A. Excitons

The dissociation of photogenerated electron–hole pairs (excitons) is a key factor for generating carriers in organic semiconductors. Exciton dissociation is affected by the relative permittivity of the solid (ϵ). The attractive force between an electron–hole pair is given by Coulomb's law.

1

Here, ϵ₀, q₁, q₂, and r are the absolute permittivity, the elementary charges on the electron and the hole, and the distance between them. In a solid with a small value of ϵ, the positive and negative charges experience a strong attractive force, whereas, in a solid with a large value of ϵ, the positive and negative charges experience a relatively weak attractive force. Inorganic semiconductors have large values for ϵ. For example, the value for Si is 11.9, and the exciton has a large diameter of 9.0 nm and is localized over about 10⁴ Si atoms (Fig. 4(a)). The thermal energy at room temperature is sufficient for this Wannier-type exciton to dissociate immediately into a free electron and a hole, thereby generating photocurrent. On the other hand, organic semiconductors have small values for ϵ. For example, the value for C₆₀ is 4.4, and the exciton diameter is just 0.50 nm and is localized over a single C₆₀ molecule (Fig. 4(b)). The thermal energy at room temperature is hardly sufficient for these Frenkel-type excitons to dissociate into free electrons and holes, and they can easily relax into the ground state (Fig. 5(a)). Therefore, organic semiconductors generate few photocarriers. This is the reason why organic solar cells fabricated before the work of Tang [6] had extremely low photocurrents, of the order of less than a microampere.

B. Donor–Acceptor Sensitization

Today's organic solar cells have overcome the aforementioned problem by combining two kinds of organic semiconductor. An electron-donating molecule (D) and an electron-accepting molecule (A) for which the energetic relationship of the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) are brought together in contact or blended. When the electron-accepting molecule (A) is excited, electron transfer from the HOMO of the electron-donating molecule (D) to the HOMO of the electron-accepting molecule (A) occurs, and as a result, the A molecule charges negatively (A⁻), and the D molecule charges positively (D⁺) (Fig. 5(c)). On the other hand, when the electron-donating molecule (D) is excited, electron transfer from the LUMO of the electron-donating molecule (D) to the LUMO of the electron-accepting molecule (A) occurs, and as a result, the A and D molecules charge negatively (A⁻) and positively (D⁺), respectively (Fig. 5(d)). Irrespective of the excitation of molecules (A) and (D), the charge transferred states (D⁺A⁻) that are obtained are the same. Thus, a charge transfer (CT) exciton is formed in which the positive and negative charges are separated on neighboring D and A molecules due to photoinduced electron transfer (Fig. 5(b)). This CT exciton can dissociate to a free electron and a hole due to thermal energy at room temperature. By utilizing this donor–acceptor (D/A) sensitization, organic semiconductors have become capable of generating photocurrents of significant magnitude, of the order of milliamperes, induced by solar radiation.

C. Exciton Diffusion

A two-layer organic solar cell (Fig. 6) [6] utilizes D/A sensitization at the heterojunction. The width of the photoactive region (shaded region) is, however, limited to around 10 nm in the vicinity of the heterojunction due to the extremely small exciton diffusion length of just several nanometers [11, 12]. So, when the thicknesses of the organic layers are increased, the so-called “masking effect” occurs, namely, a dead region develops in front of the active region in which the incident solar light is absorbed and no photocurrent is generated, and as a result, the magnitude of the photocurrent is severely suppressed. A 10-nm-thick organic film can only absorb a small part of the incident solar light. However, in order to increase the efficiency of organic solar cells, it is necessary for the whole of the incident solar light to be absorbed in the 10-nm-thick active layer. Estimation of the exciton diffusion length is discussed in Section III.A.

D. Blended Junctions

In order to overcome this problem, blended junctions, which contain both donors and acceptors, were fabricated in organic semiconductors using a co-deposition technique for small-molecule organic solar cells (Fig. 7(a)) [7, 8]. From the standpoint of photocarrier generation occurring at the molecular level, there are donor/acceptor molecular contacts that act as photocarrier generation sites due to donor/acceptor sensitization in the bulk of the co-deposited layer. Blended junctions, so-called “bulk heterojunctions,” have become crucial for polymer organic solar cells [10].

Blended junctions are physically fundamentally the same as the porous TiO₂ in the dye-sensitized solar cells proposed by Grätzel in 1991 [13]. For both solid/solid and solid/electrolyte junctions, by transmitting the incident light through a vast number of heterointerfaces, the whole of the incident solar light can be absorbed in extremely thin active layers.

Figure 7(b) shows the energetic structure of the blended junction in a small-molecule cell. Photogenerated electrons in donor molecules transfer from the LUMO of the donor molecule to that of the acceptor molecule and photogenerated holes in acceptor molecules transfer from the HOMO of the acceptor molecule to that of the donor molecule. This D/A sensitization promotes photocurrent generation. The original concept is that the positive and negative charges from ionized donors and acceptors in n- and p-type organic semiconductors, respectively, are compensated by each other, and the resulting co-deposited interlayer behaves like an intrinsic semiconductor [8]. With regard to the formation of the built-in potential in a molecular solid, the built-in electric field is distributed across an i-interlayer sandwiched between n- and p-layers, similar to the case of amorphous silicon incorporating a p–i–n junction.

A. Vertical Superlattices

An ideal nanostructure is the “vertical superlattice” (Fig. 8(a)). This structure enables the efficient dissociation of photogenerated excitons at the D/A interfaces within the exciton diffusion length (5–10 nm) and the transport of electrons and holes to the respective electrodes.

In this section, the characteristics of intentionally designed vertical organic superlattice structures are discussed [12]. In the vacuum deposition of an alternating multilayered structure of two kinds of materials, each film layer has to be deposited in sequence on the substrate (Fig. 9(a)). The thicknesses of the layers can be controlled with angstrom-order precision by monitoring with a quartz crystal microbalance (QCM), in the film thickness direction only. Here, if we consider rotating the multilayered film by 90° and making it vertical with respect to the substrate and having the cross-section of the multilayered film as the exposed surface (Fig. 9(b)), then control of the structures in the direction parallel to the surface on a nanometer scale is possible by controlling the deposition film thickness.

From the standpoint of photoelectric conversion, when a vertical superlattice structure (Fig. 8(a)) consisting of alternating perylene pigment (Me-PTC, Fig. 3) layers acting as acceptor molecules and metal-free phthalocyanine (H₂Pc, Fig. 3) layers acting as donor molecules is exposed to light, efficient charge separation occurs at the interface of the donor/acceptor organic semiconductors and the photogenerated electrons and holes are transported such that they are spatially separated.

The fabrication method is shown in Fig. 10. A substrate having a flat surface was made using an epoxy resin. Me-PTC and H₂Pc were deposited alternately on the substrate by vacuum evaporation. The multilayered film was embedded using the same epoxy resin as the substrate. By using a microtome (Leica, ULTRACUT UCT) equipped with a diamond knife (Diatome Ltd), the embedded film could be sliced perpendicular to the film surface. Epoxy film containing an embedded vertical superlattice having a height of 2 µm in the center was obtained. Finally, silver electrodes were deposited on both sides of this epoxy film. The completed cell has a structure consisting of a vertical multilayered film with height of 2 µm sandwiched by Ag electrodes (Fig. 8(a)).

When only the Me-PTC was selectively excited using monochromatic light of 480 nm wavelength under a relatively small applied electric field, the photocurrent generated in the vertical superlattice was reproducibly observed. Figure 11(a) shows the dependence of the photocurrent generation quantum efficiency on the layer width (x; Fig. 8(a)) of the vertical superlattice. The layer width (x) was varied down to 2.5 nm. A decrease in x means an increase in the number of interfaces between Me-PTC and H₂Pc. At x = 2.5 nm, the number of layers will be 4 million layers per 1 cm. The quantum efficiency rapidly increases as x decreases, and the maximum, 41%, was observed at 10∼5 nm. However, it was also found that the efficiency decreased when the layer width decreased to 2.5 nm. Thus, we could experimentally determine the optimal width, x, at which the performance of photoelectric conversion in the Me-PTC/H₂Pc system is maximized to be 10–5 nm.

The increase in the quantum efficiency by decreasing the layer width (x) is related to the exciton diffusion length. Photocurrent is generated when the photogenerated excitons dissociate into free electrons and holes when they reach the Me-PTC/H₂Pc interface (Fig. 11(b)). The excitons that could not reach the boundary surface are deactivated. The range where the excitons can reach the interface is the active area in which the photocurrent is generated (shaded area), and the remainder where the excitons are deactivated is the dead area (nonshaded area). The increase in photocurrent quantum efficiency occurs due to an increase in the active area and a decrease in the dead area (Fig. 11(b)). In other words, when x is thick, that is, the number of interfaces is small, the majority region will be dead area, and only a small photocurrent is generated. When x becomes thin, that is, the number of interfaces increases, the photocurrent gradually increases due to the increase in the active area, and at x = 10–5 nm, the dead area in Me-PTC is eliminated and the maximum photocurrent is generated (Fig. 11(b)).

