ITO/DIPO-Ph4 layer morphology
DIPO-Ph4 growth mode on indium tin oxide
As it was previously described, the interactions between the substrate and the organic layer, and inside the layer itself, lead to three diﬀerent growth modes.
To understand the ITO/DIPO-Ph4 interface, it is important to characterize the elec-trode surface: roughness, contaminations, defaults. Before all further analyses, atomic force microscopy (AFM) was used to characterize both the ITO substrates and the organic layer morphology on ITO (Appendix B.3.1). In this first part, we will present the mor-phology of the ITO surface and the eﬃciency of the used cleaning process (Appendix A.1).
AFM images of the bare ITO show many light dots, which are attributed to contam-ination clusters. After the cleaning process, no cluster appears on the ITO surface. The chemical cleaning is eﬃcient as the ITO surface appears without particle contamination. The ITO roughness is 1 nm and the surface shows a granular morphology. This was al-ready observed in the literature for sputtered ITO. [78, 79] Besides, we performed some current-sensing AFM (CS-AFM). The sample is biased and the AFM tip is grounded. This means that for a negative bias, electrons are travelling from the substrate to the tip, and in the opposite direction for a positive bias. Concerning the current image color appearance, the color scale for the negatively biased image has been reversed in compar-ison to the positively biased image so that the current minimum is blue in each case and the absolute current maximum is red. The current images are represented between 0 and (−)1 nA to correlate the observation with the ITO granular morphology, but the observed maximum current is +25 nA and −25 nA for the positively and the negatively biased im-age, respectively. CS-AFM images show that the conduction is clearly happening through the ITO grains.
We are thus able to get a cleaned ITO surface with a small roughness despite a granular surface. The CS-AFM experiments confirm the metallic behaviour of the ITO although it is a doped n-type SC. The Sn doping of the commercial ITO is eﬃcient to make ITO conductive. We will now focus the analysis of DIPO-Ph4 on ITO samples in order to investigate the organic layer morphology.
Several samples were prepared with a change in the deposited organic material DIP amount. The weight, followed by a quartz balance (QB), increases per surface unit dur-ing the time evaporation (Appendix B.1). It is then converted into a molecular sur-face density using the density given by XRD experiment.  The diﬀerent samples are named by their QB-coverage We studied three diﬀerent samples: a “thin” layer of 0.5 × 1015 molecule • cm−2, an intermediate layer of 2 × 1015 molecule • cm−2, and a “thick” layer 10 × 1015 molecule • cm−2. AFM images were used to determine the real value of the coverage and make correlations with the QB-coverage.
After the deposition of 0.5 × 1015 molecule • cm−2, the AFM image (Figure II.2a)) shows that 3D clusters cover 20 % of the ITO surface. The average cluster height is 20 nm, with an average diameter of 200 nm. Given the cluster density, an average vol-ume of DIPO-Ph4 per surface unit is obtained, leading to a molecular surface density of 0.55 × 1015 molecule • cm−2 (using the density of Ref. 58), in excellent agreement with the QB-coverage. The clusters are homogeneously distributed on the surface. For the 2 × 1015 molecule • cm−2 deposit (Figure II.2b)), the DIPO-Ph4 layer now covers 55 % of the ITO surface. The 3D growth of slightly elongated mounds increases, and their aver-age diameter is now 300 nm. Their average height increases to ∼30 nm. The molecular coverage deduced from the AFM image is 1.8 × 1015 molecule • cm−2, also in agreement with the QB-coverage. Finally, for the “thickest” layer (Figure II.2c)), corresponding to a deposit of 10 × 1015 molecule • cm−2 (QB-coverage), the film covers more than 95 % of the ITO surface and its average thickness is ∼50 nm.
Figure II.2 – AFM images for DIPO-Ph4: a1) 0.5 × 1015 molecule • cm−2; b1) 2 × 1015 molecule • cm−2; c1) 10 × 1015 molecule • cm−2; a2), b2), and c2) are the corre-sponding profiles indicated by the white straight lines in a1), b1), and c1) respectively. The molecular coverage, expressed in molecule • cm−2 (QB-coverage), is obtained from the quartz balance monitor.
estimated at 7.1 × 1015 molecule • cm−2 which is less than the QB-coverage. For this layer set, AFM measurements agree quite well with QB-coverages. An increasing error is ob-served with the increasing coverage, which may be due to an approximate QB parameter set up.
