoO3 9.4 6.9 3.0 20 MoO2 6.0 6.0 0.0 21

Ta2O5 8.9 5.2 3.9 22

WO3 9.8 6.65 3.4 23

Ag2O 6.4 5.51 1.3 24

α-NPD 5.35 n/a 3.95 20

CBP 6.00 n/a 4.0 20 1.4 - Work functions of metal oxides The work functions of transition metal oxides can be difficult to reproduce because there are so many factors that alter an oxide’s work function. For example, oxygen vacancy defects decrease an oxide’s work function because they act as n-type dopants, shifting the Fermi level closer to the conduction band and closer to the vacuum level than the stoichiometric oxide. For example, in Fig. 1.4.1 we show the effects introducing oxygen vacancies into a TiO2 sample by annealing in vacuum. The vacancies leave occupied states within the band gap of the oxide, as can be seen in the valence spectra. These states cause the Fermi level to be shifted upwards, thus reducing the work function.

Figure 1.4.1 – He I UPS spectra of a TiO2 thin film sample during annealing in vacuum at 200°C. (a) Secondary electron cut-off, (b) valence band (c) expanded valence to enhance the defect feature. As the sample is annealed, the work function

5

NATURE MATERIALS | www.nature.com/naturematerials 5

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

decreases, the valence band shifts to higher binding energy, and the defect feature increases in intensity. (d) Plot of work function versus defect feature area. (e) Energy-level diagram illustrating how defect states lower the work function.

Certain oxides are more prone to forming oxygen vacancies than others. For example, the oxygen vacancy formation energy is lower in MoO3 than TiO2, making it more difficult to obtain a stoichiometric MoO3 sample. The formation of high-work function, stoichiometric MoO3 samples requires highly oxidizing conditions, such as ozone oxidation with heating. However, the sample must be cooled before the oxygen is pumped away because heating in vacuum will result in oxygen vacancy formation. An oxide’s cation oxidation state also affects the oxide’s work function. Lower oxidation states generally cause a reduction in the oxide’s work function. If the oxide is not completely oxidized, there may be a mixture of high and low oxidation states. This is particularly important for oxides that are difficult to obtain in their highest oxidation state, like CrO3. If the conditions used to form the oxide do not produce only one oxidation state, a mixture of oxidation states is obtained and ratio of oxidation states will be difficult to reproduce, thus making the work function difficult to reproduce. In addition to the intrinsic properties of oxides, their work functions can also be altered by adsorption of gas-phase molecules. No dangling bonds are present on these oxide surfaces, as they were formed under excess oxygen conditions. Thus the sticking coefficient of most adsorbates is very low, and under a pressure of 10-10 torr adsorbate coverage occurs over the course of tens of hours. However, if leaving a sample in vacuum for several hours, changes in work function on the order of 0.5 eV can be observed. The time budget for work function measurements differs from one oxide to the next, as each oxide’s sticking coefficients for the residual gases are different. For example, NiO is found to adsorb carbonaceous species like CO, which can be generated from an ion pump, much faster than MoO3. Thus NiO will show a detectible decrease in work function over the course of 0.5 hrs while sitting in vacuum. In general, we have minimized the time our samples sat in vacuum to ensure no drop in work function from adsorbates. After oxide growth was complete, each sample is quickly measured then promptly has an organic film deposited onto it, such that each measurement is accomplished in the shortest time possible. 1.5 - Valence levels of organic semiconductors Ultraviolet photoemission spectroscopy (UPS) was used to determine the energetic difference between the HOMO level and the Fermi level, as well as the organic film’s work function. Energy-level alignment was measured by employing the layer-by-layer deposition method. This method involves sequentially depositing ultra-thin layers of organic molecules onto the substrates, with UPS measurements taken after each layer is deposited, as illustrated in Fig. 1.5.1. This is method is necessary for mapping the potential profile of an interface due to the extremely small probing depth of UPS (~ 1nm).

6

Figure 1.5.1 – Illustration of the layer-by-layer photoemission technique of interface characterization.

The layer-by-layer measurement routine generates a stack of valence and secondary cut-off spectra. Example spectra of layer-by-layer depositions for CBP, α-NPD, and 2T-NATA are shown in Fig. 1.5.2. The purpose for measuring HOMO binding energy and work function of the organic film in several steps, as opposed to using a single-point measurement is so a thickness profile can be created. The HOMO binding energy and work function of an organic film depends on the film’s thickness. We performed complete thickness profiles for each interface because we wished to determine whether there were any differences in the HOMO or work function profiles between the oxides. However, it was found that almost all profiles were identical in shape, as shown in the HOMO binding energy (ΔEH) and work function (Φ) profiles for CBP, α-NPD and 2T-NATA in Fig 1.5.3. The profiles only differ from one another in terms of a vertical off-set.

Figure 1.5.2 – Examples of stacks of UPS spectra from layer-by-layer measurements of CBP (top) α-NPD (middle) and 2T-NATA (bottom) deposited onto various substrates. The substrates used in these examples are V2O5 for the CBP spectra, NiO for the α-NPD spectra, and CuO for the 2T-NATA spectra.

7

6 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

decreases, the valence band shifts to higher binding energy, and the defect feature increases in intensity. (d) Plot of work function versus defect feature area. (e) Energy-level diagram illustrating how defect states lower the work function.

Certain oxides are more prone to forming oxygen vacancies than others. For example, the oxygen vacancy formation energy is lower in MoO3 than TiO2, making it more difficult to obtain a stoichiometric MoO3 sample. The formation of high-work function, stoichiometric MoO3 samples requires highly oxidizing conditions, such as ozone oxidation with heating. However, the sample must be cooled before the oxygen is pumped away because heating in vacuum will result in oxygen vacancy formation. An oxide’s cation oxidation state also affects the oxide’s work function. Lower oxidation states generally cause a reduction in the oxide’s work function. If the oxide is not completely oxidized, there may be a mixture of high and low oxidation states. This is particularly important for oxides that are difficult to obtain in their highest oxidation state, like CrO3. If the conditions used to form the oxide do not produce only one oxidation state, a mixture of oxidation states is obtained and ratio of oxidation states will be difficult to reproduce, thus making the work function difficult to reproduce. In addition to the intrinsic properties of oxides, their work functions can also be altered by adsorption of gas-phase molecules. No dangling bonds are present on these oxide surfaces, as they were formed under excess oxygen conditions. Thus the sticking coefficient of most adsorbates is very low, and under a pressure of 10-10 torr adsorbate coverage occurs over the course of tens of hours. However, if leaving a sample in vacuum for several hours, changes in work function on the order of 0.5 eV can be observed. The time budget for work function measurements differs from one oxide to the next, as each oxide’s sticking coefficients for the residual gases are different. For example, NiO is found to adsorb carbonaceous species like CO, which can be generated from an ion pump, much faster than MoO3. Thus NiO will show a detectible decrease in work function over the course of 0.5 hrs while sitting in vacuum. In general, we have minimized the time our samples sat in vacuum to ensure no drop in work function from adsorbates. After oxide growth was complete, each sample is quickly measured then promptly has an organic film deposited onto it, such that each measurement is accomplished in the shortest time possible. 1.5 - Valence levels of organic semiconductors Ultraviolet photoemission spectroscopy (UPS) was used to determine the energetic difference between the HOMO level and the Fermi level, as well as the organic film’s work function. Energy-level alignment was measured by employing the layer-by-layer deposition method. This method involves sequentially depositing ultra-thin layers of organic molecules onto the substrates, with UPS measurements taken after each layer is deposited, as illustrated in Fig. 1.5.1. This is method is necessary for mapping the potential profile of an interface due to the extremely small probing depth of UPS (~ 1nm).

