1. Introduction

Organic photon-to-current conversion devices are essential in various optoelectronic applications, such as organic solar cells (OSCs) [1,2], organic photodiodes (OPDs) [3,4], and organic photodetectors [5–17]. Among these devices, OPDs using narrowband organic semiconductors in the active layers are highly promising candidates for novel stacked-type full-color photodetectors or image sensors [11–17]. In a stacked-type architecture, the light detection area is increased and thanks to the organic layer, the spectral response can be controlled without the use of a top filter. Recently, highly attractive filterless narrowband organic photodetectors based on charge collection narrowing for tuning the internal quantum efficiency (IQE) have been reported [18,19], but this concept might be hardly applied to real full-color imaging due to the significant mismatch between the absorption spectra and the external quantum efficiency (EQE) spectra of the devices as well as the low EQEs. To improve the EQE of OPDs, in addition to the development of narrowband and high absorption organic semi-conductors, other optical manipulations can be brought to the device. In this regards, a gradual concentration changes from n-rich to p-rich materials in the bulk heterojunction (BHJ) active layers can be an interesting approach to enhance photon-to-current conversion efficiencies significantly, as successfully demonstrated for organic solar cells [20].

In this study, we report on filterless green-light-selective OPDs using novel dipolar donor-acceptor (D-A) molecules as narrowband and intense p-type absorbers. In addition, we demonstrate a promising method to minimize the blue absorption under 500 nm, induced by the n-type fullerene C60, by means of optical manipulation of light absorption through material concentration gradient within the active layer. The optimized OPDs reported here showed a specific detectivity $4.37 \times {10^{13}}$ $cmH{z^{1/2}}{W^{ - 1}}$ at an applied reverse bias of 3 V, corresponding to our knowledge among the highest color-selective OPD performances reported so far [3–19]. The specific green-light-selective ability of the OPD is also demonstrated by constructing a hybrid stacked-type full-color photodetector composed of the green filterless narrowband OPD top layer and a bottom silicon photodiode for the additional detection of blue and red light.

2. Results and discussion

The new dipolar D-A molecules 1a and 1b used as p-type molecules in this study are represented in Fig. 1(a). They were synthesized via the synthetic route reported earlier with very high reaction yields [17], and their molecular structures were completely characterized via nuclear magnetic resonance (NMR) spectroscopy, mass spectroscopy, and elemental analysis (see the Experimental section). As compared with our previous studies [17], an electron-rich selenophene group instead of thiophene moiety was used as the π-linker between the donating aryl amino group and the two respective accepting heterocycles to increase the strength of the donar part and the thermal stability in thin films, resulting from the enhanced intermolecular chalcogen bonding strength between the selenophene and the carbonyl moieties [21]. The dipole moment of the molecules in the ground state (μ_ground), as well as the dipole moment difference between the ground state and the excited state (Δμ), calculated by time-dependent density functional theory (TD-DFT) by using the B3LYP method with the 6-311G(2d,p) basis set in Gaussian09 [22] are reported in Table 1. As compared with dipolar molecules having a thiophene as a linker, the μ_ground further increased and the Δμ further decreased in the case of selenophene linker. The resonance parameter c² value, which is a useful parameter to characterize the charge transfer properties of D-A molecules, was also calculated from ${c^2} = 0.5[1 - \varDelta \mu {(4\mu _{tr}^2 + \varDelta {\mu ^2})^{ - 1/2}}]$ where μ_tr is transition dipole moment [23]. The c² values reached 0.44 and 0.45 for 1a and 1b, respectively. These values approaches 0.5, confirming that both polyene-like molecules are close to the cyanine limit and have therefore a high potential to exhibit intense and sharp absorption. Experimentally, the basic absorption characteristics of 1a and 1b are shown in Figs. 1(b) and 1(c), confirming the potential of the two materials to work as efficient green narrowband absorber (absorption coefficient α = 1.47 × 10⁵ cm⁻¹ at λ^abs.film_max = 560 nm, FWHM = 80 nm, absolute photoluminescence quantum yield Φ_film = 0.6%, λ^em.film_max = 687 nm for 1a and α = 1.34 × 10⁵ cm⁻¹ at λ^abs.film_max = 534 nm, FWHM = 79 nm, Φ_film = 0.5%, λ^em.film_max = 630 nm for 2b). Table 1 summarizes the basic optoelectronic properties of the two molecules. Because of the stronger acceptor strength of 1a, its lowest unoccupied molecular orbital (LUMO) level is lower than that of 1b. The highest occupied molecular orbital (HOMO) level is in contrast less affected by the change of the acceptor strength resulting in a smaller band gap for 1a and to the observed red-shift in absorption as compared with 1b. The two respective p-type D-A molecules were then thermally co-evaporated with C60 under high vacuum (< 10⁻⁷ Torr) to form the BHJ active layers [20]. The absorption profiles of both BHJ blends are represented in Figs. 1(b) and 1(c), respectively (α = 0.82 × 10⁵ cm⁻¹ at λ^abs.film_max = 552 nm, FWHM = 84 nm, Φ_film = 0.5%, λ^em.film_max = 828 nm for 1a:C60 and α = 0.58 × 10⁵ cm⁻¹ at λ^abs.film_max = 526 nm, FWHM = 80 nm, Φ_film = 0.4%, λ^em.film_max = 832 nm for 1b:C60).