In the following, we will describe the growth mode. As presented in Chapter I.2.1, when layers are deposited over homogeneous substrates, like single crystals, three types of growth are classically considered.  The first one is the Franck Van der Merwe one which leads to a layer-by-layer growth. Our results show that this is not the case for the DIPO-Ph4 layers. In the second one, the Volmer-Weber growth, the organic molecules clusters are islands, leaving bare substrate areas. That is what is seen for the DIPO-Ph4-Ph layers. It arises from the fact that the interactions between molecules (π-stacking) is much stronger than the interaction between the DIPO-Ph4 molecule and the ITO substrate. In the third mode, the Stranski-Krastanov growth, the molecule-molecule interaction competes with the molecule-substrate interaction. A thin wetting layer covers all the substrate, on top of which the 3D island growth mode takes place. If a Stranski-Krastanov mode stands, the wetting layer, if present, should not exceed ∼2 nm. The AFM resolution does not allow concluding in this section. Note that X-ray photoemission spectroscopy will help us to conclude on the growth mode (see Chapter III.2.1).
For the DIPO-Ph4 layers, as soon as deposited, islands are formed with an average height > 10 nm. The molecule/interface is thus seen only at the island rings. To perform a deeper analysis, it would be interesting to decrease the average cluster height. We were also interested in increasing the material crystallization speed. To do so, we performed some thermal annealing treatments.
Access to the first deposited layer and to the crystallized material
Two kinds of treatment were performed:
• One after the evaporation, in ambient pressure condition at a temperature near the evaporation temperature of the DIPO-Ph4 (T ≥ 170 ◦C) to perform some molecular desorption and access to the first deposited layers (ITO/DIPO-Ph4 interface).
• One during the evaporation, i.e. evaporation on hot substrate (T = 100 ◦C), to bring some energy to the system and increase the growth speed (thermodynamic equilibrium).
This study was performed while replacing the ITO substrate by a very thin (100 nm) nitride silicon (Si3N4) substrate. No cleaning treatment has been performed on these substrates before the molecular evaporation. The interface was then studied via XAS absorption measurements to determine the molecular orientation. AFM results are pre-sented in Figure II.3.
First, it is important to notice that the layer morphology on a Si3N4 is similar to the one on ITO substrate (no thermal annealing, presented in Figure II.3). The un-annealed sample morphologies are the same as the ones presented in the previous section: slightly elongated clusters with an average height of 30 nm for the 1 × 1015 molecule • cm−2 sam-ple. AFM-calculated coverages are 1.1 × 1015 molecule • cm−2 (25 % of the Si3N4 covered) and 11 × 1015 molecule • cm−2 for the 1 × 1015 and the 10 × 1015 molecule • cm−2 samples, respectively. A similar morphology between the two substrates may result from a similar surface energy.
After a post-annealing treatment at T ≥ 170 ◦C on the 1 × 1015 molecule • cm−2 sample (post thermal annealing, presented in Figure II.3a)), the average height of the clusters decreases to 13 nm. The mounds still cover 20 % of the Si3N4 surface, with an average diameter of 230 nm. The molecular surface density calculated, from the AFM image, is divided by 3 (0.3 × 1015 molecule • cm−2). The crystallization on a hot substrate (under thermal annealing, presented in Figure II.3a)) leads to the formation of twinned DIPO-Ph4 structures that are more elongated than in the previous cases.
Concerning the 10 × 1015 molecule • cm−2 sample, after a post-annealing treatment (post thermal annealing, presented in Figure II.3b)), large islands are observed with an average step of 80 nm. The DIPO-Ph4 layer does not cover the entire substrate. In the case of an under-annealing treatment (under thermal annealing, presented in Fig-ure II.3b)), the twinned clusters continuously grow to form large molecular crystals on top of the Si3N4 substrate. During an under-annealing treatment, the energy brought via the hot substrate, changes the molecular growth by increasing the molecule diﬀusion on the surface.