6

Figure 1.5.1 – Illustration of the layer-by-layer photoemission technique of interface characterization.

The layer-by-layer measurement routine generates a stack of valence and secondary cut-off spectra. Example spectra of layer-by-layer depositions for CBP, α-NPD, and 2T-NATA are shown in Fig. 1.5.2. The purpose for measuring HOMO binding energy and work function of the organic film in several steps, as opposed to using a single-point measurement is so a thickness profile can be created. The HOMO binding energy and work function of an organic film depends on the film’s thickness. We performed complete thickness profiles for each interface because we wished to determine whether there were any differences in the HOMO or work function profiles between the oxides. However, it was found that almost all profiles were identical in shape, as shown in the HOMO binding energy (ΔEH) and work function (Φ) profiles for CBP, α-NPD and 2T-NATA in Fig 1.5.3. The profiles only differ from one another in terms of a vertical off-set.

Figure 1.5.2 – Examples of stacks of UPS spectra from layer-by-layer measurements of CBP (top) α-NPD (middle) and 2T-NATA (bottom) deposited onto various substrates. The substrates used in these examples are V2O5 for the CBP spectra, NiO for the α-NPD spectra, and CuO for the 2T-NATA spectra.

7

NATURE MATERIALS | www.nature.com/naturematerials 7

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

It is interesting to note that, while the HOMO binding energy of a gas phase molecule is a material constant, when molecules coalesce to form a solid film their binding energies depend on the substrate to which they are adsorbed. This is exemplified by vertical off-set in the ΔEH and Φ profiles of Fig. 1.5.3 (a) and (b).

Figure 3 – Work function and HOMO profiles for all substrates and all organics used in this study. These profiles were extracted from stacks of UPS spectra, such as the ones demonstrated in Fig. 1.5.2. The legend at the bottom indicated the substrates used for each profile.

The ΔEH and Φ profiles also exhibit similar shapes regardless of the organic molecule used. There is always a rapid initial change in ΔEH and Φ, followed by a levelling off with thickness. The HOMO is generally closest to the Fermi level near the substrate/organic interface, and slowly moves to higher binding energy as the organic film thickens.

Although ΔEH and Φ are not material constants, the sum of them, that is, the ionization energy (IE) is constant. Figure 1.5.4 shows the IE profiles of CBP, α-NPD and 2T-NATA on all the various substrates.

The reason why HOMO binding energy is substrate dependant, while ionization energy is a constant is because the HOMO binding energy is measured with reference to the substrate Fermi level, while ionization energy is measured with reference to the vacuum level. The

8

difference between the HOMO and the Fermi level is not a constant, and depends on charge transfer between the substrate and adsorbate. However, the difference between the HOMO and the vacuum level represents a property of the molecule, and is independent of the substrate.

The trend of ionization energy versus organic film thickness generally shows a reduction in ionization energy for extremely thin organic films. This is believed to be caused by distortion of the molecular orbital electron density due to polarization by the substrates.25-26 As the films thicken, the ionization energies quickly establish their constant bulk values. The bulk IEs for the three organics, as shown in Fig. 1.5.4, are 6.00, 5.35 and 4.98 eV for CBP, α-NPD and 2-TNATA, respectively.

Figure 1.5.4 – Ionization energy versus organic film thickness profiles for each organic on each substrate. These plots were constructed from the profiles shown in Fig. 1.5.3 by summing work function with HOMO binding energy.

1.6 – Growth mode of organic films The three organics examined in this study do not crystallize easily. This is a result of their rather flexible molecular structures and inability pack closely together. They tend to form amorphous films, and as a result, they do not grow as islands. Knowledge of this property is important for interpretation of layer-by-layer profiles. If films form via an island growth mode then the thickness scale in the layer-by-layer profiles will be inaccurate. In all cases shown in this work the organic films grew in a layer-by-layer mode. This can be easily determined from the ultraviolet photoemission spectra. When a film exhibits island growth, the substrate photoemission signal persists in the spectrum even after several nanometers of overlayer have been deposited. When the film exhibits uniform coverage the substrate photoemission signal

9

8 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

It is interesting to note that, while the HOMO binding energy of a gas phase molecule is a material constant, when molecules coalesce to form a solid film their binding energies depend on the substrate to which they are adsorbed. This is exemplified by vertical off-set in the ΔEH and Φ profiles of Fig. 1.5.3 (a) and (b).

Figure 3 – Work function and HOMO profiles for all substrates and all organics used in this study. These profiles were extracted from stacks of UPS spectra, such as the ones demonstrated in Fig. 1.5.2. The legend at the bottom indicated the substrates used for each profile.

The ΔEH and Φ profiles also exhibit similar shapes regardless of the organic molecule used. There is always a rapid initial change in ΔEH and Φ, followed by a levelling off with thickness. The HOMO is generally closest to the Fermi level near the substrate/organic interface, and slowly moves to higher binding energy as the organic film thickens.

Although ΔEH and Φ are not material constants, the sum of them, that is, the ionization energy (IE) is constant. Figure 1.5.4 shows the IE profiles of CBP, α-NPD and 2T-NATA on all the various substrates.

The reason why HOMO binding energy is substrate dependant, while ionization energy is a constant is because the HOMO binding energy is measured with reference to the substrate Fermi level, while ionization energy is measured with reference to the vacuum level. The

8

difference between the HOMO and the Fermi level is not a constant, and depends on charge transfer between the substrate and adsorbate. However, the difference between the HOMO and the vacuum level represents a property of the molecule, and is independent of the substrate.