The exciton diffusion length can be obtained quantitatively. A semilogarithmic plot of the photocurrent quantum efficiency as a function of x (Fig. 11(c)) showed a clear linear relationship. This signifies that the concentration of the excitons that can contribute to the photocurrent decreases exponentially by increasing the distance from the interface. The distance at which 90% of the excitons can reach the interface in the case of Me-PTC was precisely determined to be 3.7 nm. Similarly, that for H₂Pc was determined to be 5.0 nm by the selective excitation of H₂Pc. As shown in Table I, by adopting the C₆₀/H₂Pc system, the exciton diffusion lengths for C₆₀ and H₂Pc were determined to be 1.7 and 4.6 nm, respectively.

In Fig. 11(a), when the layer width became as narrow as x = 2.5 nm, the quantum efficiency decreased. When the film is less than 2.5 nm thick, each layer is not completely formed as a single sheet, and the probability of growth remaining as island regions becomes higher. Such spatial incompleteness of the extraction route decreases the efficiency of transporting the electrons and holes without recombination over a distance of 2 µm to each electrode.

The result of this study shows that eliminating any dead areas, which do not generate photocurrent, as well as ensuring the formation of a continuous route to transport the spatially separated electrons and holes, is critically important to obtaining high photocurrent generation quantum efficiency in photoelectric conversion using organic semiconductors.

B. Phase Separation

In this section, we describe a fabrication method for pseudo-vertical superlattices similar to Fig. 8(a) by using co-evaporant third molecules that act as a solvent during vacuum deposition [16]. By introducing co-evaporant third molecules onto a substrate heated to +80 °C during film growth, phase-separated and crystallized co-deposited films that improve carrier transport can be fabricated (Fig. 8(b)). The third molecules collide with the C₆₀ and H₂Pc and decrease the density of crystalline nucleation sites on the surface and promote the crystallization/phase separation process. The third molecules are not left in the co-deposited films at elevated substrate temperatures.

Figure 8(c) and (d) shows cross-sectional SEM images of a C₆₀:H₂Pc co-deposited film. Without the third molecules, an amorphous smooth cross-section was observed for the molecular-level mixture of C₆₀ and H₂Pc (Fig. 8(c)). On the other hand, with the third molecules, a columnar structure of phase-separated and crystallized material (Fig. 8(d)) similar to the ideal vertical superlattice (Fig. 8(a)) was formed. The improved crystallinity produced by introducing the third molecules was confirmed by UV–vis absorption spectra and X-ray diffraction analyses. Photocurrent enhancement was observed, particularly for relatively thick (>400 nm) co-deposited films having greater light absorption (see Section IV.B). A striking enhancement in photocurrent generation is achieved in organic solar cells without exception, based on a variety of co-deposited films such as H₂Pc:C₆₀, PbPc:C₆₀, AlClPc:C₆₀, and rubrene:C₆₀. More than 10 kinds of low vapor pressure liquids such as polydimethylsiloxane (PDMS) and alkyldiphenylether (ADE) can be used as third molecules. Since ADE is a typical diffusion pump oil, the present effects can often be observed for co-deposition using a chamber evacuated by a diffusion pump (see Section IV.B). We believe that this method is generally applicable for growing high-quality phase-separated/crystalline co-deposited films for vacuum-deposited small-molecular-type organic solar cells.

There are reports of the successful construction of such vertical superlattices. A pseudo-vertical superlattice structure was fabricated by combining a new type of C₆₀ derivative (SIMEF, Fig. 3) with benzoporphyrin (BP, Fig. 3) [14]. BP is formed by thermal conversion from a precursor. Since fabrication by wet processing was adopted, the phase separation occurred in the presence of a solvent. After removal of the C₆₀ derivative (SIMEF) by washing in toluene, the desired interdigitated structure was clearly visible (Fig. 8(e)). The round columns of crystalline BP with diameters of around 20 nm stand almost vertically. This structure allows excitons to dissociate within their diffusion length and ostensibly inhibits the recombination of photogenerated electrons and holes. In 2013, Mitsubishi Chemical Co. Ltd succeeded in obtaining a power-conversion efficiency of 12% with this system. A commercial version using this system is illustrated in Fig. 2(a) and (b), which show a flexible see-through organic solar cell module and a demonstration building using these modules as exterior walls and windows [17]. This is a clear demonstration of the ideal vertical structure (Fig. 8(a)) showing practical conversion efficiency.

A. Seven-Nines Purification

1. Single-Crystal Sublimation

First, we focused on the high purification of organic semiconductors. Conventional p–i–n cells (Section II.D) [7, 8] incorporating a quasi-vertical superlattice (Section III.B) [16] were used to evaluate the effects of high purification.

Based on an analogy with inorganic Si, which is usually purified to eleven-nines (11N), the purity of organic semiconductors needs to at least reach the sub-ppm level in order to draw out their essential nature. Based on the aforementioned consideration, a more rigorous purification method was applied to organic semiconductors. Conventionally, organic semiconductors are purified by the “train sublimation” method under vacuum [18] and the purified samples are obtained as a powder. Alternatively, when the sublimation is performed at 1 atm, the purified samples are obtained as single crystals that are of extremely high purity due to gas convection [19].

Figure 12 shows a photograph of C₆₀ (i), C₇₀ (ii), H₂Pc (iii), 6T (iv) crystals purified by single-crystal sublimation [19]. Crystal growth was performed in a quartz tube surrounded by a three-zone furnace system (Epitech Co., Ltd) under flowing N₂ at 1 atm. For the case of C₆₀, the C₆₀ sample was set at 720 °C, and single crystals with sizes exceeding 2 mm × 2 mm were grown at around 500 °C. X-ray diffraction of the obtained crystals showed precise agreement with the reported crystal structure of C₆₀. The obtained C₆₀ crystals were used in the next single-crystal sublimation process.

2. One-Micrometer-Thick Co-deposition Cells

Highly purified organic semiconductors produced by single-crystal sublimation were incorporated in p–i–n cells (Fig. 7(a)) (Section III.D, Fig. 13(a)). A p-type H₂Pc layer (20 nm), a co-deposited C₆₀:H₂Pc i-interlayer, and an n-type layer of naphthalene tetracarboxylic anhydride (NTCDA) were successively deposited by vacuum evaporation at 1 × 10⁻³ Pa using a diffusion pump (VPC-260, ULVAC) onto an indium tin oxide (ITO) glass substrate pretreated in an air plasma. The thick NTCDA layer (600 nm) also acts as a transparent protection layer that prevents electrical shorting of the cells due to metal migration into the organic film during metal deposition [20, 21]. The co-deposition was performed on a substrate heated to +80 °C. The optimized C₆₀:H₂Pc ratio was 1.13:1. ADE, which acted as a co-evaporant third molecule (Section VI.B), was automatically introduced from the diffusion pump. A phase-separated/crystalline nanostructure (Fig. 8(d)) was confirmed to be formed for the present C₆₀:H₂Pc co-deposited film.

Figure 13(b) shows the current–voltage (J–V) characteristics of the cells in Fig. 13(a) with co-deposited layer thicknesses, X, of 250, 600, 960 nm, and 1.2 µm, incorporating a C₆₀ sample purified three times by single-crystal sublimation. Figure 14 shows the dependence of the fill-factor (FF) and the short-circuit photocurrent density (J_sc) on X. Surprisingly, FF hardly decreases even for an extremely thick i-co-deposited layer of 1.2 µm (open circles). Simultaneously, J_sc increases with X and reaches a maximum value of 19.1 mA/cm². On the contrary, when the C₆₀ is purified by conventional train sublimation under vacuum, FF monotonically decreases with co-deposited layer thickness (Fig. 14(a), open squares) [22]. At X = 960 nm, a J_sc value of 18.3 mA/cm² and a conversion efficiency of 5.3% were observed [23, 24].¹ The internal quantum efficiency reaches around 90% in the region from 400 to 700 nm for the X = 960 nm cell (Fig. 15(a)).