 

Fig. 1. (a) Molecular structures of 1a, 1b and C60. Absorption coefficient (solid line) and photoluminescence (dashed line) of (b) 1a and 1a:C60 and (c) 1b and 1b:C60.

Download Full Size | PDF

Table 1. Molecular optoelectronic properties of 1a and 1b

View Table | View all tables in this article

The optoelectronic potential of 1a and 1b was evaluated by fabricating inverted-structure green-light-selective OPDs consisting of indium tin oxide (ITO)-coated glass substrates, the respective BHJs of 1a or 1b and C60 (volume ratio, 1:1), a hole-extraction layer of molybdenum oxides (MoO_x), and an ITO contact (see the experimental section and Fig. 2(a) inset). It is to be noted that transparent ITO electrodes were used on both sides of the OPDs to assess for the green-light-selectivity. Representative current density-voltage (J-V) curves under dark conditions of the OPDs having a 85-nm-thick BHJs are presented in Fig. 2(a) and the important device parameters are reported in Table 2. Among the principal characteristics, at the reverse bias of 3 V used in this study, the diodes exhibited rather low dark current (J_d) of 11.1 nA/cm² for 1a:C60 and 62.2 nA/cm² for 1b:C60 and remarkably high selectivity to the external green light, displaying FWHMs as narrow as 97 nm for 1a:C60 and 102 nm for 1b:C60, which is attributed to the narrow absorption profile of the BHJs (see Figs. 1(b) and 1(c)). Figure 2(b) shows the representative EQE curves of the OPDs driven at the 3 V bias (left-hand side) and the plots of the EQE as a function of applied reverse electric-field (right-hand side). While rather low EQEs of 21.1% for 1a:C60 and 24.0% for 1b:C60 were obtained without any external reverse bias (0 V), the efficiency drastically increased at the reverse bias of 3 V reaching 51.8% for 1a:C60 and 46.7% for 1b:C60. As shown in Table 2, these enhanced EQE values upon applied bias are the results of better charge-separation efficiency (η_CS) and charge-collection efficiency (η_CC), attributed to the minimization of charge recombination loss under electric field-assisted charge carrier mobilities [16,17]. It is noted that IQE was estimated from EQE based on measured absorptance (η_A) of the BHJ in OPD, and η_CS and η_CC were estimated from IQE and field-dependent photoluminescence quenching efficiency [24].

 

Fig. 2. (a) Representative current density-voltage (J-V) curves under dark condition. The inset shows the device structure. (b) EQE spectra at the reverse bias of 3 V. The right-hand figure shows the maximum EQE values as a function of applied electric fields. GIWAXS spectra of (d) 1a:C60 and (d) 1b:C60. (e) Internal quantum efficiency (IQE), charge separation efficiency (CSE), and charge collection efficiency (CCE) as a function of applied voltage. (f) Calculated maximum EQE values at various thicknesses of an organic layer.