The trend of ionization energy versus organic film thickness generally shows a reduction in ionization energy for extremely thin organic films. This is believed to be caused by distortion of the molecular orbital electron density due to polarization by the substrates.25-26 As the films thicken, the ionization energies quickly establish their constant bulk values. The bulk IEs for the three organics, as shown in Fig. 1.5.4, are 6.00, 5.35 and 4.98 eV for CBP, α-NPD and 2-TNATA, respectively.

Figure 1.5.4 – Ionization energy versus organic film thickness profiles for each organic on each substrate. These plots were constructed from the profiles shown in Fig. 1.5.3 by summing work function with HOMO binding energy.

1.6 – Growth mode of organic films The three organics examined in this study do not crystallize easily. This is a result of their rather flexible molecular structures and inability pack closely together. They tend to form amorphous films, and as a result, they do not grow as islands. Knowledge of this property is important for interpretation of layer-by-layer profiles. If films form via an island growth mode then the thickness scale in the layer-by-layer profiles will be inaccurate. In all cases shown in this work the organic films grew in a layer-by-layer mode. This can be easily determined from the ultraviolet photoemission spectra. When a film exhibits island growth, the substrate photoemission signal persists in the spectrum even after several nanometers of overlayer have been deposited. When the film exhibits uniform coverage the substrate photoemission signal

9

NATURE MATERIALS | www.nature.com/naturematerials 9

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

disappears after one or two monolayers have been deposited. The uniform growth mode of these films has also been confirmed using AFM measurements. 1.7 - Interfacial chemical reactions There are several reports in the literature of certain substrates forming chemical bonds with organic overlayers.27-31 These interfacial reactions lead to a localized chemical bond, and thus a new chemical state, which can be clearly seen in valence and core-level photoemission spectra. None of the substrates investigated in this study had such a strong chemical interaction with any of the organic overlayers. None of the interfaces exhibited any additional chemical states. Overlayer and substrate spectra at the interface were identical to spectra of pristine samples except for some spectral broadening of organic films close to the interface. This broadening is likely a result of a physisorption-type of interaction. Thus we conclude that all interfaces seen in this study were weakly interaction or physisorptive interactions. The oxides were not expected to bond strongly to the organics, as they were all oxygen terminated, and not expected to possess any dangling bonds. The clean metal surfaces are more reactive, but were not found to form chemical bonds with any of the organics. This was likely due to the fact that all metals were transition metals, and had valence bands composed of narrow d-bands, which do not interact strongly with the organic molecular orbitals. 1.8 – Oxide film thickness and energy-level alignment Oxide film thickness affected energy-level alignment only to the extent that oxide work function depended on oxide thickness, and if oxides were too thick to allow electrical conduction, in which cases measurements could not be made using photoemission spectroscopy. In general, the work functions of very thin oxide films are lower than the bulk films, and work function increases as the films thicken, as shown in Fig. 1.8.1 (a). This happens because thin films have many defects at the substrate-oxide interface. For example, for MoO3 grown on Mo metal, cations in lower oxidation states exist close to the Mo/MoO3 interface, as shown in Fig. 1.8.1 (b) and (c). These lower oxidation state cations contribute occupied valence states within the oxide band gap and lower the oxide work function, as shown in Fig. 1.8.1 (d) and (e).

Figure 1.8.1 – (a) plot of work function versus oxide film thickness for a MoO3 film grown on a Mo substrate. (b) Mo3d5/2 XPS spectra of bulk MoO3, clean Mo metal, a 0.6 nm thick MoO3 film and the difference spectrum made by subtracting the Mo metal spectrum from the 0.6 nm spectrum. (c) Peak fit of the 0.6 nm MoO3 spectrum, showing the Mo6+ and Mo4+ components. (d) Valence UPS spectrum of MoO3 films of increasing thickness. (e) Expanded view of the defect features in the UPS spectrum, revealing two new states at the Mo/MoO3 interface.

10

2. Additional experimental details 2.1 Work Function Measurements Work function was measured using the secondary-electron cut-off of photoemission spectra. Secondary electrons are generated when photoelectrons, travelling within the sample, become inelastically scattered on their way to the surface. Since the amount of energy loss per scattering event is arbitrary, inelastically scattered electrons give a smooth, almost featureless background signal. This signal increases exponentially towards the high binding energy (low-kinetic energy) side of the spectrum, as shown in the UPS spectrum of gold in Fig. 2.1.1. As one moves to lower kinetic energy in the spectrum, there is a cut-off in the in the scattered electron intensity. This cut-off represents the point at which electrons escape the sample with zero kinetic energy.

Figure 2.1.1 – UPS spectrum of sputter-cleaned Au foil, illustrating the secondary electron background, secondary electron cut-off, the valence density of states and the Fermi edge.

In Fig. 2.1.2 the ultraviolet photoemission spectrum is re-plotted using electron kinetic energy for the x-axis. One can see that the electrons from the Fermi level are escaping the sample with ~ 15.8 eV of kinetic energy. Since the photon energy is 21.22 eV, the remaining energy represents the sample’s work function of 5.42 eV.

Figure 2.1.2 – UPS spectrum of sputter-cleaned Au foil, with the x-axis plotted in terms of electron kinetic energy to illustrate how work function is determined from a UPS spectrum. Note, the zero in kinetic energy appears at the secondary electron cut-off.

When measuring work function, the sample must be placed under a negative applied bias relative to the spectrometer so that the very slow electrons are capable of being detected. Without

11

10 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

disappears after one or two monolayers have been deposited. The uniform growth mode of these films has also been confirmed using AFM measurements. 1.7 - Interfacial chemical reactions There are several reports in the literature of certain substrates forming chemical bonds with organic overlayers.27-31 These interfacial reactions lead to a localized chemical bond, and thus a new chemical state, which can be clearly seen in valence and core-level photoemission spectra. None of the substrates investigated in this study had such a strong chemical interaction with any of the organic overlayers. None of the interfaces exhibited any additional chemical states. Overlayer and substrate spectra at the interface were identical to spectra of pristine samples except for some spectral broadening of organic films close to the interface. This broadening is likely a result of a physisorption-type of interaction. Thus we conclude that all interfaces seen in this study were weakly interaction or physisorptive interactions. The oxides were not expected to bond strongly to the organics, as they were all oxygen terminated, and not expected to possess any dangling bonds. The clean metal surfaces are more reactive, but were not found to form chemical bonds with any of the organics. This was likely due to the fact that all metals were transition metals, and had valence bands composed of narrow d-bands, which do not interact strongly with the organic molecular orbitals. 1.8 – Oxide film thickness and energy-level alignment Oxide film thickness affected energy-level alignment only to the extent that oxide work function depended on oxide thickness, and if oxides were too thick to allow electrical conduction, in which cases measurements could not be made using photoemission spectroscopy. In general, the work functions of very thin oxide films are lower than the bulk films, and work function increases as the films thicken, as shown in Fig. 1.8.1 (a). This happens because thin films have many defects at the substrate-oxide interface. For example, for MoO3 grown on Mo metal, cations in lower oxidation states exist close to the Mo/MoO3 interface, as shown in Fig. 1.8.1 (b) and (c). These lower oxidation state cations contribute occupied valence states within the oxide band gap and lower the oxide work function, as shown in Fig. 1.8.1 (d) and (e).

