High‐Performance Filterless Blue Narrowband Organic Photodetectors

Modern image sensors necessitate three color‐selective photodetectors: red, green, and blue (RGB) sensing units. Among those, blue light‐specific photodetectors are comparably much less reported. Here, a fully thermal‐evaporated organic photodetector (OPD) with narrowband blue detection based on exploring the dual role of the hole‐transporting layer is elegantly demonstrated. By incorporating a MoO3‐doped underlayer, the narrowband OPD achieves high external quantum efficiency up to 50% at 0 V with thin‐film devices. By adopting an intrinsic underlayer in planar junction configuration along with efficient hole and electron blocking layers, an ultralow dark current (2.46 × 10−12 A cm−2 at −0.1 V) is achieved. The device has a calculated specific detectivity (D*) reaching a record‐high value of 6.35 × 1014 Jones, outperforming commercial silicon photodetectors, and an ultrahigh linear dynamic range of 205 dB. This work paves the way for high‐sensitivity narrowband organic photodetectors in the visible spectral region.


Introduction
[7][8] Fine-tuning the parameters for customized commercial criteria is pivotal for the success of any sensing application.Organic semiconductors, albeit their lower charge carrier mobility and high exciton binding energy, outperform their inorganic DOI: 10.1002/adfm.202308719counterparts in terms of nontoxicity, flexibility, low cost, and molecular design versatility. [9,10]Organic photodetectors (OPDs), therefore, are attracting continuous attention.A traditional image sensor, typically comprised of red, blue, and green photosensitive components, is achieved by placing optical filters in front of the broadband inorganic PDs.Such an approach adds to the overall manufacturing cost, limits sensitivity in certain wavelength region and compromises photon intensity when integrating miniaturized sensing units. [11,12]Other promising approaches employing dispersive elements, like prisms and gratings, also encounter challenges like spatial resolution and high cost, despite their high performance. [12,13]Given the advantages of molecular building blocks and device engineering, OPDs have showcased superior filterless narrowband detection by means of charge collection narrowing (CCN), [14,15] charge injection narrowing (CIN), [16,17] novel narrowband absorbing materials, [18,19] self-filtering, [20,21] and microcavity integration. [22,23]Those methods propel the realization of spectroscopic sensing with phenomenal full-width half-maximum down to 7 nm. [24]Nevertheless, all approaches contain intrinsic drawbacks.A CCN device typically requires a relatively thick active layer (>1 μm), for which usually longer wavelength narrowband detection is achieved.In terms of CIN, it indeed obtains more photoresponse due to photomultiplication integration, but such strategy inevitably slows down the speed.Realizing a high detectivity yet fast photoresponse narrowband OPD remains a formidable challenge.
To achieve a high-performing image sensor, color discrimination is of paramount importance.While significant effort has been made to realize green OPDs [25][26][27] as well as longer wavelength narrowband OPDs, [8,28,29] few breakthroughs have been reported in the blue absorbing region.Li et al. demonstrated employing Rubrene and C 60 as the photoactive components to build a blue-channel visible light communication system. [30]They reported ultrahigh speed up to 950 kHz at −1 V, contributing to an exceptional data transmission rate of 530 kb s −1 for an all-organic system.Lu et al. designed a novel blue light filter in conjunction with a PffBT4T-2OD:PC 71 BM bulk heterojunction (BHJ) OPD, which exhibits ≈180 dB of linear dynamic range (LDR) under blue light illumination. [31]The same group later successfully transfers blue OPDs to a flexible substrate and Figure 1.a) Energy level of materials [33,35] used in the OPD devices.b) Absorbance spectrum of amorphous Rubrene, C 60 , BPAPF, and BF-DPB, and c) device structure of OPDs investigated, impinging photons come from the bottom.
enables low-voltage control.Segi et al. chose to thermally evaporate pyrazol derivative and 2,5-bis(5-tert-butyl-benzoxazol-2-yl) thiophene (BBOT)/bathocuproine to form a planar heterojunction OPD for blue detection. [32]They found that inserting merely 1 nm of Rubrene between the donor and acceptor boosts the photocurrent by 3.4 times while maintaining low dark current density ≈5 × 10 −11 A cm −2 .Exciting as those results are, compared to commercial inorganic PDs, those blue OPDs still lag behind predominantly due to low photocurrent.