Figure 15(b) shows the spectral dependence of the absorption ratio of the cells. For a thin C₆₀:H₂Pc layer (X = 180 nm, curve A), a large portion of the visible light, especially around 500 nm, cannot be absorbed due to the low absorbance of C₆₀. For an extremely thick C₆₀:H₂Pc layer (X = 960 nm, curve C), 95% of the visible light from 300 to 800 nm is absorbed. Figure 15(c) shows photographs of the cells with X = 180 nm (left) and 960 nm (right). For X = 180 nm, the cell color is a transparent green, that is, a large portion of the visible light is not absorbed and therefore cannot be utilized. For X = 960 nm, the cell color is an opaque dark brown, that is, almost all of the visible light is absorbed. The most important feature of the present cells is the incorporation of an extremely thick (1 µm) C₆₀:H₂Pc co-deposited layer into the cell without decreasing FF. This allows the utilization of the entire visible region of solar light.

To evaluate the purity of the C₆₀ crystals at the ppm level, secondary ion mass spectroscopy with a Cs⁺ ion source (SIMS, ULVAC-PHI, 6650M) was used. Figure 16 shows the negative ion mass spectrum (i) and the depth profile (ii). Most of the peaks are assigned to carbon (C₁, C₂, etc. and their isotopes). As impurities, only oxygen (O) and hydrogen (H) were detected. The intensity of 10¹ in the depth profile corresponds to the detection limit for the elements. For O atoms, the detection limit corresponds to a concentration of about 8 × 10¹⁵ atoms/cm³ [25]. Taking this value into account, we conclude that the purity of the C₆₀ has reached at least seven-nines (99.99999%, 7N). A purity of 7N was also confirmed by temperature-programmed desorption mass spectroscopy. The main impurity is revealed to be oxygen. C₆₀ molecules interacting with oxygen seem to be the main impurities. Oxidized C₆₀ (C₆₀O_x) has been reported to act as an electron trap [26]. It is probable that the absence of C₆₀O_x traps greatly enhances the electron diffusion length (L = (Dτ)^1/2; D: diffusion coefficient, τ: electron lifetime) by increasing the lifetime (τ).

The overall results in this section suggest that both high purification and phase separation are necessary to fabricate 1-µm-thick C₆₀:H₂Pc co-deposited cells.

B. pn-Control by Doping

2. Method of ppm-Level Doping

Organic semiconductor samples of at least 7N purity were used (Fig. 12). MoO₃ (Alfa Aesar, 99.9995%) and V₂O₅ (Aldrich, 99.99%), and Cs₂CO₃ (Aldrich, 99.995%) were used as dopants for acceptors and donors, respectively (Fig. 17(a)).

Multiple component co-evaporation techniques were employed to simultaneously evaporate organic semiconductors and dopants. In the case of doping into single organic semiconductor films, a two-component co-evaporation technique was employed. In the case of doping into C₆₀:H₂Pc and C₆₀:6T co-deposited films, a three-component co-evaporation technique (Fig. 17(a)) was employed. An oil-free vacuum evaporator (Epitech Inc., ET300-6E-HK) was used for co-evaporation on ITO glass substrates at a chamber pressure of 10⁻⁵ Pa.

Precise monitoring of the deposition rate using a QCM equipped with a computer monitoring system (ULVAC, CRTM-6000G/Depoview) allowed us to introduce the dopants down to a very low concentration of 9 ppm by volume [44]. Figure 17(b) shows an example of the total thickness signal for the QCM versus time relationship as monitored by a PC display for 50 ppm MoO₃ doping. There was a very slow cyclical fluctuation in the material due to temperature fluctuations in the cooling water for the QCM caused by on/off cycling of the chiller. However, a reproducible increase in the baseline (dark grey line), which was only observed during MoO₃ evaporation for a prolonged timescale of 3500 s was observed (1.8 × 10⁻⁶ nm/s). The evaporation rate of the organic semiconductors was maintained at 0.2 nm/s. Therefore, a doping concentration of 9 ppm by volume can be obtained (1.8 × 10⁻⁶/0.2 = 9 × 10⁻⁶). Doping concentrations of 1 ppm were realized by reducing the dopant evaporation rate using rotating disks having slits with aperture ratios of 1/10 (Fig. 17(c)).

The Fermi level (E_F) of the 100-nm-thick organic semiconductor films was measured using a Kelvin vibrating capacitor apparatus (Riken-Keiki, FAC-1). Both the evaporation chamber and the Kelvin probe were built into a glove box (Miwa, DBO-1.5) purged with N₂ gas (O₂ < 0.2 ppm, H₂O < 0.5 ppm). During the film deposition and the E_F and photovoltaic measurements, none of the organic films were exposed to air at any time. Removal of the influence of O₂ is indispensable for obtaining accurate E_F measurements of organic semiconductor films. The E_F values were easily perturbed if the organic films were exposed to air even once, especially by the ingress of O₂ into the films, and then reproducible results could not really be obtained.

3. pn-Control of Single C₆₀ Films

In Fig. 18, energy diagrams of the C₆₀ (left side) and MoO₃ (right side) films are shown. The MoO₃ showed a remarkably positive value of E_F at 6.69 eV, which is more positive than the upper edge of the valence band (VB) of C₆₀ (6.4 eV), as determined by X-ray photoelectron spectroscopy [45, 46]. The value of E_F for nondoped C₆₀ (middle line) is located at 4.6 eV, near the lower edge of the conduction band (CB), suggesting that this film is n-type in nature. When MoO₃ was doped at a concentration of 3000 ppm, the value of E_F shifted toward the positive direction and reached 5.88 eV, which is close to the upper edge of the VB (6.4 eV). This result strongly suggests that MoO₃-doped C₆₀ is p-type.

To investigate the kinds of interactions that occur between C₆₀ and MoO₃, the absorption spectra of a co-deposited film with a ratio of 1:1 were obtained. Though single films of MoO₃ and C₆₀ are transparent and weak-yellowish transparent, respectively, a new strong absorption from 500 to 1800 nm appeared for the co-deposited MoO₃:C₆₀ (1:1) film and the film color changed to black (Fig. 19(a), upper photograph). This new absorption can be attributed to CT absorption between C₆₀ and MoO₃. Absorption spectrum of the CT band appeared in Ref. [47]. Based on the energy diagram (Fig. 18), it is reasonable to infer that MoO₃ extracts electrons from the VB of C₆₀. The middle figure in Fig. 19(a) shows the mechanism of p-type C₆₀ formation. A CT complex, that is, C₆₀⁺–MoO₃⁻, is formed. Here, the negative charge on the MoO₃⁻ group can be regarded as a spatially fixed ion, that is, an ionized acceptor. On the other hand, the positive charge on C₆₀⁺ can be liberated from the negative charge on the MoO₃⁻ by heat energy at room temperature and can migrate into the C₆₀ film and act as a free hole in the VB of C₆₀. This increase in hole concentration causes the large positive shift of E_F that is observed (Fig. 18). This is a process similar to the formation of free holes in p-type silicon (Fig. 19(a), lower). V₂O₅ was also confirmed to act as an acceptor in C₆₀.

For the 3000 ppm Cs₂CO₃-doped C₆₀ film, the value of E_F shifted negatively to 4.40 eV, which is close to the lower edge of the CB of C₆₀ (4.0 eV) [45]. A thick co-deposited film of C₆₀:Cs₂CO₃ in the ratio 10:1 changed color to dark grey (Fig. 19(b), upper photograph), that is, it showed a new broad CT absorption. Since the work function of Cs₂CO₃ (2.96 eV) is more negative than the CB of C₆₀ (4.0 eV), it is reasonable that Cs₂CO₃ donates an electron to C₆₀ and forms a CT complex, that is, C₆₀⁻–Cs₂CO₃⁺. The formation of n-C₆₀ by Cs₂CO₃ doping is caused by the opposite mechanism to MoO₃ doping (Fig. 19). The donor ability of Cs₂CO₃ did not disappear even after exposure to air.

4. pn-Homojunction Formation in Single C₆₀ Films

Since both p- and n-type C₆₀ were formed, we tried to fabricate pn-homojunctions in single C₆₀ films [48, 49]. Three types of cells with different thickness combinations of p- and n-doped layers, that is, 250/750 nm (a), 500/500 nm (b), and 750/250 nm (c) were fabricated (Fig. 20). The total thickness of the C₆₀ was maintained at 1 µm for all cells. Figure 21 shows the action spectra for the three types of cells. Under irradiation onto the ITO electrode (Fig. 20(a), hν(ITO)) and with the homojunction located at the ITO side (Figs. 3.20(a) and 3.21(a, curve A)), photocurrent mainly appeared between 400 and 500 nm, which corresponds to the visible absorption region of the C₆₀ film (black curve). For the homojunction at the center of the cell (Figs. 3.20(b) and 3.21(a, curve B)), the photocurrent decreased and shifted to longer wavelength and the main peak was located at the edge of the C₆₀ absorption. Finally, for the homojunction on the Ag side (Figs. 3.20(c) and 3.21(a, curve C)), the low magnitude photocurrent shifted to a wavelength far longer than 500 nm, where there is little C₆₀ absorption. The observed systematic change in the shape of the action spectra with respect to the distance of the homojunction from the light-irradiated electrode can be attributed to the so-called “masking effect.” This means that photocarrier generation occurs mainly in the neighboring regions of the p/n-doped interface (active zone) and that, by retracting this homojunction from the light-irradiated ITO surface, a dead layer is gradually grown in front of this active zone.