Download Full Size | PDF

Table 2. Numerical properties of inverted-structure green-light-selective OPDs with the BHJs (1a or 1b:C60 = volume ratio 1:1, total thickness 85 nm).

View Table | View all tables in this article

To gain information on molecular stacking, grazing incidence wide angle X-ray scattering (GIWAXS) experiments were carried out and the diffraction patterns of the blends 1a:C60 and 1b:C60 are shown in Figs. 2(c) and 2(d), respectively. The blends are amorphous but a degree of local arrangement nonetheless exists for both systems. For 1a:C60, local periodicity is observed preferentially in the z-direction, parallel to the surface. For 1b:C60, the periodicity is not limited to the z-direction, but also distributed in the other directions. These differences might originate from the different molecular structures of 1a and 1b. The larger acceptor unit of 1a might favour the preferred local alignment of blend domains with the surface, while the smaller molecule size of 1b might lead to a more random distribution of locally stacked domains. In terms of the efficiencies of the OPDs, which can be described by the expression $EQE = {\eta _A} \times IQE = {\eta _A} \times {\eta _{CS}} \times {\eta _{CC}}$, the main difference between the two p-type D-A molecules is the higher IQE obtained with 1b:C60 at low reverse biases (< 2 V) as shown in Fig. 2(e) and Table 2. This better IQE performance is mainly attributed to the higher η_CS and η_CC values thanks to the higher charge-carrier mobilities in the 1b:C60 blend. One sees that in the blend 1b:C60, an even distribution of locally arranged domains in all directions positively impacts charge-carrier mobilities for the better IQE at the low reverse biases. For larger bias, these small differences in stacking no longer influence the transport properties and the IQEs have the same levels (83.6% for 1a:C60 and 83.4% for 1b:C60 at 3 V). The EQEs of the OPD composed of 1a:C60 gets therefore higher than those composed of 1b:C60 over the reverse bias of 2 V, thanks to the higher η_A (61.9% for 1a:C60 and 56.0% for 1b:C60). On the other hand, the J_d increased significantly in the OPD composed of 1b:C60 under the reverse biases, not only due to the larger portion of stacked domains but also due to the lower hole–injection barrier from the cathode ITO in the 1b:C60 blend composition, which has a negative impact on the OPD performance [15,17].

Finally, we calculated the EQE of the device as function of the thickness of organic layer from the relation $EQE = {\eta _A} \times IQE$. The absorptance was calculated from optical simulation using transfer matrix method [25]. For the calculation, the refractive index of each layer in the device was measured by ellipsometry (see the Experimental Section). The thickness-dependent IQE at the reverse bias of 3 V was obtained from the voltage-dependent IQE shown in Fig. 2(e) according to the relation ${d_{thickness}}[\mbox{nm}] = 3 \times {10^5}/{V_{bias}}[\mbox{V}]$. Figure 2(f) shows that the maximum EQE in the device with the material 1a:C60 is much higher than that with 1b:C60 within the thickness of 190 nm. We experimentally confirmed the EQE in OPDs with the optimal organic layer thickness. Since the IQE in the device predominantly depends on the charge separation efficiency, the EQE estimation in the device with different BHJ thickness by simple calculation is valid.

As the new D-A molecule 1a gave improved OPD characteristics at the reverse bias of 3 V, especially a better η_A (or EQE) and a lower J_d, its green-light-selective OPD characteristics were further investigated by means of optical manipulation of light absorption within the active layers. To reduce the optical loss, an anti-reflective coated glass was used for encapsulation and an additional AlO_x layer was coated on the upper ITO electrode to reduce the light reflection at the air/ITO interface (see Fig. 3(a) inset). Figure 3(a) shows the periodical fluctuation of the absorptance as a function of the AlO_x thickness, obtained by optical calculation for the wavelengths of 560 nm (filled circles) and 450 nm (open circles), respectively. At 560 nm, the absorptance reaches a first maximum at a AlO_x thickness of 45 nm. For this condition, with a BHJ thickness of 130 nm, the EQE reached 70% at 560 nm (see black filled circles in Fig. 3(b)). The drawback of the AlO_x layer is also to induce an increase of the absorptance at 450 nm, leading to an EQE of about 25% at this wavelength.