Figure 1.8.1 – (a) plot of work function versus oxide film thickness for a MoO3 film grown on a Mo substrate. (b) Mo3d5/2 XPS spectra of bulk MoO3, clean Mo metal, a 0.6 nm thick MoO3 film and the difference spectrum made by subtracting the Mo metal spectrum from the 0.6 nm spectrum. (c) Peak fit of the 0.6 nm MoO3 spectrum, showing the Mo6+ and Mo4+ components. (d) Valence UPS spectrum of MoO3 films of increasing thickness. (e) Expanded view of the defect features in the UPS spectrum, revealing two new states at the Mo/MoO3 interface.

10

2. Additional experimental details 2.1 Work Function Measurements Work function was measured using the secondary-electron cut-off of photoemission spectra. Secondary electrons are generated when photoelectrons, travelling within the sample, become inelastically scattered on their way to the surface. Since the amount of energy loss per scattering event is arbitrary, inelastically scattered electrons give a smooth, almost featureless background signal. This signal increases exponentially towards the high binding energy (low-kinetic energy) side of the spectrum, as shown in the UPS spectrum of gold in Fig. 2.1.1. As one moves to lower kinetic energy in the spectrum, there is a cut-off in the in the scattered electron intensity. This cut-off represents the point at which electrons escape the sample with zero kinetic energy.

Figure 2.1.1 – UPS spectrum of sputter-cleaned Au foil, illustrating the secondary electron background, secondary electron cut-off, the valence density of states and the Fermi edge.

In Fig. 2.1.2 the ultraviolet photoemission spectrum is re-plotted using electron kinetic energy for the x-axis. One can see that the electrons from the Fermi level are escaping the sample with ~ 15.8 eV of kinetic energy. Since the photon energy is 21.22 eV, the remaining energy represents the sample’s work function of 5.42 eV.

Figure 2.1.2 – UPS spectrum of sputter-cleaned Au foil, with the x-axis plotted in terms of electron kinetic energy to illustrate how work function is determined from a UPS spectrum. Note, the zero in kinetic energy appears at the secondary electron cut-off.

When measuring work function, the sample must be placed under a negative applied bias relative to the spectrometer so that the very slow electrons are capable of being detected. Without

11

NATURE MATERIALS | www.nature.com/naturematerials 11

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

an applied bias, slow electrons will not reach the detector because they do not have enough energy to overcome the contact potential difference between the sample and the spectrometer. Without the applied bias there will still be an apparent cut-off in the spectrum, but the cut-off will be un-related to the sample’s work function, and will only represent the spectrometer’s work function. 2.2 - Oxidation procedure

Oxidation was carried out in an isolated oxidation chamber by evacuating the chamber to ~ 10-9 torr, introducing the sample, then back-filling the chamber with ~ 760 torr of 99.99 % pure O2. Faster oxidation rates and high-oxidation-state oxides were achieved by heating the substrates during oxidation using a 300 W halogen light bulb, positioned 2 cm above the sample, and by generating ozone during oxidation by shining UV-light through a UV-transparent quartz window from an external UV-lamp. Oxidation times and temperatures varied depending on the oxidation rate of the metal state, and the desired oxidation state. Reduced and oxygen-deficient oxides were achieved by annealing oxides in vacuum. 2.3 – Validation of photoemission measurements As mentioned in section 1.5, organic interfaces are generally characterized using layer-by-layer photoemission spectroscopy. This measurement method requires samples to be repeatedly exposed to ionizing radiation. This radiation can cause minor damage to the molecules, such as chemical changes, and can result in subtle changes to the spectra, which are often insignificant. However, throughout an experiment, if care is not taken to minimize radiation exposure, the subtle changes from each measurement step can lead to significant artefacts in the observed trends in the binding energy and work function profiles. For this reason, it is important to validate one’s measurement procedures to ensure that the data are reliable. Here we will illustrate some of our validation procedures. We found that it was necessary to develop an operating procedure that minimizes radiation times. This required that we only used UPS (without XPS) to measure energy-level alignment. XPS could not be used for layer-by-layer characterizations because it has much lower signal intensity and requires a much longer acquisition time. Instead, we limited our layer-by-layer characterizations to UPS spectra, in which each spectrum required only about 10 seconds of collection time. Our validation procedures confirmed that this method causes insignificant changes to the sample work functions and binding energies. Fig. 2.3.1 shows layer-by-layer work function measurements of an α-NPD film deposited onto a CuO substrate, where XPS spectra of the C 1s, N 1s, O 1s and Cu 2p3/2 photoemission features were collected after each deposition step. A control sample was used, which did not have XPS spectra measured from it, and was only irradiated for the time necessary to take a work function measurement (~ 10 seconds per measurement). In Fig. 2.3.1 (a) it can be seen that the collection of spectra causes a systematic decrease in the organic film’s work function, while the non-irradiated sample’s work function remains relatively constant after about 1 nm of α-NPD is

12

deposited. Fig. 2.3.1 (b) shows the XPS spectra during the layer-by-layer deposition. The binding energies of the C 1s and N1s peaks shift throughout the measurement, while the O 1s and Cu 2p3/2 peaks remain constant. This indicates that irradiation is causing artefacts in the organic film, and one must be careful when determining potential profiles from photoemission measurements.

Figure 2.3.1 – (a) Plot of work function versus organic film thickness to demonstrate the effects of prolonged x-ray irradiation on organic film work function. The irradiated sample had C1s, N1s, O1s and Cu2p3/2 spectra collected after each organic film deposition step. The control sample was only irradiated for the time needed for work function measurements. (b) The XPS spectra of the ‘irradiated sample’ from (a). Note the peaks of the over-layer spectra (C1s and N1s) shift to higher binding energy, while the substrate spectra (O1s and Cu2p3/2) do not change throughout the deposition sequence.