In the case of well-separated 3D-clusters, post-annealing treatment leads to molecular desorption. This is the case for the 1 × 1015 molecule • cm−2. The average height of the clusters decreases while the overall area coverage and the average cluster diameter stay constant. With such a treatment, we can access the first deposited layers which are repre-sentative of the initial stages of the evaporation. As far as the 10 × 1015 molecule • cm−2 sample is concerned, there is no homogeneous desorption as the intermolecular interaction compete with the desorption eﬀect. In this case, the post-annealing treatment leads to the formation of large DIPO-Ph4 islands which do not entirely cover the substrate. Note that, the 10 × 1015 molecule • cm−2 sample prepared under thermal annealing can be used as a reference sample for crystallized DIPO-Ph4 structure.
To perform XAS spectroscopy, it is important to understand the electronic mechanisms which take place during the absorption process. Density functional theory (DFT) calcula-tions were used to describe the empty molecular states (LUMOs) localized on the carbon and the oxygen atoms of DIPO-Ph4 structure (Figure II.4). Furthermore, it was useful to visualize the concerned molecular orbital (MO) to obtain information on the molecular orientation. Indeed, to determine the molecular orientation inside the clusters, we must be sure that there is a MO with one main direction in the molecule geometry.
For C 1s transitions (Figure II.4a)), the lowest unoccupied molecular orbital (LUMO) and unoccupied MOs of higher energy (LUMO+1 to LUMO+6) are described via DFT calculations. LUMOs are mainly composed of π∗ MOs which are perpendicular to the DIP core or to the phenyl (Ph) groups. They can be divided in 3 groups. The first one is centred at ∼0.3 eV above the LUMO and composed of LUMO, LUMOs +1, and +2. The second one is centred at 1.3 eV above the LUMO and composed of LUMOs +3 and +4.
Figure II.4 – Electronic scheme of the a) C 1s and b) O 1s transition for DIPO-Ph4. Indicated energies are related to the LUMO and three MOs are represented with an isovalue of 0.04 (LUMO, LUMO+3 and LUMO+5). The principal weight localization of the MOs is emphasized with ellipses.
and +6. Looking at the localization of these groups, it appears that the first one is more localized on the DIP core, whereas the second one is more localized on the Ph groups. For the last one, the MOs are localized, with the same weight, on the DIP core and on the Ph groups. These three groups will be now called πDIP∗, πPh∗, and πDIP∗+Ph. Concerning the O 1s transitions (Figure II.4b)), there are two main transitions as only LUMO and LUMO+5 are localized on O atoms.
The geometry of the free molecule was previously described: four Ph groups almost parallel to the DIP core. Consequently, there is one main plane following the DIP core and four other planes following each of the four Ph groups. Thus, for the πDIP∗ MOs there is one main direction which is perpendicular to the DIP core. On the contrary, for the πPh∗ MOs, there is no main direction. Finally, the πDIP∗+Ph MOs have no main orientation because they are localized both on the DIP core and on the Ph groups.
In the following, we will describe the absorption transition. The X-ray absorption results in the electron transition from the core level to the empty state. Thus, the ab-sorption eﬃciency depends on the energy level of the empty states. It will also rely on the relative geometry between the electrical component of the X-ray beam and the direction of the MOs. Therefore, for the DIPO-Ph4 structure, the absorption will be high if the polarization is parallel to the πDIP∗, i.e. perpendicular to the molecule. In other terms, with an X-ray beam normal to the surface (polarization parallel to the surface) there will be a high absorption at πDIP∗ energy if DIPO-Ph4 molecules are perpendicular to the surface (πDIP∗ parallel to the surface).
Thanks to absorption spectroscopy we saw that we can access the molecular orientation inside the layer. Such information will be important to know, as a good crystallization can lead to a better charge conduction. In this context, we can stress upon the significance of an adapted characterization tool, such as scanning transmission X-ray microscopy (STXM) that has the potential to disentangle molecular domain orientation information in thin films.  STXM, and other synchrotron based soft X-ray characterization tools, such as resonant soft X-ray scattering, are already been widely used in conjunction with other laboratory based techniques to investigate the complex morphology of organic devices. [81, 82, 83, 84, 85, 86, 87]
Absorption analyses were performed at the PolLux beamline (SLS synchrotron) on 4 samples (Appendix B.3.2):
• Sample A: 1 × 1015 molecule • cm−2 QB-coverage with no thermal treatment.