 

Fig. 3. (a) Calculated absorptance at 560 and 450 nm as a function of the thickness (d_opt) of the AlO_x layer. The inset shows the device structure. (b) Measured EQE spectrum with homogeneous layer (1a:C60 ratio of 1:1) and inhomogeneous layer (1a:C60 ratio of 1.1:1, 1.9:1, 1.5:1) as a bulk-heterojunction layer.

Download Full Size | PDF

We introduced three ratios of 1a:C60 in the BHJ, as illustrated schematically in Fig. 3(b) inset. By choosing an inhomogeneous BHJ, the EQE at 450 nm was reduced to 17% without affecting the maximum EQE of 70% at 560 nm. The p-rich composition near the middle of BHJ can lead to a decrease in blue absorption predominately. In Fig. 4(a), the black lines represent the calculated absorption profile as a function of depth in a homogeneous BHJ in the device. The profiles actually depend on the wavelength; the absorption at 560 nm exponentially decreases with the depth length, while the absorption at 450 nm is maximized near the middle of 1a:C60 BHJ. Because of the low absorption coefficient of 1a:C60 at 450 nm, as compared to 560 nm, optical interferences are generated between the incident light and the reflected light at the BHJ/bottom ITO and bottom ITO/Glass interfaces at this wavelength. This wavelength-dependence of the absorption profiles in the organic layer can therefore be advantageously exploited to reduce the blue EQE only, by decreasing the concentration in C60 in the middle of the BHJ. In the inhomogeneous BHJ shown in Fig. 3(b) inset, a p-rich layer was therefore inserted in the middle of the BHJ, between the two other adjacent layers.

 

Fig. 4. (a) Absorption profile as a function of the depth in a homogeneous (H) and an inhomogeneous (I) 1a:C60 bulk-heterojunction (BHJ) in the device at 560 and 450 nm. (b) Internal quantum efficiency (IQE) as a function of the BHJ thickness at the reverse bias of 3 V, extracted from voltage-dependent IQE in the device. (c) Comparison of estimated EQEs at 560 nm and 450 nm in homogeneous and inhomogeneous devices. The inset shows schematics of OPD with an inhomogeneous BHJ consisting of three different 1a:C60-ratio layers. (d) Comparison of measured and calculated EQE spectrum at the reverse bias of 3 V in the optimized inhomogeneous BHJ device.

Download Full Size | PDF

The EQE spectrum was advertently estimated to optimize the thickness and the 1a:C60 ratio of each layer in inhomogeneous three-layer BHJ. The absorption of BHJ layers was calculated by optical simulation and the thickness-dependent IQE was measured from devices with various 1a:C60 ratios as shown in Fig. 4(b). The EQE spectrum was calculated with the 1a:C60 ratio (α, β, γ) from 1:1 to 2:1 and the thickness (d₁, d₂, d₃) from 0 to 100 nm in inhomogeneous BHJ shown in Fig. 4(c) inset. In Fig. 4(c), the red dots show the calculated EQE at 450 nm and 560 nm when varying the thickness (d₁, d₂, d₃) of each layer in the condition with α of 1.1:1, β of 1.9:1, γ of 1.5:1, and total thickness (d₁+d₂+d₃) of 120 nm. The black line shows the calculated EQE at various thicknesses with the homogeneous 1a:C60 ratios of 1:1, 1.5:1 and 1.9:1, respectively. It is clearly shown that the inhomogeneous BHJ can reduce the EQE at 450 nm without the EQE loss at 560 nm. At the EQE of 70%, the EQE at 450 nm can be further reduced to 17% (red arrow in Fig. 4(c)), for a three layer combination having the respective 1a:C60 ratio of 1.1:1 (thickness d₁ of 40 nm), 1.9:1 (d₂ of 70 nm), and 1.5:1 (d₃ of 10 nm). The light absorption profile in Fig. 4(a), corresponding to this inhomogeneous BHJ composition, show that the absorptance at 450 nm can be considerably reduced while the absorptance at 560 nm slightly changes only. Finally, a real device was fabricated with all optimized parameters, and the measured EQE reached 70.3% at 560 nm and 17.2% at 450 nm, which is in good agreement with the estimated values as shown in Fig. 4(d). The average blue EQE between the wavelengths of 440 nm and 480 nm was reduced from 24.1 to 17.7%, which corresponds to a reduction of 26%.