As accurate peak determination from XPS measurements requires too much sample irradiation, we could not include XPS measurements to our layer-by-layer interface characterizations. However, XPS measurements are not absolutely necessary for an interface characterization if the valence levels can be reliably measured using UPS. Therefore, we needed to test whether ultraviolet radiation also causes similar artefacts, and whether UPS measurements can be taken without significant film degradation. Fig. 2.3.2 (a) shows the UPS spectra of an α-NPD film during continual ultraviolet irradiation. The off-set between each spectrum represents 5 minutes of irradiation time. The HOMO feature shifts to higher binding energy and broadens as the sample is irradiated. A plot of HOMO binding energy versus irradiation time is shown in Fig. 2.3.2 (b). The HOMO binding energy shifts by ~ 0.3 eV during 45 minutes of UV irradiation. This is a significant amount of error for an interface characterization. However, as only about 30 seconds is required for one valence band measurement, a layer-by-layer characterization can likely be performed with minimal irradiation artefacts if special care is taken. To confirm that indeed, using our operating procedure, radiation damage has insignificant impact on the measured binding energy, we compare several HOMO binding energy profiles in

13

12 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

an applied bias, slow electrons will not reach the detector because they do not have enough energy to overcome the contact potential difference between the sample and the spectrometer. Without the applied bias there will still be an apparent cut-off in the spectrum, but the cut-off will be un-related to the sample’s work function, and will only represent the spectrometer’s work function. 2.2 - Oxidation procedure

Oxidation was carried out in an isolated oxidation chamber by evacuating the chamber to ~ 10-9 torr, introducing the sample, then back-filling the chamber with ~ 760 torr of 99.99 % pure O2. Faster oxidation rates and high-oxidation-state oxides were achieved by heating the substrates during oxidation using a 300 W halogen light bulb, positioned 2 cm above the sample, and by generating ozone during oxidation by shining UV-light through a UV-transparent quartz window from an external UV-lamp. Oxidation times and temperatures varied depending on the oxidation rate of the metal state, and the desired oxidation state. Reduced and oxygen-deficient oxides were achieved by annealing oxides in vacuum. 2.3 – Validation of photoemission measurements As mentioned in section 1.5, organic interfaces are generally characterized using layer-by-layer photoemission spectroscopy. This measurement method requires samples to be repeatedly exposed to ionizing radiation. This radiation can cause minor damage to the molecules, such as chemical changes, and can result in subtle changes to the spectra, which are often insignificant. However, throughout an experiment, if care is not taken to minimize radiation exposure, the subtle changes from each measurement step can lead to significant artefacts in the observed trends in the binding energy and work function profiles. For this reason, it is important to validate one’s measurement procedures to ensure that the data are reliable. Here we will illustrate some of our validation procedures. We found that it was necessary to develop an operating procedure that minimizes radiation times. This required that we only used UPS (without XPS) to measure energy-level alignment. XPS could not be used for layer-by-layer characterizations because it has much lower signal intensity and requires a much longer acquisition time. Instead, we limited our layer-by-layer characterizations to UPS spectra, in which each spectrum required only about 10 seconds of collection time. Our validation procedures confirmed that this method causes insignificant changes to the sample work functions and binding energies. Fig. 2.3.1 shows layer-by-layer work function measurements of an α-NPD film deposited onto a CuO substrate, where XPS spectra of the C 1s, N 1s, O 1s and Cu 2p3/2 photoemission features were collected after each deposition step. A control sample was used, which did not have XPS spectra measured from it, and was only irradiated for the time necessary to take a work function measurement (~ 10 seconds per measurement). In Fig. 2.3.1 (a) it can be seen that the collection of spectra causes a systematic decrease in the organic film’s work function, while the non-irradiated sample’s work function remains relatively constant after about 1 nm of α-NPD is

12

deposited. Fig. 2.3.1 (b) shows the XPS spectra during the layer-by-layer deposition. The binding energies of the C 1s and N1s peaks shift throughout the measurement, while the O 1s and Cu 2p3/2 peaks remain constant. This indicates that irradiation is causing artefacts in the organic film, and one must be careful when determining potential profiles from photoemission measurements.

Figure 2.3.1 – (a) Plot of work function versus organic film thickness to demonstrate the effects of prolonged x-ray irradiation on organic film work function. The irradiated sample had C1s, N1s, O1s and Cu2p3/2 spectra collected after each organic film deposition step. The control sample was only irradiated for the time needed for work function measurements. (b) The XPS spectra of the ‘irradiated sample’ from (a). Note the peaks of the over-layer spectra (C1s and N1s) shift to higher binding energy, while the substrate spectra (O1s and Cu2p3/2) do not change throughout the deposition sequence.

As accurate peak determination from XPS measurements requires too much sample irradiation, we could not include XPS measurements to our layer-by-layer interface characterizations. However, XPS measurements are not absolutely necessary for an interface characterization if the valence levels can be reliably measured using UPS. Therefore, we needed to test whether ultraviolet radiation also causes similar artefacts, and whether UPS measurements can be taken without significant film degradation. Fig. 2.3.2 (a) shows the UPS spectra of an α-NPD film during continual ultraviolet irradiation. The off-set between each spectrum represents 5 minutes of irradiation time. The HOMO feature shifts to higher binding energy and broadens as the sample is irradiated. A plot of HOMO binding energy versus irradiation time is shown in Fig. 2.3.2 (b). The HOMO binding energy shifts by ~ 0.3 eV during 45 minutes of UV irradiation. This is a significant amount of error for an interface characterization. However, as only about 30 seconds is required for one valence band measurement, a layer-by-layer characterization can likely be performed with minimal irradiation artefacts if special care is taken. To confirm that indeed, using our operating procedure, radiation damage has insignificant impact on the measured binding energy, we compare several HOMO binding energy profiles in

13

NATURE MATERIALS | www.nature.com/naturematerials 13

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

Fig. 2.3.3. Here, the grey data points represent HOMO binding energies measured with our conventional layer-by-layer method. The black point represents the HOMO binding energy of a thick α-NPD film deposited in a single step and having no prior exposure to radiation. Here it can be seen that the even without prior radiation exposure, the black point falls along the general trend of the layer-by-layer data points. This plot validates that our measurement procedure is sufficient to minimize radiation artefacts and produce reliable energy-level profiles.

Figure 2.3.2 – (a) Valence UPS spectra collected during continuous UV irradiation of a 5 nm thick α-NPD film. Each spectrum differs by 5 minutes of irradiation time. (b) Plot of HOMO binding energy versus UV irradiation time.