• Sample B: 1 × 1015 molecule • cm−2 QB-coverage with a post-annealing treatment.
• Sample C: 10 × 1015 molecule • cm−2 QB-coverage with no thermal treatment.
• Sample D: 10 × 1015 molecule • cm−2 QB-coverage with an under-annealing treat-ment.
The STXM results have been used to confirm the thickness of the layers (using Equa-tion) and thus to confirm the AFM results.
The obtained image at hν = 285 eV, corresponding to the maximum absorption at the C-K edge, is extracted from the [X; Y ; hν] image stack to obtain the optical density (OD) images ([X; Y ]) at 285 eV. This image is used to calculate the corresponding thickness using (Equation (I.3)). In Figure II.5, images are STXM images obtained from the ab-sorption measurements, but are presented in the same way as AFM images for comparison purpose.
Figure II.6 – Sample A: a) STXM decomposition mapping; b1) Absorption spectra ex-tracted for each region (the substrate (Sub) spectrum is not represented as there is no absorption from this region); b2) π∗ area after subtraction of the step function background and the σ∗ area.
thermal annealing in Figure II.5b)). [X; Y ] STXM analyses are in great agreement with AFM results.
Further investigations of the [X; Y ; hν] stacks were performed with respect to the pho-ton energy related variation (spectroscopy). Using principal component analyses (PCA), we determined the main components (absorption spectra) of the stacks. Then, the first four principal components have been used to decompose the STXM stacks in four main regions. For each region, the absorption spectra were then extracted. These spectra are the average one for the found regions. After normalization, spectra have been fitted as described in the Appendix B.5.1. The potential ionization (IP = 288.9 eV) value was de-termined from the XPS C 1s binding energy and the measurement from the work function (see next section, Chapter III.2). Results for Sample A are presented in Figure II.6 and Table II.1.
The decomposition of the [X; Y ; hν] stack for Sample A (Figure II.6a)) leads to four regions precisely localized on the sample. The main region is the one localized between the organic clusters. There is no absorption (or at least too low to be detected with the STXM thickness resolution). This corresponds to the response from the substrate. The organic mounds are decomposed in three regions: one at the top and two on the edge. These regions will now be named Top, E2, and E1 respectively.
for comparison (Figure II.6b1)). A clear distinction of the π∗ and the σ∗ regions is evident in the figure. After subtraction of the background and the σ∗ region, the π∗ region is emphasized in Figure II.6b2) for clarity. It has been fitted with three Gaussian curves (G-FWHM values are indicated in Table II.1). The first one, entered at ∼ 285 eV is attributed to the first transition group πDIP∗. πPh∗ is at ∼286 eV, and πDIP∗+Ph is centred at ∼288 eV. DFT calculations for a free molecule gives a πPh∗ −πDIP∗ of ∼1.3 eV and a πDIP∗+Ph −πDIP∗ of ∼2.3 eV (see Figure II.4a)). Experimentally, the value of the πPh∗ −πDIP∗ is ∼1.2 eV, which is very close to the theoretical calculation. For the πDIP∗+Ph transition, the diﬀerence with the πDIP∗ is ∼3 eV. This value is 1 eV higher than the theoretical one. More LUMOs than the 2 calculated ones (+5 and +6) must thus participate to the πDIP∗+Ph transition.
In the absorption spectra, we also focused on the contribution of the π∗ transition in comparison to the σ∗ one (Table II.1). The intensity of the σ∗ transition remains identical between spectra extracted from Top to E1 regions. This intensity was used as a reference to evaluate the π∗ transition in the diﬀerent regions. The π∗/σ∗ ratio decreases from Top to E1 (from 26 % to 21 %). This evolution is mainly due to a decrease of the πDIP∗ transition (from 11 % to 8 %). This transition is the only one which can be used to access the molecular orientation. Thus, in the organic layer, there is an evolution in this orientation from the edge to the top of the cluster. In the latter (top region), molecules are perpendicularly oriented to a higher degree than in the edge. This implies a change in the orientation between the layer near the substrate and the top layers.