Thanks to the higher transmittance of OPD in the blue wavelengths, a hybrid stacked-type full-color photodetector composed of the green narrowband OPD and a bottom silicon photodiode exhibited fairly balanced EQEs in the blue, green, and red colors (see Fig. 5(a)). In addition, the device dark current was 0.165 nA/cm² and the specific detectivity (D*) was $4.37 \times {10^{13}}$ $cmH{z^{1/2}}{W^{ - 1}}$, which is estimated from the expression ${D^\ast } = EQE/{E_{ph}}/{(2q{J_d})^{0.5}}$ where E_ph is the incident photon energy in electron volts, at a reverse bias of 3 V. These device performances are very good considering the state of the art and are comparable to those obtained with a silicon photodiode used in commercial CMOS image sensor with color filters [26]. Among other device characteristics, the thermal stability was also measured. The dark current and EQE remained unchanged for at least 3 hours at 160 °C. The thermal stability is critical for OPD manufacturing, during the process stages of thin film encapsulation, layers coating for flattening, and micro-lenses conformation. The response behavior of the photocurrent density versus the light intensity exhibited good linearity (see Fig. 5(b)).

 

Fig. 5. (a) Calculated EQE spectra of red, green, and blue signal in the full stacked image sensor with 1.4 μm pixel size. The blue- and red-color EQE were calculated by FDTD simulation (Ref. [27]) and the IQE of the underlying Si photodiode. The inset shows the schematics of a stacked image sensor comprising Si and organic photodiodes. (b) Linear response behavior of the photocurrent density versus the light intensity

Download Full Size | PDF

3. Conclusions

In summary, we presented a high-performing green-selective OPD that reached an EQE over 70% at an operating voltage of 3 V, while the dark current was only 6 e/s/μm². Those performances result from the use of newly synthesized dipolar p-type compounds having high absorption coefficient and from light manipulation within the device. The p-type materials were designed with a properly balanced accepting/donating strength and selenophene as a linker to develop a cyanine-like molecules characterized by intense and sharp absorption. The blue-color EQE induced by the use of C60 as the n-type materials can be further reduced to 17% by light manipulation by choosing an inhomogeneous distribution of p-type:C60 materials in the active layer. In addition, important OPD device specifications such as a high thermal stability were also confirmed. This level of performance is in many aspects comparable to that obtained with a Si photodiode. Considering the dual function of the active organic layer as filter and absorber, simplifying the lensing process, it therefore appears that organic photodiode could be integrated in the next generation of image sensors, as an alternative to conventional Si-photodiode image sensor.

4. Experimental section

4.1 Syntheses of 1a and 1b

To a mixture of 5-(naphthalen-1-yl(phenyl)amino)selenophene-2-carbaldehyde (1 mmol) and 1H-cyclopenta[b]naphthalene-1,3(2H)-dione (1.1 mmol) for 1a or 1H-indene-1,3(2H)-dione (1.1 mmol) for 1b was added anhydrous ethanol (EtOH) (20 mL) at room temperature. The reaction mixture was stirred at the room temperature for 30 min, to which was added a drop of piperidine by using a syringe. The reaction mixture was further stirred at 85°C for 6 h and was cooled to room temperature. After the mixture was quenched with H2O, the resulting solid was collected via filtration and was thoroughly washed with H2O and hexanes, to afford reddish brown solids. The solid products were further purified via column chromatography on silica gel (chloroform/ethyl acetate = 9/1). The resulting product was recrystallized from dichloromethane/hexanes (yield = 93–95%).