Figure 4 – Plot of HOMO binding energy of α-NPD on CuO, as determined from numerous independent measurements. The Black cross indicates a measurement taken of a thick film with no prior irradiation. This plot shows that indeed there is a thickness dependence to HOMO binding energy (i.e. not all HOMO binding energy changes are caused by radiation damage). All measurements in this plot were taken using the minimum sample irradiation needed.

3 - Additional discussion 3.1 - Discussion of reference levels Reference levels are an important consideration in interpreting photoemission spectra, and often a point of confusion.32-36 In gas phase spectroscopy, binding energies are calibrated using optical methods. For example, the binding energy for a particular electron state can be determined by measuring the energy of the corresponding emission line in a Rydberg series. By doing so, the binding energy is calibrated with reference to the vacuum level (i.e. the energy of an electron at rest, infinite distance away from any other charged particles), such that the vacuum level is zero on the scale.

14

In solid state photoemission spectroscopy, the situation is more complicated. In general, when a sample contacts a spectrometer, the electron chemical potentials of the two equilibrate, resulting in Fermi level alignment. We will illustrate the concepts of reference levels in solid state spectroscopy using a step-wise example, as shown in Fi.g 3.1.1. Part (A) represents the spectrometer and sample in complete isolation from one another, and from any other source of charge. Here, the two materials both have their own distinct Fermi energies defined by the energies of their highest filled electron states. These Fermi energies can in principle (although not in practice) be defined the absolute energy scale. The zero of this scale is indicated by the horizontal dashed line in Fig. 3.1.1, labelled “vac. level”. In this case, and neglecting surface dipoles of the sample and spectrometer, we can say their vacuum levels are aligned. The next step to consider is grounding the spectrometer, as illustrated in (B). By grounding, we mean that the spectrometer is in electrochemical equilibrium with the electron reservoir known as “ground” or “earth”. In this case, the Fermi level of the spectrometer aligns with the Fermi level of ground, causing the ‘local’ vacuum level of the spectrometer to deviate from the absolute reference level. The energy of the Fermi level of ground on the absolute energy scale has been determined experimentally to be ~ 4.09 eV.32 In the next step, we consider what happens when a sample contacts the spectrometer, as shown in (C). When contact is made, electrons will move in order to establish equilibrium between the sample and ground. In this case, we have shown a sample with a work function greater than the energy of ground. To establish equilibrium, electrons move from the spectrometer to the sample. The excess negative charge on the sample shifts its local vacuum level above the absolute vacuum level, and generates an electrostatic field outside the sample surface. The ‘local’ vacuum is very important because it is what is observed experimentally when one measures the secondary electron cut-off of a sample. One cannot directly measure the absolute vacuum level in solid state photoemission spectroscopy. From the above description one can see why the Fermi level is more suitable as a reference level in solid state photoemission spectroscopy. The Fermi level remains constant due to the constraints of thermodynamic equilibrium. However, as a result of this equilibrium the local vacuum level (and only measureable vacuum level) becomes variable and sample dependent, and is thus not a suitable reference level. We will now consider what happens when a molecule adsorbs to the sample surface. Fig. 3.1.1 (D) shows the situation of the spectrometer and substrate at equilibrium, and the molecule in the gas phase at infinite distance away from the substrate. The molecule’s vacuum level is aligned with the absolute vacuum level, and the HOMO binding energy is the same as the gas phase ionization energy. When the molecule comes close to the sample surface, as shown in Fig. 3.1.1 (E), it feels the electrostatic field of the sample’s local vacuum level. This field shifts all electron energy levels of the molecule upward. When the molecule adsorbs to the surface, a so called dipole, or shift in the sample’s vacuum level can occur, due to the push-back effect; however, this not shown in the figure for simplicity. If no charge transfer between the molecule and substrate

15

14 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

Fig. 2.3.3. Here, the grey data points represent HOMO binding energies measured with our conventional layer-by-layer method. The black point represents the HOMO binding energy of a thick α-NPD film deposited in a single step and having no prior exposure to radiation. Here it can be seen that the even without prior radiation exposure, the black point falls along the general trend of the layer-by-layer data points. This plot validates that our measurement procedure is sufficient to minimize radiation artefacts and produce reliable energy-level profiles.

Figure 2.3.2 – (a) Valence UPS spectra collected during continuous UV irradiation of a 5 nm thick α-NPD film. Each spectrum differs by 5 minutes of irradiation time. (b) Plot of HOMO binding energy versus UV irradiation time.

Figure 4 – Plot of HOMO binding energy of α-NPD on CuO, as determined from numerous independent measurements. The Black cross indicates a measurement taken of a thick film with no prior irradiation. This plot shows that indeed there is a thickness dependence to HOMO binding energy (i.e. not all HOMO binding energy changes are caused by radiation damage). All measurements in this plot were taken using the minimum sample irradiation needed.

3 - Additional discussion 3.1 - Discussion of reference levels Reference levels are an important consideration in interpreting photoemission spectra, and often a point of confusion.32-36 In gas phase spectroscopy, binding energies are calibrated using optical methods. For example, the binding energy for a particular electron state can be determined by measuring the energy of the corresponding emission line in a Rydberg series. By doing so, the binding energy is calibrated with reference to the vacuum level (i.e. the energy of an electron at rest, infinite distance away from any other charged particles), such that the vacuum level is zero on the scale.