The analysis of the post-annealed sample (Sample B) could allow us to extract in-formation on the first evaporated layers.
Figure II.7 – STXM images for 1 × 1015 molecule • cm−2 DIPO-Ph4 with a 30° tilted angle (A’: no thermal annealing; B’: post thermal annealing). The molecular coverage, expressed in molecule • cm−2 (QB-coverage), is obtained from the quartz balance measurement. The dotted line for Sample B’ indicates the position for the absorption linescan.
material in this sample makes the analysis harder, due to a weak signal at normal X-ray incidence. To overcome this issue, we tilted the sample (30° between the normal and the sample surface). To limit beam damage, we only performed a linescan along the organic clusters (see Figure II.7). From this line, we extracted the edge (E) and Top spectra. The values of the π∗ transitions are given in Table II.1. To compare the two samples, we per-formed the same tilted incidence angle experiments on Sample A. Results are presented in Figure II.7, Figure II.8, and Figure II.9.
From the thickness STXM images of Sample A and B with a tilted X-ray incidence angle (A’ & B’), we observe that 3D clusters appear with a bigger average height: 40 nm and 20 nm, for A’ and B’, respectively. Indeed, in this case there is a bigger depth penetration in the molecular material. It results in a bigger absorption which leads to an increase of the apparent thickness.
The analysis of Sample A analysis with the tilted incidence angle leads to the same results as with the normal incidence angle (see Table II.1). There are no big changes from the previous results as we chose a small tilt angle.
Figure II.10 – Sample C: Absorption spectra extracted for the top, plateau, and valley areas. π∗ area after subtraction of the step function background and the σ∗ area. (P and V referred to the points on the dotted line in STXM images Figure II.5b)).
to top agrees with the change of molecular orientation inside the organic cluster. Besides, the high σ∗ transition results from the several DIPO-Ph4 molecules that are parallel to the surface.
Therefore, at the beginning of the evaporation on substrates kept at room temper-ature, molecules remain parallel to the substrate. Substrate-molecule interactions are minimized in this configuration. In this case, there is a small π∗ absorption and a greater σ∗ absorption. After a few layers, molecules start to straighten and crystallize, driven by the π-stacking interaction. There is a change in the molecular orientation resulting in an increase of the π∗ absorption. The edges of the mounds reflect this change, as they are close to the first deposited layers. Besides, crystallization leads to molecular orientation as described in Figure I.7a). It shows that molecules are not all parallel to each other: they are V-shaped. Consequently, σ∗ transition is still present and the Top absorption spectrum is equivalent to that of the crystallized organic material.
To pursue the analysis, we performed the experiments on the 10 × 1015 molecule • cm−2 QB-coverage (Sample C). Linescan was performed across a high cluster and results are presented in Figure II.10. Points P and V are localized respectively on a plateau and on a valley, as depicted in STXM image (Figure II.5b)) along the dotted line.
In this sample, π∗ absorption is greater at the Top and on the plateau (P). However, in a valley (V), the π∗ absorption decreases from 30 % to 26 %. This confirms the change of morphology.
Let us focus now on Sample D to get the absorption spectra of a crystallized cluster.
Figure II.12 – Schematic representation of the molecular orientation for Samples A, B & D.
crystallized from the first deposited layers. The substrate temperature brings enough energy to start crystallization and overcome molecule-substrate interaction. We present in Figure II.12, the schematic molecular orientation inside the organic cluster.
With these samples, we obtained two opposed molecular orientations: one which present parallel molecules on top of the surface (Sample B) and another in which the organic mounds are directly crystallized as soon as deposited on the substrate (Sample D). We are thus able to tailor the molecular orientation thanks to the evaporation conditions. With the perspective of organic electronics, it is important to control the morphology and the molecular orientation to enhance, for example, the charge mobility through the organic layer.
We assume that on ITO, the same observation would be seen. Indeed, the surface energy between an oxide and an air exposed silicon nitride must be similar. To verify this hypothesis, STXM experiments on Si3N4/ITO/DIPO-Ph4 samples have been planned but the proposal was not yet accepted.