1a: Melting point: 336°C. 1H-NMR (600 MHz, Methylene Chloride-d2): δ 8.24 (s, 1H), 8.15 (s, 1H), 8.09–7.99 (m, 5H), 7.93 (s, 1H), 7.86 (d, J = 5.0 Hz, 1H), 7.70–7.54 (m, 8H), 7.49–7.44 (m, 2H), 7.37–7.33 (m, 1H), 6.43 (d, J = 5.0 Hz, 1H). 13C-NMR (600 MHz, Methylene Chloride-d2): δ 190.69, 190.54, 179.50, 152.31, 146.41, 142.25, 140.91, 136.61, 136.57, 135.97, 130.67, 130.62, 130.57, 130.24, 130.11, 129.48, 128.88, 128.86, 128.58, 128.30, 127.79, 127.54, 127.40, 126.86, 124.93, 123.42, 122.52, 122.33, 119.25, 115.16. HRMS (ESI+): calculated for C34H22NO2Se [M + H+]: 556.0816; found: 556.0810. Elemental analysis: calculated for C34H21NO2Se: C – 73.65%, H – 3.82%, N – 2.53%; found: C – 73.69%, H – 3.74%, N – 2.55%.

1b: Melting point: 276°C. 1H-NMR (600 MHz, Methylene Chloride-d2): δ 8.08–7.99 (m, 3H), 7.84 (s, 1H), 7.81–7.74 (m, 2H), 7.70 (dd, J = 6.0, 2.1 Hz, 1H), 7.69–7.54 (m, 8H), 7.48–7.40 (m, 2H), 7.32 (t, J = 7.5 Hz, 1H), 6.39 (d, J = 5.0 Hz, 1H). 13C-NMR (600 MHz, Methylene Chloride-d2): δ 191.20, 191.14, 178.05, 151.29, 146.56, 142.41, 141.84, 140.79, 140.21, 135.96, 134.36, 134.29, 130.54, 130.25, 130.07, 129.43, 128.20, 128.00, 127.48, 127.47, 126.85, 124.79, 123.49, 122.28, 122.13, 117.17, 114.37. HRMS (ESI+): calculated for C30H20NO2Se [M + H+]: 506.0659; found: 506.0652. Elemental analysis: calculated for C30H19NO2Se: C – 71.43%, H – 3.80%, N – 2.78%; found: C – 71.55%, H – 3.72%, N – 2.86%.

4.2 Fabrication

C60 (Frontier Carbon Corp. >99.5% purity) was used as received without further purification. For the device fabrication, the active layer was prepared on indium tin oxide (ITO)-coated glass substrate having a sheet resistance of 15 Ω per square by vacuum thermal evaporation by co-depositing the p- and n-type materials under a pressure of 10⁻⁷ Torr. A 20-nm-thick hole transport layer made of molybdenum oxide was deposited on the active layer. Finally an ITO top electrode was evaporated through a shadow mask by using a customized facing targets sputtering equipment. To reduce the optical loss of the device, an additional AlO_x layer on the top-ITO electrode was deposited by thermal atomic layer deposition (ALD) using water and trimethylaluminum precursors.

4.3 Characterization

The GIWAXS measurements were carried out at the 3C1 SAXSI beamline at Pohang Light Source II (PLS-II, Pohang). The EQE spectra were obtained by using a spectral incident photon-to-electron conversion efficiency measurement system under monochromatic light generated by an ozone-free Xe lamp equipped with an optical filter and having a chopper frequency of 30 Hz. The intensity of the incident light was about 0.2 mWcm⁻² at the wavelength of 540 nm. The HOMO levels of all organic thin films were measured using an AC-2 photoelectron spectrophotometer (Hitachi High Tech) and the LUMO levels were determined from the optical band gap calculated from the edge of the absorption spectra. The absorbance, transmittance, and reflectance of the toluene solutions, thin films, and OPD devices were measured by UV-visible spectrophotometer (Shimadzu UV-240). For optical calculation, the refractive index (n) and extinction coefficient (k) values of each layer in the device were obtained by variable angle spectroscopic ellipsometry (VASE) measurements (J. A. Woollam M-2000). The photoluminescence was measured by a time-correlated single photon counting (TCSPC) setup (FluoTime 300, PicoQuant GmbH) by photoexcitation with a 510-nm picosecond laser (LDH-P-C-510B, PicoQuant GmbH) operated at 40 MHz.