14

In solid state photoemission spectroscopy, the situation is more complicated. In general, when a sample contacts a spectrometer, the electron chemical potentials of the two equilibrate, resulting in Fermi level alignment. We will illustrate the concepts of reference levels in solid state spectroscopy using a step-wise example, as shown in Fi.g 3.1.1. Part (A) represents the spectrometer and sample in complete isolation from one another, and from any other source of charge. Here, the two materials both have their own distinct Fermi energies defined by the energies of their highest filled electron states. These Fermi energies can in principle (although not in practice) be defined the absolute energy scale. The zero of this scale is indicated by the horizontal dashed line in Fig. 3.1.1, labelled “vac. level”. In this case, and neglecting surface dipoles of the sample and spectrometer, we can say their vacuum levels are aligned. The next step to consider is grounding the spectrometer, as illustrated in (B). By grounding, we mean that the spectrometer is in electrochemical equilibrium with the electron reservoir known as “ground” or “earth”. In this case, the Fermi level of the spectrometer aligns with the Fermi level of ground, causing the ‘local’ vacuum level of the spectrometer to deviate from the absolute reference level. The energy of the Fermi level of ground on the absolute energy scale has been determined experimentally to be ~ 4.09 eV.32 In the next step, we consider what happens when a sample contacts the spectrometer, as shown in (C). When contact is made, electrons will move in order to establish equilibrium between the sample and ground. In this case, we have shown a sample with a work function greater than the energy of ground. To establish equilibrium, electrons move from the spectrometer to the sample. The excess negative charge on the sample shifts its local vacuum level above the absolute vacuum level, and generates an electrostatic field outside the sample surface. The ‘local’ vacuum is very important because it is what is observed experimentally when one measures the secondary electron cut-off of a sample. One cannot directly measure the absolute vacuum level in solid state photoemission spectroscopy. From the above description one can see why the Fermi level is more suitable as a reference level in solid state photoemission spectroscopy. The Fermi level remains constant due to the constraints of thermodynamic equilibrium. However, as a result of this equilibrium the local vacuum level (and only measureable vacuum level) becomes variable and sample dependent, and is thus not a suitable reference level. We will now consider what happens when a molecule adsorbs to the sample surface. Fig. 3.1.1 (D) shows the situation of the spectrometer and substrate at equilibrium, and the molecule in the gas phase at infinite distance away from the substrate. The molecule’s vacuum level is aligned with the absolute vacuum level, and the HOMO binding energy is the same as the gas phase ionization energy. When the molecule comes close to the sample surface, as shown in Fig. 3.1.1 (E), it feels the electrostatic field of the sample’s local vacuum level. This field shifts all electron energy levels of the molecule upward. When the molecule adsorbs to the surface, a so called dipole, or shift in the sample’s vacuum level can occur, due to the push-back effect; however, this not shown in the figure for simplicity. If no charge transfer between the molecule and substrate

15

NATURE MATERIALS | www.nature.com/naturematerials 15

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

occurs (i.e. when the Fermi level of the substrate lies within the gap of the molecule) as shown in (E) then the binding energies of the molecular levels will shift in response to changes in the substrate’s local vacuum level. This is why the HOMO binding energy of molecules shifts linearly with substrate work function when the work function is less than the molecule’s ionization energy. When the work function is greater than the molecule’s ionization energy, as shown in Fig. 3.1.1 (F), charge transfer from the HOMO of the molecule to the substrate occurs, allowing for equilibrium of the molecule with the Fermi level. As a result, the binding energies of the molecule are no longer dependant on the work function of the substrate.

Figure 5 – Energy-level diagrams illustrating how Fermi levels and vacuum levels align when (A) a spectrometer and sample are isolated, (B) a spectrometer equilibrates with ground, (C) a sample comes into contact and equilibrates with a grounded spectrometer, (D) a gas phase molecule not in contact with the grounded sample, (E) a molecule adsorbed to the surface of the grounded sample, where charge transfer cannot occur, (F) a molecule adsorbed to a grounded sample, where charge transfer can occur.

16

3.2 - Volta potential at substrate surfaces and how it affects binding energy When two metals come into contact, electrons are transferred from the metal with the low

work function to the metal with the high work function. The result is a contact potential difference between the two metals, known as the Volta potential. A potential difference is established between any two materials in electrochemical equilibrium, in which charge transfer was required to reach equilibrium. A thin metal oxide film in contact with a metal will have a contact potential difference. Since most oxides have work functions that are much higher than their parent metals, electron transfer goes from the metal to the oxide in order to align the two materials’ Fermi levels. This produces a negative potential on the oxide surface, which causes an upward shift of the oxide’s vacuum level. An upward shift in the oxide’s vacuum level is equivalent to saying that there is a negative electric field at the oxide/vacuum interface. When a molecule approaches the oxide surface, the molecules ‘feels’ the negative field of the oxide surface. This field shifts all electron levels upward in energy, thereby moving their binding energies to lower values, closer to the Fermi level of the substrate.

1 John F. Moulder, W. F. S., Peter E. Sobol. Handbook of X-ray Photoelectron Spectroscopy. (Perkin Elmer, Physical Electronics Division, 1993).

2 Lu, G., Bernasek, S. L. & Schwartz, J. Surf. Sci. 458, 80-90 (2000). 3 Zimmermann, R. et al. J. Phys.-Condes. Matter 11, 1657-1682 (1999). 4 Ingle, N. J. C., Hammond, R. H. & Beasley, M. R. Journal of Applied Physics 89, 4631-

4635 (2001). 5 T. J. Chuang, C. R. B., D. W. Rice. Surf. Sci. 59, 413-429 (1976). 6 Ghijsen, J. et al. Phys. Rev. B 38, 11322-11330 (1988). 7 Grunert, W. et al. J. Phys. Chem. 95, 1323-1328 (1991). 8 Yamada, M., Yasumaru, J., Houalla, M. & Hercules, D. M. Journal of Physical

Chemistry 95, 7037-7042 (1991). 9 Lecuyer, S., Quemerais, A. & Jezequel, G. Surf. Interface Anal. 18, 257-261 (1992). 10 Santucci, S., Cantalini, C., Crivellari, M., Lozzi, L., Ottaviano, L., Passacantando, M.

Jounal of Vacuum Science and Technology A 18, 1077-1082 (2000). 11 Deng, H. H., Qian, P., Sanada, N., Yoneya, M. & Nanjo, H. J. Electrochem. Soc. 150,

B336-B341 (2003). 12 Nanjo, H. et al. Electrochim. Acta 55, 4685-4693 (2010). 13 Kwok, Y. S., Zhang, X. X., Qin, B. & Fung, K. K. Appl. Phys. Lett. 77, 3971-3973

(2000). 14 Jain, S., Adeyeye, A. O., Chan, S. Y. & Boothroyd, C. B. J. Phys. D-Appl. Phys. 37,

2720-2725 (2004). 15 Bhargava, G., Gouzman, I., Chun, C. M., Ramanarayanan, T. A. & Bernasek, S. L. Appl.

Surf. Sci. 253, 4322-4329 (2007). 16 Uozumi, T., Okada, K. & Kotani, A. J. Phys. Soc. Jpn. 62, 2595-2599 (1993). 17 Laubach, S. et al. Phys. Chem. Chem. Phys. 9, 2564-2576 (2007). 18 Zhai, H. J., Li, S., Dixon, D. A. & Wang, L. S. J. Am. Chem. Soc. 130, 5167-5177 (2008). 19 Vanelp, J. et al. Phys. Rev. B 44, 6090-6103 (1991).