Under light irradiation onto the Ag electrode (Fig. 20, hν(Ag)), the homojunction approaches the illuminated electrode in the order of cells (a), (b), and (c). In this case, completely the reverse tendency, namely, an increase in the magnitude of the photocurrent and a shift toward shorter wavelength of the action spectra, were observed (Fig. 21(b), curves C, B, and A). This means that the dead layer between the active zone and the illuminated Ag electrode gradually disappeared. Apparently, the photoactive zone moves together with the homojunction. Even if the Ca donors were substituted with Cs₂CO₃ donors, fundamentally the same result was obtained. Since the width of the depletion layer pn-homojunction at the present doping concentrations of 5000 ppm MoO₃/5000 ppm Cs₂CO₃ is calculated to be only 29 nm, which is far smaller than the total cell thickness of 1 µm, a strong masking effect (Fig. 21) was observed.

In Fig. 22, the energy structure of a pn-homojunction (doping concentration: 3000 ppm MoO₃/500 ppm Cs₂CO₃) as measured by Kelvin band-mapping (see Section IV.D) is shown. Since there is a significant difference in E_F, a built-in potential can be created by contacting the MoO₃- and the Cs₂CO₃-doped C₆₀ films (Fig. 18, left side). As a result, a pn-homojunction is formed. The observed direction of the photovoltage, whereby ITO/MoO₃ is positive and BCP/Ag is negative, is consistent with this energy structure. The present results clearly show that pn-homojunctions were fabricated in the single C₆₀ films by doping alone. In other words, the photovoltaic properties of organic semiconductor films could be intentionally designed by doping. We also confirmed the formation of pn-homojunctions for metal-free phthalocyanine (H₂Pc) [50].

6. Band-Mapping by Kelvin Probe

The concentrations of carriers created by doping can be evaluated by using the Kelvin vibrating capacitor method [44, 51, 52]. Figure 23 shows the principle of band-mapping by Kelvin probe. When p-doped organic semiconductors are in contact with ITO electrodes, the E_F values are aligned. Accordingly, the vacuum level (E_VAC) is bent upward and the value of the work function, which is defined as the difference between E_VAC and E_F (double arrows), changes with the thickness of the films. Thus, the band-bending can be directly mapped (Fig. 23, lower) by measuring the work function using a Kelvin probe for changing thicknesses of doped films (Fig. 23, middle). Since the band-bending gives the depletion layer width (W_dep) and the built-in potential (V_bi), the carrier concentration (N) can be obtained by using the following Eq. (3.2).

2

Here, ϵ, ϵ₀, and e are the relative dielectric constant, the dielectric constant of a vacuum, and the elementary charge.

Figure 24 shows the dependences of the work functions of the doped C₆₀ films on the film thicknesses. In the case of Cs₂CO₃ doping, the work function was shifted toward the negative direction and reached close to the lower edge of the CB (3.9 eV) (triangles). In the case of MoO₃ doping, the work function was shifted toward the positive direction and reached near the upper edge of the VB (6.4 eV) (circles). For both dopants, as the doping concentration increased, the film thickness at which the shift in the work function finished became thinner, and the magnitude of the energy shift became larger. Namely, W_dep decreased and V_bi increased. These band-bendings can be fitted by quadratic curves (Fig. 24, solid lines) based on the Poisson equation, and the values of W_dep and V_bi can be precisely determined. For example, in the cases of the n- and p-type band-bending of C₆₀ films doped with Cs₂CO₃ (500 ppm) and with MoO₃ (5000 ppm), electron and hole concentrations of 2.5 × 10¹⁷ cm⁻³ and 9.6 × 10¹⁷ cm^{−3 2} were obtained, respectively, from W_dep values of 24 and 21 nm and V_bi values of 0.29 and 0.87 V using a ϵ value of 4.4 for C₆₀ [54].

Figure 25(a) shows the dependence of carrier concentration on the doping concentration. When the Cs₂CO₃ doping concentration increased, the electron concentration rapidly increased and reached 10¹⁹ cm⁻³ at a doping concentration of 10,000 ppm. On the other hand, when MoO₃ was used as the dopant, the carrier concentration showed a minimum value, that is, 4.3 × 10¹⁵ cm⁻³ at 500 ppm, and the hole concentration increased and reached 2.7 × 10¹⁸ cm⁻³ at 10,000 ppm. The minimum carrier concentration at 500 ppm with MoO₃ doping suggests that the holes created by MoO₃ doping compensate the inherent n-type nature of C₆₀.

Figure 25(b) shows the doping efficiency, which is defined as the ratio of the induced carrier concentration to the doped molecular concentration. The doping efficiencies of Cs₂CO₃ and MoO₃ doping are about 10% and 3%, respectively. The doping process can be explained by the formation of a CT complex and its subsequent ionization (Fig. 19(middle) and Fig. 26(b)). So, the doping efficiency is expressed by the product of the rates of CT complex formation and ionization. In the case of Cs₂CO₃, since Cs₂CO₃ is a substantial molecule, by assuming that Cs₂CO₃ evaporates molecularly and the rate of CT complex formation with C₆₀ is close to unity,³ the observed doping efficiency of 10% can be regarded as the ionization efficiency, which is significantly smaller than the value of 100% obtained for the donor dopant P in Si at room temperature. The orbitals of the electrons around the positive charge on the ionized donor in the cases of P doping in Si (a) and of Cs₂CO₃ doping in C₆₀ (b) are shown in Fig. 26. Based on the fact that the ϵ value of Si is 11.9, the radius of the orbital of the electron around the positive charge of an ionized donor (P⁺) is calculated to be 3.3 nm. This situation is fundamentally similar to the Wannier exciton (Fig. 4(a)), and the electron is easily liberated from the positive charge by thermal energy at room temperature, and thus the ionization efficiency reaches unity (Fig. 26(a)). The only difference to the exciton is that the positive charge is spatially fixed in the crystal lattice.⁴ In the case of Cs₂CO₃ doping in C₆₀, since the ϵ value of C₆₀ is 4.4, the electron experiences a stronger attractive force from the positive charge. The radius of the Frenkel exciton is only 0.5 nm for C₆₀ (Fig. 4(b)). However, a CT complex is formed by employing Cs₂CO₃ doping, that is, [Cs₂CO₃⁺–C₆₀⁻] and the charges are separated on the neighboring molecules (Fig. 26(b)). This situation is fundamentally the same as the CT exciton (Fig. 5(b)) and the negative charge on C₆₀ can be liberated by thermal energy at room temperature. Thus, significant values for the ionization efficiencies of electrons of about 10% were observed, though they were lower than the case of Si.

In the case of MoO₃, a doping efficiency of 3% was obtained under the assumption that MoO₃ forms the trimer (Mo₃O₉) [55]. Though a similar mechanism to that in Fig. 26(b) can also be applied in this case, in addition, the formation of larger MoO_x clusters lowers the efficiency of CT complex formation, which seems to lower the total doping efficiency.

7. Organic/Metal Ohmic Junction

It is very important to make the two organic/metal contacts in a photovoltaic cell ohmic. When the region in the vicinity of a metal electrode is heavily doped, even if there is a Schottky barrier, its width becomes extremely thin allowing charge-carrier tunneling and, as a result, an ohmic contact is expected to be formed similar to that on heavily p⁺- or n⁺-doped inorganic semiconductors [43, 56, 57]. Here, + means heavily doped. Moreover, an ohmic contact can be formed irrespective of the electrode material used, since tunneling is less dependent on the metal work function, enabling the cell structure to be inverted. This technique would allow flexibility in the design of the cell structure.

As a test case, two-layer cells consisting of C₆₀ and H₂Pc (Fig. 27(a) and (b)) were examined [58]. The electrode materials were ITO and Ag. Heavy doping of the order of 10,000 ppm (1%) and 50,000 ppm (5%) were applied to thin 10 nm regions close to the C₆₀ and H₂Pc/metal interfaces. Figure 27(c) and (d) shows the current–voltage (J–V) characteristics for the cells with heavily doped regions in Fig. 27(a) and (b) (curves A). For the cell in Fig. 27(a), the FFreaches a value of 0.59 and clear rectification characteristics can be seen for the heavily doped regions (curves A). Without the heavily doped regions (curves B), however, FF is only 0.29, and the forward current is significantly suppressed. For the inverted cell (Fig. 27(b)), without heavily doped regions (Fig. 27(d), curves B), photovoltaic and rectification behavior are scarcely perceptible. However, with the heavily doped regions (curves A), the FF recovers, reaching a value of 0.49, and rectification is clearly observed. Clearly, the photovoltaic properties of the cells with thin heavily doped regions at the interfaces are independent of the type of electrode material used. Thus, H₂Pc/C₆₀ cells are invertible using this interfacial heavy-doping technique.