Here, we further reveal the potential of Rubrene-based OPDs for blue photon detection.By exploring different wide-gap holetransporting layer (HTL) combinations, we unravel and exploit the dual role of HTL in achieving narrowband detection through effective exciton quenching.By doping Rubrene with MoO 3 , external quantum efficiency (EQE) increases by 50%, with the peak value up to 60% under −1 V.The high charge injection barrier endowed by both HTL and electron transport layer efficiently suppresses dark current, which leads to an ultrahigh specific detectivity (D * ) of 6.35 × 10 14 Jones at zero bias, one of the highest reported for blue OPDs, and exceeds that of commercial silicon PD.The devices employing similar structures can also achieve 414 kHz cutoff frequency, which is comparably fast for this class of devices.Furthermore, the facile manufacturing process and common material choice portend the great potential for commercialization.

Results and Discussions
To realize narrowband blue absorption, Rubrene and C 60 are chosen as the donor and acceptor components in the photoactive layer.The absorbance spectrum is shown in Figure 1a, with the respective chemical structures also attached.Noticeably, the primary absorption within the blue region is mainly due to C 60 , with minor contribution from Rubrene at around 500 nm.The device structure is depicted in Figure 1c, specifically ITO (90 nm)/BF-DPB:NDP9 or BPAPF:NDP9 (3 wt% doping for both, 40 nm)/BF-DPB or BPAPF (5 nm)/underlayer (20 nm, Rubrene or Rubrene:MoO 3 (6:1))/active layer (BHJ/PHJ)/HATNA-Cl 6 (10 nm)/HATNA-Cl 6 :W 2 (hpp) 4 (3 wt%, 10 nm)/Al (100 nm).The whole device stack is fully thermal-vapor deposited, and the overlap between the top and bottom contacts defines the active area of 6.44 mm 2 .Fabrication details are further specified in the Experimental Section.For optimization and performance purposes, both BHJ/PHJ configurations are investigated.While the planar structure consists of 25 nm of Rubrene and 20 nm of C 60 , different C 60 :Rubrene mixing ratios (10 wt%, 25% wt% donor to acceptor) form the BHJ.As dark current is pivotal to determine the D * , we meticulously chose those transporting layers with high injection barriers to prevent unwanted species injection from the relevant electrodes.Specifically, BPAPF and BF-DPB have shallow lowest unoccupied molecular orbital (LUMO) levels of −2.2 and −2.3 eV, respectively. [33,34]The ultradeep highest occupied molecular orbital of HATNA-Cl 6 (≈−7.3 eV) ensures minimum hole injection from Al under dark conditions. [35]To further suppress dark species generation, a smaller size mask is deployed when evaporating the doped layers, which reduces the contribution of lateral flow current to the overall dark current. [36]esides, the purpose of introducing an underlayer is to investigate the effect of MoO 3 doping on the OPD performance.Such a strategy prevents the omnipresence of dopants in the photoactive layer, which may disrupt morphology and induce severe recombination.
To fine-tune the detection spectrum suitable for blue light discrimination, the bulk needs to comprise a decent amount of C 60 to compensate for the lack of absorption of Rubrene in the range of 400-500 nm.However, the strong absorption of C 60 around 300-400 nm limits its applications in narrowband light detection.Therefore, we devise different HTL combinations to manipulate the absorption range.As plotted in Figure 2a,b, PHJ devices with BF-DPB and MoO 3 doped underlayer peak at 445 nm with EQE of 41%, which is 8% higher than the pure Rubrene underlayer.A similar EQE boosting effect is also observed in the case of BPAPF HTL.Since BPAPF and BF-DPB have roughly an electrical gap of 3.4 and 3.1 eV, respectively, the latter relates and quenches the absorption more in the wavelength closer to the blue region (Figure 1b), which manifests in an even narrower peak.The dual role of HTL, namely charge transport and photon quenching, is more pronounced for BHJ devices, with the highest value exceeding 50% at 455 nm.Such promotion is observed in all BHJ blend ratio, details can be found in the Supporting Information.Under −1 V reverse bias, such device can achieve 60% of EQE while maintaining narrowband spectral feature.Given BHJ its profuse amount of donor-acceptor interface and percolated pathways, we postulate that the charge generation and collection are more efficient for the case in BHJ device.However, the drastic enhancement in EQE upon MoO 3 doping in the underlayer requires further investigation.We examine the photocurrent (J ph ) trend with effective voltage (V eff ), [37][38][39] in which where J light is the current under one sun illumination and J dark is the dark current.V eff = V 0 −V, where V 0 is the voltage at which the J ph = 0, and V is the applied bias.J sat is usually determined at high revere bias when the photocurrent saturates.However, as plotted in Figure 2c,d, J ph continues to increase under high effective field.In order to demonstrate the relative comparison among the samples, we substitute the J sat with J ph,−2 V , which is the photocurrent obtained at V = −2 V. Therefore, the revised charge dissociation efficiency at is defined as: where J ph,sc is the photocurrent under short-circuit condition.