17

16 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

occurs (i.e. when the Fermi level of the substrate lies within the gap of the molecule) as shown in (E) then the binding energies of the molecular levels will shift in response to changes in the substrate’s local vacuum level. This is why the HOMO binding energy of molecules shifts linearly with substrate work function when the work function is less than the molecule’s ionization energy. When the work function is greater than the molecule’s ionization energy, as shown in Fig. 3.1.1 (F), charge transfer from the HOMO of the molecule to the substrate occurs, allowing for equilibrium of the molecule with the Fermi level. As a result, the binding energies of the molecule are no longer dependant on the work function of the substrate.

Figure 5 – Energy-level diagrams illustrating how Fermi levels and vacuum levels align when (A) a spectrometer and sample are isolated, (B) a spectrometer equilibrates with ground, (C) a sample comes into contact and equilibrates with a grounded spectrometer, (D) a gas phase molecule not in contact with the grounded sample, (E) a molecule adsorbed to the surface of the grounded sample, where charge transfer cannot occur, (F) a molecule adsorbed to a grounded sample, where charge transfer can occur.

16

3.2 - Volta potential at substrate surfaces and how it affects binding energy When two metals come into contact, electrons are transferred from the metal with the low

work function to the metal with the high work function. The result is a contact potential difference between the two metals, known as the Volta potential. A potential difference is established between any two materials in electrochemical equilibrium, in which charge transfer was required to reach equilibrium. A thin metal oxide film in contact with a metal will have a contact potential difference. Since most oxides have work functions that are much higher than their parent metals, electron transfer goes from the metal to the oxide in order to align the two materials’ Fermi levels. This produces a negative potential on the oxide surface, which causes an upward shift of the oxide’s vacuum level. An upward shift in the oxide’s vacuum level is equivalent to saying that there is a negative electric field at the oxide/vacuum interface. When a molecule approaches the oxide surface, the molecules ‘feels’ the negative field of the oxide surface. This field shifts all electron levels upward in energy, thereby moving their binding energies to lower values, closer to the Fermi level of the substrate.

1 John F. Moulder, W. F. S., Peter E. Sobol. Handbook of X-ray Photoelectron Spectroscopy. (Perkin Elmer, Physical Electronics Division, 1993).

2 Lu, G., Bernasek, S. L. & Schwartz, J. Surf. Sci. 458, 80-90 (2000). 3 Zimmermann, R. et al. J. Phys.-Condes. Matter 11, 1657-1682 (1999). 4 Ingle, N. J. C., Hammond, R. H. & Beasley, M. R. Journal of Applied Physics 89, 4631-

4635 (2001). 5 T. J. Chuang, C. R. B., D. W. Rice. Surf. Sci. 59, 413-429 (1976). 6 Ghijsen, J. et al. Phys. Rev. B 38, 11322-11330 (1988). 7 Grunert, W. et al. J. Phys. Chem. 95, 1323-1328 (1991). 8 Yamada, M., Yasumaru, J., Houalla, M. & Hercules, D. M. Journal of Physical

Chemistry 95, 7037-7042 (1991). 9 Lecuyer, S., Quemerais, A. & Jezequel, G. Surf. Interface Anal. 18, 257-261 (1992). 10 Santucci, S., Cantalini, C., Crivellari, M., Lozzi, L., Ottaviano, L., Passacantando, M.

Jounal of Vacuum Science and Technology A 18, 1077-1082 (2000). 11 Deng, H. H., Qian, P., Sanada, N., Yoneya, M. & Nanjo, H. J. Electrochem. Soc. 150,

B336-B341 (2003). 12 Nanjo, H. et al. Electrochim. Acta 55, 4685-4693 (2010). 13 Kwok, Y. S., Zhang, X. X., Qin, B. & Fung, K. K. Appl. Phys. Lett. 77, 3971-3973

(2000). 14 Jain, S., Adeyeye, A. O., Chan, S. Y. & Boothroyd, C. B. J. Phys. D-Appl. Phys. 37,

2720-2725 (2004). 15 Bhargava, G., Gouzman, I., Chun, C. M., Ramanarayanan, T. A. & Bernasek, S. L. Appl.

Surf. Sci. 253, 4322-4329 (2007). 16 Uozumi, T., Okada, K. & Kotani, A. J. Phys. Soc. Jpn. 62, 2595-2599 (1993). 17 Laubach, S. et al. Phys. Chem. Chem. Phys. 9, 2564-2576 (2007). 18 Zhai, H. J., Li, S., Dixon, D. A. & Wang, L. S. J. Am. Chem. Soc. 130, 5167-5177 (2008). 19 Vanelp, J. et al. Phys. Rev. B 44, 6090-6103 (1991).

17

NATURE MATERIALS | www.nature.com/naturematerials 17

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.

18

20 Kroger, M. et al. Organic Electronics 10, 932-938 (2009). 21 Greiner, M. T. et al. Appl. Phys. Lett. 96, 213302 (2010). 22 Soriano, L., Abbate, M., Alders, D. & Sanz, J. M. Solid State Commun. 91, 551-554

(1994). 23 Meyer, J. et al. Appl. Phys. Lett. 96 (2010). 24 Tjeng, L. H. et al. Phys. Rev. B 41, 3190-3199 (1990). 25 Hill, I. G., Makinen, A. J. & Kafafi, Z. H. Journal of Applied Physics 88, 889-895 (2000). 26 Amy, F., Chan, C. & Kahn, A. Organic Electronics 6, 85-91 (2005). 27 Betti, M. G., Kanjilal, A. & Mariani, C. J. Phys. Chem. A 111, 12454-12457 (2007). 28 Kahn, A., Koch, N. & Gao, W. Y. J. Polym. Sci. Pt. B-Polym. Phys. 41, 2529-2548

(2003). 29 Koch, N. J. Phys.-Condes. Matter 20 (2008). 30 Schlaf, R., Merritt, C. D., Picciolo, L. C. & Kafafi, Z. H. Journal of Applied Physics 90,

1903-1910 (2001). 31 Zou, Y. et al. Surf. Sci. 600, 1240-1251 (2006). 32 Anderson, S. E. & Nyberg, G. L. J. Electron Spectrosc. Relat. Phenom. 52, 293-302

(1990). 33 Broughton, J. Q. & Perry, D. L. Surf. Sci. 74, 307-317 (1978). 34 Carley, A. F., Joyner, R. W. & Roberts, M. W. Chem. Phys. Lett. 27, 580-582 (1974). 35 Hagstrum, H. D. Suface Science 54, 197-209 (1976). 36 Sitte, J. Chem. Phys. Lett. 42, 131-132 (1976).

18 NATURE MATERIALS | www.nature.com/naturematerials

SUPPLEMENTARY INFORMATION DOI: 10.1038/NMAT3159

© 2011 Macmillan Publishers Limited. All rights reserved.