Since electron extraction from C₆₀ to the ITO electrode is crucial for the operation of the inverted cell (Fig. 27(d) (curves A)), we estimated the interfacial energy band structure of ITO/10,000 ppm Cs₂CO₃-doped C₆₀ by Kelvin band-mapping (Fig. 28) (Section IV.B.6). There is a distinct barrier to electrons with a height of 0.34 eV from the CB of C₆₀ to the ITO. However, since the band bends down steeply within 5 nm of the interface, photogenerated electrons can tunnel through this barrier. Heavily doped C₆₀ acts as an n⁺-type semiconductor and makes the n⁺-C₆₀/ITO junction ohmic. Organic/metal ohmic junctions can be fabricated by making tunneling contacts with heavy interfacial doping.

C. pn-Control of D/A Co-deposited Films by Doping

1. pn-Control of Co-deposited Films

Co-deposited films consisting of D/A organic semiconductors are essential in order to generate photocurrent densities of significant magnitude based on the dissociation of excitons by the photoinduced electron transfer process (Figs 3.5(b)–(d), 3.7). So, for the application of this doping technique to organic photovoltaic cells, doping should be performed by regarding a co-deposited film consisting of two kinds of D/A organic semiconductors as a single semiconductor. We believe that the formation of a built-in potential by direct doping in the bulk of co-deposited films, where the generation and transport of photocarriers occurs, has the potential to enhance the efficiency of these cells [52]. Moreover, in these types of cells, a short exciton diffusion length is no longer a factor that limits cell performance.

Figure 29(a) shows the energy diagram for H₂Pc:C₆₀ co-deposited films. With donor (Cs₂CO₃) doping, E_F has shifted from the undoped value of 4.48 eV (middle broken line) to 4.22 eV (upper broken line) and is close to the C₆₀ conduction band (CB_C60). Conversely, for acceptor (V₂O₅) doping, E_F has shifted to 4.95 eV (lower broken line) and is close to the H₂Pc VB (). Clearly, the conduction properties of the co-deposited films are controlled to both n- and p-type. It should be noted that the E_F shift occurs within the band gap of co-deposited H₂Pc and C₆₀ films, that is, in between CB_C60 and . For n-type doping (Fig. 29(b)), free carriers (electrons) are produced in both H₂Pc and C₆₀, and they relax to the CB of C₆₀. E_F is fixed to 4.22 eV and is constant throughout the H₂Pc and C₆₀ (upper broken line). For p-type doping (Fig. 29(b)), holes are produced that relax to the top of the H₂Pc VB. E_F is fixed to 4.95 eV and is constant throughout the H₂Pc and C₆₀ (lower broken line). As a result, the shifts in E_F are within the “bandgap of the co-deposited film.” Thus, control of the doping in the co-deposited film was accomplished [59, 60].

2. Organic/Organic Ohmic Junction

By making an n⁺p⁺ heavily doped double layer, organic/organic ohmic junctions can be fabricated [49, 61]. Figure 30(a) shows an n⁺p⁺-homojunction device fabricated in a C₆₀:6T co-deposited film. MoO₃ and Cs₂CO₃ were heavily doped (50,000 ppm (5%)) for the p⁺ and n⁺-regions, respectively. Obviously, the n⁺p⁺-homojunction showed good ohmic properties (Fig. 30(b)).

3. Tandem Cells Formed by Doping

Tandem organic solar cells were first reported by the author in 1990 [62], and many types of cells that connect two unit cells have been fabricated [63]. The tandem organic solar cell, in which two single p⁺in⁺-homojunctions are connected by a heavily doped n⁺p⁺-ohmic interlayer, were formed in co-deposited films consisting of C₆₀ and sexithiophene (6T) simply by doping [61]. Figure 31(a) and (b) shows the structures of a single p⁺in⁺-homojunction cell and a tandem cell connecting two p⁺in⁺-homojunctions formed in co-deposited C₆₀:6T films. The following cell parameters were obtained: J_sc of 3.0 mA/cm², V_oc of 1.69 V, FF of 0.47, and a conversion efficiency of 2.4% (Fig. 31(c), curves B). Compared with the single cell (Fig. 31(c), curves A), V_oc has doubled from 0.85 to 1.69 V. Obviously, the n⁺p⁺-homojunction in the tandem cell acts as an ohmic interlayer to connect the unit cells as shown in Fig. 30.

The overall energy band diagram of a tandem cell is shown in Fig. 32. In the case of the p⁺in⁺-homojunction cells formed in co-deposited C₆₀:6T films, there are C₆₀/6T heterointerfaces at which photoinduced electron transfer can occur throughout the entirety of the cells. Thus, under light irradiation, electrons and holes are photogenerated in both the front and back cells. Electrons and holes that move toward the n⁺p⁺-ohmic interlayer neutralize each other due to recombination or tunneling. This process is consistent with the observed photovoltaic properties of the present tandem cell (Fig. 31). It should be stressed that the present results proved that built-in potentials, such as homojunctions and tandem junctions, can be intentionally formed in bulk co-deposited films induced by doping alone. Methods for band-mapping in organic pn-homojunctions are explained in Section V.D.

D. Band-Mapping of Organic pn-Homojunctions

In this section, precise band-mapping in organic pn-homojunctions with various balances between the doping concentrations of the p- and n-type layers, that is, (i) a p⁺n⁺-homojunction in 6T:C₆₀ (p⁺-layer doped with MoO₃ at 50,000 ppm and n⁺-layer doped with Cs₂CO₃ at 50,000 ppm), (ii) a pn-homojunction in 6T:C₆₀ (p-layer doped with MoO₃ at 3000 ppm and n⁺-layer doped with Cs₂CO₃ at 500 ppm), and (iii) a pn⁺-homojunction in H₂Pc:C₆₀ (p-layer doped with V₂O₅ at 1000 ppm and n⁺-layer doped with Cs₂CO₃ at 10,000 ppm) was performed [64]. Here, “+” means heavily doped.

In order to obtain the potential profiles of the pn-homojunctions, the film thickness dependence of the work function in both p-type layers on n-type layers and vice versa were measured using a Kelvin probe. The shift in the work function on the p-side can be observed by repeated deposition and measurement of the work function of a p-type layer on a 100-nm-thick n-type layer. Similarly, the shift in the work function for the n-side can be observed by repeated deposition and measurement of the work function of an n-type layer on a 100-nm-thick p-type layer.

3

For the heavily doped p⁺n⁺-homojunction (i), the bulk values of E_F in 100-nm-thick p⁺- and n⁺-layers were determined to be 5.58 and 4.38 eV, respectively. Variations in the work function (φ) with film thickness in the n⁺- and p⁺-type regions of the junction are shown in Fig. 33(a) and (b), respectively. On the n⁺-type side, φ becomes rapidly less positive as we move away from the interface, converging toward a value of 4.44 eV at a depth of about 3 nm, which is close to the value of E_F for 100-nm-thick n⁺-6T:C₆₀ (4.38 eV). This indicates that a space charge layer (SCL) is formed up to 3 nm into the n⁺-type region. On the other hand, on the p⁺-type side, φ becomes more positive as we move away from the interface, converging toward a value of 5.63 eV at a depth of about 15 nm, which is close to the value of E_F for 100-nm-thick p⁺-6T:C₆₀ (5.58 eV). Thus, the SCL is formed up to 15 nm into the p⁺-type region. Since the values toward which φ converge and the values of E_F in the 100 nm films agree well, we conclude that, in thermal equilibrium, full alignment of the Fermi level between the p⁺- and n⁺-type films is realized. Thus, we can say that the depletion layer widths (W_n) for the n⁺-type region and (W_p) for the p⁺-type region are 3 and 15 nm, respectively, that is, the total depletion layer width (W) is 18 nm, and the built-in potential (V_bi), which is the difference between the bulk E_F of n⁺-6T:C₆₀ and that of p⁺-6T:C₆₀, is 1.22 eV. Using these parameters, the ionized dopant concentrations for the n⁺- and p⁺-type regions, N_D⁺ and N_A^–, respectively, can be estimated. According to the standard theory for uniformly doped inorganic semiconductors, the total depletion layer width (W) of a pn-junction is given by Eqs (3.3) and (3.4) [65].