The plots and corresponding charge dissociation efficiencies are shown and indicated in Figure 2c,d.For both BHJ and PHJ cases, the MoO 3 effect on charge dissociation is significantly pronounced for devices with BF-DPB, with 18% and 53% boosting.Upon doping, MoO 3 and Rubrene form charge-transfer complexes, which also introduce additional intragap states that absorb in the long wavelength region. [40]To preserve the narrowband absorption, we therefore adopt a much lower doping ratio.Given that MoO 3 doping effectively deepens the Fermi level of Rubrene (≈0.2 eV difference), [30] a larger built-in potential upon doping is expected, which leads to more efficient charge extraction and dissociation.Thus, incorporating a MoO 3 -doped underlayer is an effective strategy to boost EQE.To gain more insight into the doped underlayer impact on device performance, we also substitute the dopant to F 6 -TCNNQ, [41] a common p-type dopant.The EQEs when adopting F 6 -TCNNQ, however, do not present drastic changes upon doping compared with MoO 3 .In comparison, it suggests the superior charge dissociation enhancement endowed by MoO 3 .All electrical characterizations are specified in Figures S5 and S6 of the Supporting Information.
For high-performing organic photodetectors, achieving an ultralow dark current remains a formidable challenge.Herein, we plot the current density-voltage (J-V) characteristic curve for light and dark conditions.As shown in Figure 3a, PHJ with BPAPF HTL exhibits an ultralow dark current density (J d ) of 2.46 × 10 −12 A cm −2 at −0.1 V (3.08 × 10 −11 A cm −2 at −2 V).Along with its decent short-circuit current density of 1.86 mA cm −2 , such device presents phenomenal diode-rectifying performance, among the best OPDs to date.At around 0 V, according to theoretical modeling, [42] J-V characteristics show predominantly shuntinduced current contribution.In our case, however, we observe a clear exponential relationship even near 0 V, which indicates the ultrahigh shunt resistance.The MoO 3 -doped sample, comparably, also presents a low J d (6.18 × 10 −10 A cm −2 at −0.02 V) near 0 V but higher J d (6.67 × 10 −7 A cm −2 ) at −2 V, which is more than four orders of magnitude larger than the undoped sample.By contrast, it was previously reported two orders of magnitude J d suppression due to the shifted Fermi level upon MoO 3 doping. [30]n our case, such effect might be overshadowed by the high injection barrier of our HTLs.Considering the fact that V oc remains unchanged regardless of doping in PHJ (detailed value tabulated in Table S1, Supporting Information), the strong voltage dependence of J d can be related to the worsened charge injection and charge recombination induced by MoO 3 .Such trend also manifests in BHJ devices.The lowest J d down to 5.54 × 10 −12 A cm −2 at −0.02 V (3.05 × 10 −9 A cm −2 at −2 V) is obtained from the device with an intrinsic underlayer.As reported in previous literature, the donor/acceptor interface is responsible for dark species generation. [43,44]PHJ effectively reduced such interface to suppress possible thermalization processes mediated by the intragap states. [43,45]Furthermore, the effect of the doped underlayer is consistent regardless of device structure or HTL choices.Doping also worsens BF-DPB-containing devices less than BPAPF counterparts by roughly one order of magnitude.The relatively high J d showed in BF-DPB devices can be ascribed to the slightly shallower LUMO level compared to BPAPF, which blocks dark current injection not as efficient as the latter case.Interestingly, the V oc in terms of BHJ MoO 3 doped devices present almost the same values (0.70 V and 0.71 V) for BF-DPB and BPAPF containing devices.Compared with the V oc (0.89 V for BPAPF and 0.92 V for BF-DPB) of pure Rubrene underlayer, it suggests significant voltage losses upon doping.