The ITO XPS overview (Figure III.1) shows all the diﬀerent core levels that can be analysed. The C 1s peak due to carbon contamination is small in comparison with the indium peak. We will further discuss this point in the study of the ITO/DIPO-Ph4 interface.
The In 3d5/2 spectra of the bare, chemically cleaned surface are presented in Fig-ure III.2a). To reach “surface sensitive” conditions, a photon energy of 600 eV is used, corresponding to a photoelectron kinetic energy (KE) of ∼155 eV and a calculated IMFP λITO of ∼0.56 nm (see Figure I.16). More “bulk sensitive” conditions are obtained with photons of energy 825 eV, corresponding to a photoelectron KE of ∼380 eV and a λITO ∼0.92 nm. Both spectra present an asymmetry towards higher binding energy (BE), but the spectrum measured in more ‘bulk sensitive” conditions at hν = 825 eV is more asym-metric. This indicates that the electronic structure changes from the surface to the inner layers. The observed asymmetry is related to the electronic structure of the material and not to surface contamination. Indeed the In 3d spectra, unlike the O 1s ones, are not sensitive to adsorbed species.  The attribution of the observed asymmetry to plasmon losses has now gained wide acceptance. [38, 89] The plasmon frequencies (below 1 eV) in ITO are significantly smaller than that of classical metals, due to the lower electron density. Fitting procedures, however, diﬀer according to authors. Christou et al.  and Körber et al.  fit the In 3d5/2 spectrum with two Voigt components. The Voigtian is a weighted sum of a Gaussian and of a Lorentzian (the Lorentzian fraction as a free parameter), the low binding energy component being prevalently Gaussian. On the other hand, Gassenbauer et al.  use three Voigt functions, constrained to have the same width and the same Gaussian/Lorentzian ratio, one main peak and two plasmon losses (~ωp and 2~ωp). Naturally the value of ~ωp depends on the procedure. Here we adopt the same procedure as that used by Gassenbauer et al.,  i.e. a fit with three Voigt components of equal widths (Table III.1), corresponding to the well-screened peak (at lower BE) and to the first and second plasmon peaks.
Table of contents :
I Metal/organic interfaces
I.1 Semiconductor description
I.1.1 Inorganic semiconductor
a) Theoretical description
b) Electrode for organic electronics
I.1.2 Organic semiconductor
I.2 Metal/organic interface
I.2.1 Layer interactions
I.2.2 Energetic level alignment
I.3 X-ray characterizations
I.3.1 Physical principles
I.3.2 Sampling depth
I.3.3 X-ray absorption
II DIPO-Ph4 layer on ITO substrate
II.1 ITO/DIPO-Ph4 layer morphology
II.1.1 DIPO-Ph4 growth mode on indium tin oxide
II.1.2 Access to the first deposited layer and to the crystallized material
II.2 Molecular orientation
II.2.1 Molecular description
II.2.2 Absorption spectroscopy
III.1 ITO characterization
III.1.1 Electronic properties
III.1.2 Carrier concentration
III.1.3 Electron energy level scheme
III.2 ITO/DIPO-Ph4 interface
III.2.1 Core levels XPS spectroscopy
a) Indium and tin core levels
b) Carbon and oxygen levels
c) DFT calculation correlation
III.2.2 Valence band energy level
III.2.3 Electron energy level scheme
III.3 Charge transfer from DIPO-Ph4 to ITO
III.3.1 Resonant photoemission spectroscopy
III.3.2 ITO/DIPO-Ph4 interface
a) C K-edge
b) O K-edge
III.3.3 Pump-probe experiments
IV DIPO-Ph4 as interfacial layer
IV.1 Solar devices
IV.1.1 Photovoltaic mechanism
IV.1.2 Photovoltaic devices
a) Photovoltaic generations
b) Organic solar cells
c) Interfacial layer
IV.2 Organic electronics application
IV.2.1 Energetic alignment
IV.2.2 Photovoltaic response
V DIP heteroatom effect
V.1 ITO/DIP layer morphology
V.1.1 Molecular description
V.1.2 DIPS-Ph4 and DIPSe-Ph4 growth mode
V.2 Electronic properties
V.2.1 Core levels XPS spectroscopy
V.2.2 Valence band energy level
V.2.3 Electron energy level scheme