4

Here, ϵ_r, ϵ₀, q, x_p, and x_n denote the relative dielectric constant of the semiconductor, the dielectric permittivity in a vacuum, the elementary charge, and the depletion layer width in the p- and n-type regions, respectively. Furthermore, from Eq. (3.4), since x_n = W_n = 3 nm and x_p = W_p = 15 nm, we obtain the relationship N_D⁺ = 5 N_A^–. Together with V_bi = 1.22 eV, we find from Eq. (3.3) that N_D⁺ = 1.1 × 10¹⁹ cm⁻³ and N_A^– = 2.2 × 10¹⁸ cm⁻³.

The electric potential distribution in the p⁺n⁺-junction can be described by integrating Poisson's equation twice (3.5) and (3.6) with the values of W_n, W_p, N_D⁺, and N_A^–, as shown in Fig. 34 (solid curve).

5

6

Here, the left solid and right solid parts of the curve are the electric potentials in the n⁺- and p⁺-type regions, respectively.

Based on this distribution, we attempted to reproduce the observed work function change (Fig. 33, black squares). When an n⁺-type layer is deposited stepwise on the p⁺-type layer, an energy band is bent not only in the n⁺-type layer but also simultaneously in the underlying p⁺-type layer. Taking account of the fact that the total amounts of charge due to ionized dopants in the n⁺-region (N_D⁺x_n) and the p⁺-region (N_A^–x_p) are identical, by depositing a x_n-thick n⁺-layer on the p⁺-layer, band-bending up to a depth of x_n in the n⁺-layer and up to a depth of x_p in the underlying p⁺-layer develop simultaneously. So, in order to calculate the change in work function, the potential change that occurred in n⁺-layer (left solid double-headed arrow) and that which occurred in the p⁺-layer (left broken double-headed arrow) should be added together as shown in the n⁺-region in Fig. 34. Here, the light grey broken curve is obtained by inverting and shrinking in width the light grey solid curve from x_p to x_n. The calculated work function (Fig. 33(a), black solid curve) clearly reproduces the observed work function (Fig. 33(a), black squares).

Conversely, when a p⁺-type layer was deposited stepwise on the n⁺-type layer, the band is not only bent in the p⁺-type layer but also in the underlying n⁺-type layer. In order to calculate the change in work function, the potential change that occurred in the p⁺-layer (right solid double-headed arrow) and that which occurred in the n⁺-layer (right broken double-headed arrow) should be added together as shown in the p⁺-region in Fig. 34. Here, the dark grey broken curve is obtained by inverting and extending in width the dark grey solid curve from x_n to x_p. Again, the calculated work function (Fig. 33(b), black solid curve) clearly reproduces the observed work function (Fig. 33(b), black squares). Irrespective of whether the n⁺-layer is deposited on the p⁺-layer or the p⁺-layer is deposited on the n⁺-layer, the change in the work function can be reproduced by the calculation. This supports the validity of the values of W_p, N_A^–, W_n, and N_D⁺ obtained and the potential distribution in Fig. 34 (solid curve).

The energy band diagram for the p⁺n⁺-homojunction can be illustrated precisely by turning the potential distribution shown in Fig. 34 (solid curve) upside down (Fig. 35). Since this p⁺n⁺-homojunction has an extremely narrow depletion layer width (W) of 18 nm, it confirms that it behaves as an organic/organic ohmic junction due to tunneling (see Sections IV.C.2 and IV.C.3).

For lightly doped pn-homojunctions (ii), as in the case of the first sample, the simulated curve (Fig. 36, solid curves) reproduces the observed changes in the work function (Fig. 36, black squares), and the value of both W_n and W_p was 22 nm. As a result, N_D⁺ and N_A^– were both determined to be 5.0 × 10¹⁷ (Fig. 36(a) and (b)). These values of N_D⁺ and N_A^– almost accord with those obtained from band-bending measurements made in Schottky junctions, that is, ITO/n-6T:C₆₀ (6.0 × 10¹⁷ cm⁻³) [66] or p-6T:C₆₀ (3.0 × 10¹⁷ cm⁻³). The consistency between the parameters in the pn-homojunction and the corresponding Schottky junctions supports the validity of the obtained ionized dopant concentrations.

For the pn⁺-homojunction (iii), since the doping concentration of the n⁺-region is considerably greater than that of the p-region, it is assumed that in the case of an n⁺p-junction, which forms a one-sided abrupt junction, the SCL spreads predominantly into the p-region. As shown in Fig. 37(a) and (b), as expected, the band-bending mainly occurred in the p-region (Fig. 37(b)). On the other hand, a sharp decrease in the work function was observed to a depth of 2 nm in the n⁺-region. This means that the positive charge within 2 nm of the interface is sufficient to compensate for the negative charge in the depletion layer of the p-region. Conversely, deposition of an n⁺-layer of only 2 nm thickness on the p-layer can induce band-bending of 92 nm width in the underlying p-layer. The values of W_p and W_n that were obtained were 92 and 2 nm (see inset in Fig. 37(a)), respectively. As a result, N_D⁺ and N_A^– were determined to be 1.9 × 10¹⁸ and 4.2 × 10¹⁶ cm⁻³, respectively (Fig. 37(a) and (b)). The values of N_D⁺ and N_A^– accord with those obtained from band-bending measurements made on a Schottky junction, that is, ITO/p-H₂Pc:C₆₀ (4.3 × 10¹⁶ cm⁻³), which has a similar energy structure to an n⁺p-junction (Fig. 37(c)). Again, the simulated curves (Fig. 37(b), solid curves) precisely reproduce the observed changes in work function (Fig. 37(b), open squares).

The energy band diagram in the present n⁺p-homojunction based on the simulated curve is shown in Fig. 37(c). Since the ionized donor concentration is significantly greater than the ionized acceptor concentration, a one-sided abrupt junction, in which the SCL spreads predominantly into the p-type region, is formed.

We have demonstrated precise mapping of the band-bending in three cases, a p⁺n⁺-homojunction (i), a pn-homojunction (ii), and a pn⁺-homojunction (iii), irrespective of the doping concentrations of the n- and p-layers and the balance between the doping concentrations of these layers. It was revealed that the consideration of the change in band-bending in the underlying n-(or p-)type layer when a p-(or n-)type layer is gradually accumulated on it is necessary. The validity of the conventional theory for the SCL suggests that the dopants are spatially fixed even in organic semiconductor films. The present results clearly show that the precise configuration of the built-in electric field in organic photovoltaic cells is fundamentally identical to that in inorganic cells.

E. Doping Sensitization

Sensitization of the ionization for both n- and p-type doping in organic semiconductor co-deposited films was observed [53]. A significantly high ionization efficiency of 97%, comparable to that for silicon (100%), was achieved.

By making accurate estimations of the carrier concentrations generated by impurity doping by means of Kelvin band-mapping, we found evidence of higher doping efficiencies for co-deposited films compared with those for the respective single films. Thus, we tried to clarify the reasons for the difference between the doping efficiencies in co-deposited films and single films of organic semiconductors. As a typical case, we chose metal-free phthalocyanine: fullerene (H₂Pc:C₆₀, H₂Pc:C₇₀) and metal-free phthalocyanine:perylene pigment (H₂Pc:Me-PTC) co-deposited films because of their high capability for exciton dissociation in organic solar cells. Cesium carbonate (Cs₂CO₃) and iron chloride (FeCl₃) were used as donor and acceptor dopants, respectively. The deposition rate for the organic semiconductors was kept to 0.1 nm/s, and those for the dopants were adjusted from 10⁻⁵ to 2 × 10⁻³ nm/s to give doping concentrations from 0.0002 to 0.05 in molecular ratio (MR), that is, 50–10,000 ppm by volume.

Figure 38 shows an energy band diagram of the organic semiconductors and dopants used. Cs₂CO₃ can donate an electron to the CBs (LUMOs) of H₂Pc (3.2 eV) and the fullerenes (3.9 eV) [67], which can be measured by near ultraviolet inverse photoemission spectroscopy [68], and thus acts as a donor since its work function is located at a shallower position (2.96 eV). In practice, by Cs₂CO₃ doping (MR = 0.02), the Fermi level (E_F) shifted toward the positive direction and reached 3.97, 4.00, and 3.48 eV for C₆₀, C₇₀, and H₂Pc, respectively. Both the fullerenes and the H₂Pc became n-type. On the other hand, FeCl₃ can withdraw an electron from the VBs (HOMOs) of H₂Pc (5.1 eV) and Me-PTC (5.4 eV) and act as an acceptor since its work function is located at a deeper position (5.52 eV). By FeCl₃ doping (MR = 0.03), E_F shifted toward the negative direction and reached 4.72 and 5.04 eV for H₂Pc and Me-PTC, respectively. Both H₂Pc and Me-PTC became p-type.