To unravel the mechanism of the doped underlayer, we further measure the ultra-sensitive EQE to investigate the subgap features below the charge-transfer state (E CT ).In Figure S4a,b of the Supporting Information, no pronounced additional subgap states are observed when adopting the doped underlayer.We can identify two distinctive features, for which the MoO 3 -doped sample shows a relatively higher signal in PHJ configuration.In the case of BHJ, such feature is not so distinguishable with or without doping.Noticeably, when adopting doped underlayers, EQE curves suggest almost identical photoresponse in the subgap region.We further fit those energetic states, which are in the range of 1.0-1.2eV for E sub1 , and 0.7-0.9eV for E sub2 .It is worth mentioning that E sub1 states are energetically higher than the midgap.Through fitting and calculation according to previously reported methods, [46] we further quantify the nonradiative losses (ΔV oc,nr ).MoO 3 -doped underlayer samples, suffer significantly from ΔV oc,nr , with a striking value of 596 mV, are 75% higher than the undoped BHJ (340 mV).Therefore, we can attribute the several orders of magnitude higher J d to the dopantinduced recombination losses.
We further compare the speed performance of the devices.The −3 dB cutoff frequency is the figure of merit for evaluating a device's transient response, which is defined as: [28] 1 where f RC is the RC cutoff frequency determined by the intrinsic electrical property of the device, and f tr is the transient cutoff frequency for charge carriers.For the latter, it also can be expanded into [47] f tr = 3.5 in which t tr is the transient time of the charge carriers, μ eff denotes the effective mobility, V is the applied voltage, V bi presents the built-in voltage, and L refers to the active layer thickness.We adopt the rise time to estimate the t tr and derive the f tr , [48] which are plotted as dashed lines in Figure 3c,d.As seen in Figure 3c, we first compare HTL influence.BF-DPB outperforms BPAPF by a very small value, but they both achieve high cutoff frequencies of 412 kHz and 386 kHz at 0 V.We address the difference to higher doping efficiency in BF-DPB, which leads to increased conductivity and better transient characteristics.In line with the observation in frequency response, the rise and fall time of BF-DPBcontaining devices are also faster than those of BPAPF, with the value of 861 ns/1.38 μs (BF-DPB) and 956 ns/3.09μs (BPAPF), respectively.In Figure 3d, the effect of MoO 3 doping on speed performance is also compared.Both samples cutoff at roughly the same frequency, 397 kHz for underlayer doped with MoO 3 and 412 kHz for the device without.Such independence also reflects on the rise and fall characteristics, with negligible difference between those two.Noticeably, the higher energy subgap states remain the same with or without doping, which explains the similar speed performance that was previously reported to be mediated by shallow traps of similar energy levels. [49]As can be seen in both cases in Figure 3c,d, f -3db is approaching f tr , which indicates a high μ eff in the measured system.Given the ultralow J d performance for our OPDs, we further obtain the LDR, which measures the photoresponse linearity with regard to light intensity.It is formulated as: [50][51][52] LDR = 20 * log J lin,max where J lin,max and J lin,min are the maximum and minimum photocurrent densities that define the linearity region.Figure 4a shows the LDR of the PHJ device without doping.By changing light intensity over almost ten orders of magnitude, the linearity retains (slope = 0.999), which renders an ultrahigh LDR of 205 dB without bias.This promising result is in line with its ultralow J d .In fact, we plot the LDR for silicon against our OPD, as shown in Figure S5 of the Supporting Information, the silicon diode photoresponse also deviates at ≈10 −13 A cm −2 , which indicates potential setup limitation and even theoretically higher LDR for our OPD.The LDR for the highest EQE sample is also measured and plotted in Figure S6 of the Supporting Information.Despite its BHJ configuration and incorporation of MoO 3 -doped underlayer, it also demonstrates a striking high LDR of 192 dB.Revisiting its J-V characteristics, we associate such high value with the comparatively low J d in the vicinity of 0 V. Specific detectivity (D * ), as the prime figure of merit when comparing OPD performances, is also calculated as follows: [53] D * = R √ A * Δf I noise (6)   in which R is the spectral responsivity, A is the device area, Δf is the noise measurement bandwidth, and I noise denotes the noise current.The noise current can be further specified as: [53][54][55] I noise = √ I 2 sn + I 2 jn + I 2 fn (7) Adapted under the terms of the CC-BY license. [28]Copyright 2021, Wiley-VCH GmbH.