Figure 39(a) shows the film thickness dependence of the work function for Cs₂CO₃-doped H₂Pc:C₆₀ (1:1) co-deposited films (circles) and the respective Cs₂CO₃-doped single films of C₆₀ (diamonds) and H₂Pc (triangles). The shifts in work function due to the band-bending by doping are fitted by quadratic functions based on the Poisson equation (solid curves), which excludes the effect of a dipole layer of a few nanometers thickness in the vicinity of the ITO electrode [69]. Clearly, the depletion layer width (W_dep) in the H₂Pc:C₆₀ co-deposited film has shrunk to 10 nm from 20 to 30 nm in the respective C₆₀ and H₂Pc single films (arrows). This observation is a sign of an increase in the carrier concentration. Concentrations of 3.2 × 10¹⁸, 4.3 × 10¹⁷, and 4.6 × 10¹⁷ cm⁻³ were obtained for the co-deposited film and for the C₆₀ and H₂Pc films, respectively. By mixing C₆₀ and H₂Pc, the number of carriers created by Cs₂CO₃ doping was increased by a factor of around 10.

Essentially the same result was obtained for the H₂Pc:Me-PTC (1:1) system (Fig. 39(b)). FeCl₃ acting as an acceptor was introduced into a H₂Pc:Me-PTC co-deposited film and its respective Me-PTC and H₂Pc single films. Concentrations of 8.7 × 10¹⁸, 5.3 × 10¹⁶, and 6.9 × 10¹⁷ cm⁻³ were obtained for the co-deposited film and the Me-PTC and H₂Pc films, respectively. By mixing Me-PTC and H₂Pc, the number of carriers created by FeCl₃ doping was also increased by a factor of about 100 for Me-PTC.

Figure 40 shows the dependence of the carrier concentration and the doping efficiency on the doping concentration for H₂Pc:C₆₀ (a,b) and H₂Pc:C₇₀ (c,d) co-deposited films and their component films. The doping efficiency is defined by the ratio of the carrier concentration created to the molecular concentration of Cs₂CO₃. Obviously, the carrier concentration is an order of magnitude higher for the co-deposited films (Fig. 40(a) and (c), curves A and D) compared with the respective component single films (Fig. 40(a) and (c), curves B, C, and E). The doping efficiencies of single films of C₆₀, C₇₀, and H₂Pc are around 10% (Fig. 40(b) and (d), diamonds and triangles). In contrast, the doping efficiencies of the co-deposited films of H₂Pc:C₆₀ and H₂Pc:C₇₀ are around 50% (Fig. 40(b) and (d), circles). Thus, we conclude that the doping efficiency was significantly enhanced in the co-deposited films.

Essentially, the same phenomenon was observed for acceptor doping. By combining H₂Pc and Me-PTC, a considerable increase in doping efficiency to 30% was obtained from the 2% and 0.2% efficiencies obtained for H₂Pc and Me-PTC films, respectively (Table II).⁵ Thus, we concluded that the increase in doping efficiency is universal and irrespective of whether the doping is donor or acceptor doping.

The doping process for Cs₂CO₃ in H₂Pc can be explained by a CT complex, that is, the formation of H₂Pc⁻–Cs₂CO₃⁺ and its ionization. Here, the positive charge on the Cs₂CO₃⁺ is a spatially fixed positive ion, that is, an ionized donor. The negative charge on the H₂Pc is liberated by thermal energy and acts as a free electron in the CB (Fig. 41(a)). Thus, the doping efficiency can be expressed by the product of the rates of the CT complex formation and ionization. Since Cs₂CO₃ exists as a molecule having a structure of CsO(CO)OCs, it is reasonable to assume that Cs₂CO₃ evaporates molecularly and the rate of the CT complex formation with H₂Pc, C₆₀ and C₆₀:H₂Pc films is close to unity. Thus, we consider that the observed doping efficiency represents the ionization efficiency.

We propose a charge separation superlattice model to explain the enhancement in ionization rate. Figure 41(a) shows the energy structures of H₂Pc and C₆₀ single films before contact. The lower edge of the CB was determined by inverse photoelectron spectroscopy [68]. The activation energy (ΔE_D) of 0.12 eV was determined from the ionization rate of 10%.⁶ Before contact, 10% of the electrons are excited to the CB, and the other electrons are captured by donor levels (Fig. 41(a)).

By making contact with C₆₀, since the electron transfer from the CB of H₂Pc to that of C₆₀ (Fig. 41(a)) occurs accompanied with energetic relaxation (0.7 eV), the electron concentration in H₂Pc decreases. Due to the shift in equilibrium, electron liberation from the donor levels accelerates, that is, the ionization rate for donors increases only in the H₂Pc region. Taking the molecular ratio of the H₂Pc:C₆₀ co-deposited film (1:1) into consideration, the ionization rate of the H₂Pc region in the co-deposited film can be estimated to be 100%, while that in the C₆₀ region is unchanged (10%), since the observed ionization rate of 55% for the H₂Pc:C₆₀ co-deposited film can be obtained by calculation, that is, 100% × 0.5 + 10% × 0.5 = 55%. Thus, we concluded that for the H₂Pc region, the ionization rate increased from 10% to 100%, an increase by a factor of 10 (Table II). The ionization rate of 100% supports the assumption that the rate of formation of the CT complex is unity.

Figure 41(b) and (c) shows the three-dimensional and cross-sectional energy structures of charge separating H₂Pc/C₆₀ superlattices after contact, respectively. Due to the faster liberation of electrons from the donors, all of the donors in the H₂Pc region of the co-deposited film positively ionize, and the energy band of H₂Pc is bent to a depth of 3.3 nm.⁷ In this model, the H₂Pc regions act as electron-supplying layers to the C₆₀ regions. The C₆₀ regions act as electron transporting highways. Based on this model, we expect that the total ionization rate in H₂Pc:C₆₀ co-deposited films would increase further by increasing the H₂Pc ratio.

Figure 42(a) and (b) shows the Kelvin band-mapping results and observed ionization rates for different ratios of H₂Pc in H₂Pc:C₆₀ co-deposited films doped with Cs₂CO₃. As the H₂Pc ratio increases, the ionization rate increases proportionally, with good agreement to the calculated line (broken light grey line, Fig. 42(b)),⁸ and reached 97% at a H₂Pc:C₆₀ ratio of 99:1.

Interestingly, the positions of E_F after band-bending systematically shift in the negative direction with decreasing C₆₀ (Fig. 42(a)). This is a sign that there is a more concentrated accumulation of liberated electrons in the thinner C₆₀ regions, which lifts E_F by 0.38 eV for H₂Pc:C₆₀ (99:1) (compare the left and right plots in Fig. 41(c)). Since there should be a vast number of free electrons in the C₆₀ CB under E_F, we assume that a confined electron gas exists on the C₆₀ sides of the H₂Pc/C₆₀ interfaces, similar to that reported for inorganic AlGaAs/GaAs superlattices [70]. This systematic observation also supports the proposed charge separation superlattice model.

For acceptor doping, the opposite mechanism occurs (Fig. 41(d)). Being in contact with H₂Pc, the holes in the VB of Me-PTC energetically relax (0.3 eV) to the VB of H₂Pc (dark grey arrows) and their concentration in Me-PTC decreases. As a result, there is faster liberation of holes from the acceptors. For the Me-PTC region, the ionization rate increases from 0.2% to 60%, a 300-fold increase (Table II).

Direct ionization processes might occur when the ionized donor is located just at the C₆₀/H₂Pc molecular interface (Fig. 43). For donors in the single C₆₀ and H₂Pc regions, the activation energy, that is, the binding energy of an electron around a positively ionized donor (Cs₂CO₃⁺) (ΔE_D) is 0.12 eV (Fig. 41(a)),⁶ which corresponds to an electron orbital radius of 1.6 nm (Fig. 43, circles A).⁹ This situation resembles the CT exciton. The only difference is that the positive charge of the ionized donor is fixed. For donors just on the H₂Pc side of the C₆₀/H₂Pc molecular interface (Fig. 43, center), a cascade-like second electron transfer from the LUMO of the H₂Pc molecule to that of the C₆₀ molecule occurs (see Fig. 38). Compared to the binding energy (ΔE_D) of 0.12 eV, the relaxation energy of 0.58 eV¹⁰ is sufficient to liberate electrons by increasing the electron orbital radius to over 6.5 nm, which has a binding energy comparable to the thermal energy at room temperature (kT = 0.026 eV) (Fig. 43, semicircle B).¹¹ This situation resembles the Wannier exciton. It should be noted that a similar electron transfer from H₂Pc (D) to C₆₀ (A) (Fig. 38) is utilized in organic solar cells to increase exciton dissociation, that is, so-called donor/acceptor (D/A) sensitization (Section III.B). Thus, the present doping sensitization phenomenon can be regarded as analogous to D/A sensitization for dopant ionization.

F. ppm-Doping Effects

The effects of doping reaching the 1 ppm level in organic photovoltaic cells were clarified using simple n⁺p-homojunctions [71]. With doping from 0 to 10 ppm, the FF increased due to the appearance of majority carriers. From 10 to 100 ppm, the photocurrent density increased due to the increase in the built-in potential, that is, the formation of an n⁺p-homojunction. The photocurrent was increased by a factor of 1.3 by directly doping the photoactive co-deposited layer with acceptors at 100 ppm.

We believe that direct doping of the photoactive co-deposited layer where the generation and transport of photocarriers occur has the potential to enhance the efficiency of these organic solar cells. Based on this consideration, we attempted to introduce a pn-homojunction into a C₆₀:6T co-deposited film. For the simplest cell, we adopted an n⁺p-homojunction that has a one-sided abrupt junction (Fig. 44(a)) (Section V.D).

The Cs₂CO₃ doping concentration of the n⁺-layer was kept constant at 10,000 ppm. The FeCl₃ doping concentrations of the p-layer were varied, that is, 0, 1, 10, 100, and 1000 ppm. The doping concentrations of 10 and 1 ppm were realized by reducing the dopant evaporation rate using rotating disks having slits with aperture ratios of 1/10 and 1/100, respectively. The doping concentration of 1 ppm corresponds to a molecular ratio (MR) of 3.7 × 10⁻⁶.

Figure 44(b) shows the energy diagram of the C₆₀:6T co-deposited film. The bandgap in this film is determined by the CB of C₆₀ (CB_C60) and the VB of 6T (VB_6T). In order to fabricate an n⁺-layer, heavy doping with Cs₂CO₃ (10,000 ppm) was performed. For the p-type layers formed on the n⁺-layer, FeCl₃ doping concentrations of 10, 100, and 1000 ppm caused E_F to shift toward VB_6T (5.50 eV) with more positive values of 5.01, 5.15, and 5.21 eV, respectively. These observations suggest that an n⁺p-homojunction is expected to form as the FeCl₃ concentration is gradually increased.

Figure 45 shows the current–voltage characteristics for n⁺p-homojunction cells with p-layer doping concentrations of 0, 1, 10, 100, and 1000 ppm (Fig. 44(a)). The magnitude of the short-circuit photocurrent density (J_sc) increases systematically with FeCl₃ doping of the photoactive C₆₀:6T layer as follows: 3.51 for 0 ppm, 4.02 for 10 ppm, and 4.48 mA/cm² for 100 ppm. Simultaneously, the magnitude of the forward current density increases systematically (black, green, and red broken curves). That is, the cell resistance (R_s) decreases systematically with FeCl₃ doping and the minimum R_s of 2.35 Ω cm² is obtained at 100 ppm (see Fig. 47(b)). J_sc increases by a factor of 1.3 and the maximum efficiency of 1.51% is at 100 ppm. However J_sc decreases and R_s increases with a further increase in doping concentration to 1000 ppm. This confirms that the photocurrent can be increased significantly by directly doping the photoactive co-deposited layer with extremely low concentrations at the ppm level.

Band-bending occurs only on the p-side of the n⁺p-homojunction since the n⁺-layer is heavily doped (10,000 ppm) (see Section V.D). Therefore, the width of the depletion layer (W_dep) formed in the p-type layer depends on the FeCl₃ doping concentration in the layer. Figure 46(a) shows the dependence of the work function on the p-type layer thickness for various doping concentrations. In the cases of 0 and 1 ppm, the observed work function has a constant value of 4.20 eV, which is the same as the Fermi level in the n⁺-layer, that is, no band-bending was observed. This shows that layers with these doping levels act as intrinsic layers. The observation that there was no change in work function at the extremely low concentration of 1 ppm doping (Fig. 46(a)) may be attributed to trap-filling, that is, the doping induced carriers are captured by the traps [72, 73]. In the cases of 10, 100, and 1000 ppm, band-bending was observed. This clearly shows that layers doped at these levels act as p-type layers. The width of the depletion layer (W_dep) and the built-in potential (V_bi) can be determined from the points (shown by arrows) at which band-bending ends. For example, W_dep = 50 nm and V_bi = 1.0 V at a doping concentration of 100 ppm. At 100 ppm, N_h was calculated to be 1.8 × 10¹⁷ cm⁻³. The dependences of R_s, V_bi, and N_h on doping concentration are shown in Fig. 47(b)–(d), respectively. The energy band diagrams of the n⁺p-homojunction cells (Fig. 46(b)) can be found by turning the curves in Fig. 46(a) upside down.

Figure 47(a) shows the dependences of J_sc and the FF on doping concentration. The doping effects can be divided into three regions. From 0 to 10 ppm, both FF and J_sc increase rapidly. From 10 to 100 ppm, J_sc still increases while FF maintains a constant value. From 100 to 1000 ppm, both J_sc and FF decrease.

Figure 47(b) shows the dependences of the FF and the cell resistance (R_s) on doping concentration from 0 to 100 ppm. Clearly, R_s decreases and FF increases from 0 to 10 ppm. Once acceptor doping is performed, holes and electrons inevitably act as majority carriers and minority carriers, respectively. At 10 ppm, the number (N_h) of majority carriers (holes) is 2.3 × 10¹⁶ cm⁻³. Thus, we conclude that the increase in FF at 10 ppm doping is due to the appearance of majority carriers in the p-layer.

Figure 47(c) shows the dependences of J_sc and the built-in potential (V_bi) on doping concentration from 0 to 1000 ppm. Clearly, there is a close relationship between J_sc and V_bi. In particular, there are simultaneous increases in J_sc and V_bi at relatively low doping concentrations from 10 to 100 ppm. An increase in V_bi is also shown in the energy band diagrams for 10 and 100 ppm in Fig. 46(b), and one can see that J_sc increases with increasing V_bi. Based on these considerations, we conclude that the increase in J_sc is due to the increase in V_bi, that is, the formation of an n⁺p-homojunction.

From 100 to 1000 ppm, both FF and J_sc decrease (Fig. 47(a)). Simultaneously, R_s increases considerably from 2.35 to 50 Ω cm². The mobility of the majority carriers (holes), μ_h, can be calculated based on Eq. (3.7)

7

Here, N_h and σ_h are the hole concentration, which is obtained by Kelvin band-mapping (Fig. 46(a)) and the hole conductivity, which is obtained from the cell resistance (σ_h = LR_s⁻¹; L: cell thickness) determined from the forward dark current for the n⁺p-homojunction cells (n⁺-C₆₀:6T/p-C₆₀:6T/MoO₃) (Fig. 45, broken curves), respectively. The mobility, μ_h, was calculated to be 1.5 × 10⁻⁴ and 1.1 × 10⁻⁶ cm²/V/s at 100 and 1000 ppm, respectively. Figure 47(d) shows the dependences of the hole concentration (N_h) and the hole mobility (μ_h) on doping concentration from 10 to 1000 ppm. μ_h decreases at higher concentrations from 100 to 1000 ppm (light grey solid curve). We think this is due to the disturbance of hopping transport of holes by the negatively ionized and neutral accepter dopants, although there still remains the possibility that it is due to a change in morphology on the nanoscale. Harada et al. [74, 75] reported decreasing mobility for higher doping concentrations in C₆₀ films. The decrease in μ_h dominates the cell resistance (R_s) in spite of the increase in N_h, causing R_s to increase and, as a result, FF to decrease. Thus, we conclude that the decrease in FF is due to the decrease in hole mobility (μ_h). As shown in Fig. 46(b), the width of the depletion layer (W_dep) reduces from 50 to 20 nm as the doping concentration is increased from 100 to 1000 ppm. Thus, we conclude that the decrease in J_sc is due to the decrease in W_dep, which acts as the photoactive layer.

In conclusion, by directly doping a photoactive co-deposited layer at the ppm level, we were able to increase both the photocurrent and the FF. The increases in FF and J_sc from 0 to 100 ppm are due to the decrease in R_s following the introduction of majority carriers and the increase in V_bi, respectively. The decreases in FF and J_sc from 100 to 1000 ppm doping are caused by the decrease in mobility of the majority carriers and by the decrease in W_dep, respectively. The most important technical consequence of this doping is possibility for the intentional design of built-in potentials in cells. We think that a new design concept including this doping technology should be developed to realize a high-performance cell.