where the noise can decompose into the contribution of shot noise (I sn ), thermal noise (I jn ), and flicker noise (I fn ).I dark is the dark current, q is the element unit charge, k B is the Boltzmann constant, T is the temperature, and R SH is the shunt resistance.Flicker noise (I fn ) describes the random fluctuation of carrier concentration or mobility, which is inversely proportional to the frequency. [56]It is imperative to emphasize the importance of impartially calculating D * using the above equation, because simple approximation of shot noise domination can lead to significant overestimation of D * near 0 V. [6,57] As is depicted in Figure S7a,b of the Supporting Information, the measured noise spectral density is indistinguishable from that of the background noise level, indicating the actual I noise is below the measurement sensitivity limit.At 0 V, the contribution from I sn is negligible compared to I jn , from which we calculate the D * to be a record high value of 6.4 × 10 14 Jones for PHJ adopting BPAPF and pure underlayer.We further measure the I noise at −2 V (Figure S8b, Supporting Information), the noise spectrum barely presents any increase despite high reverse bias, proving the reliability of the measured ultralow J d and high device quality.For the high EQE sample, which adopts BF-DPB with MoO 3 doped underlayer, we compare the D * value derived from calculated noise and experimentally measured noise.As shown in Figure S7c of the Supporting Information, the noise value in such case is distinguishable from the background noise level.The comparison is plotted in Figure S8a of the Supporting Information, which shows less than an order of magnitude of difference, indicating good agreement between the theory and the experiment.Despite the high R value (0.21 A W −1 at −0.5 V), its high J d drags down its overall detection sensitivity and ends up with 5.6 × 10 12 Jones at 0 V.To date, those devices are among the highest D * reported for narrowband blue OPDs.The highest D * achieved by PHJ, as plotted in Figure 4b, beats commercial silicon photodiode by about two orders of magnitude, showcasing superior performance and potential for commercial application.

Conclusion
We demonstrate the possibility of achieving excellent narrowband blue organic photodetectors through intended exciton quenching, which is realized by exploring the optical and electrical properties of HTL as well as introducing MoO 3 -doped underlayer to the device.Such filterless strategy ensures a high EQE of up to 50% at 0 V in thin-film devices.Doping can further improve EQE by assisting charge carrier dissociation.However, doping induces extra nonradiative recombination in the case of BHJ and affects the charge injection profile, which upsets the overall dark current.However, ultralow J d can be obtained by PHJ devices, which renders an ultrahigh LDR of 205 dB and recordhigh D * of 6.35 × 10 14 Jones for blue OPDs.Such excitement can complement recent success of green narrowband OPDs to propel the advancement of high-performance filterless fully organic RGB sensors.

Experimental Section
Device Fabrication: All devices were thermal evaporated in the vacuum chamber system (Kurt J. Lesker, UK) with base pressure of less than 10 −7 mbar.ITO substrates were purchased from Thin Film Devices with ≈90 nm of ITO on 1.1 mm of glass.Prior to processing, substrates were thoroughly cleaned and plasma cleaning was also performed.Shadow masks and shutters were employed to make required thickness and device structures.Doped layers were done by coevaporation of different materials concurrently.To ensure long-term stability, the devices were all encapsulated with transparent glasses on top of the device utilizing an epoxy resin (Nagase ChemteX XNR 5592, Japan) cured by UV light.A moisture getter (Dynic Ltd., UK) was inserted in the between to prevent degradation.

Figure 2 .
Figure 2. EQE spectrum of the devices at 0 V for a) PHJ OPDs with different HTLs and b) BHJ C 60 :Rubrene 10 wt% OPDs with different HTLs.The inset graph in (b) shows the EQE of the device with BF-DPB HTL and MoO 3 -doped underlayer under different biases.Photocurrent density versus effective voltage characteristics for c) PHJ and d) BHJ.Revised dissociation efficiencies of different conditions are calculated and denoted on the top left corner.

Figure 3 .
Figure 3. J-V characteristics for devices with a) PHJ structure and b) BHJ structure (C 60 :Rubrene 10 wt%).Blue and yellow lines denote the ones with BPAPF or BF-DPB as HTL, solid and dashed lines denote whether the underlayer is doped with MoO 3 .Light and dark response can be distinguished by the magnitude difference.Frequency responses of BHJ devices: c) comparison of effect of HTL layer choices.d) Comparison of the effect of MoO 3 doping on underlayers.(c,d) Insets are OPD transient responses to a 455 nm LED with an intensity of 49.50 mW cm −2 operating at 20 kHz.The photoresponse is normalized to compare.

Figure 4 .
Figure 4. a) LDR of the device adopting BPAPF HTL with an intrinsic underlayer, the light source is a 455 nm LED.The dashed line is the linear fit of the data, with a slope of 0.999.b) D * comparison of this work and other reported work.The dashed lines indicate silicon, InGaAs, and Ge PD, respectively.For the top red point, D * for BPAPF with undoped underlayer.Lower point is the D * for BF-DPB with MoO 3 -doped underlayer.Both D * are calculated based on thermal noise domination at 0 V. Adapted under the terms of the CC-BY license.[28]Copyright 2021, Wiley-VCH GmbH.