Enhancing sub-bandgap external quantum efficiency by photomultiplication for narrowband organic near-infrared photodetectors

Detection of electromagnetic signals for applications such as health, product quality monitoring or astronomy requires highly responsive and wavelength selective devices. Photomultiplication-type organic photodetectors have been shown to achieve high quantum efficiencies mainly in the visible range. Much less research has been focused on realizing near-infrared narrowband devices. Here, we demonstrate fully vacuum-processed narrow- and broadband photomultiplication-type organic photodetectors. Devices are based on enhanced hole injection leading to a maximum external quantum efficiency of almost 2000% at −10 V for the broadband device. The photomultiplicative effect is also observed in the charge-transfer state absorption region. By making use of an optical cavity device architecture, we enhance the charge-transfer response and demonstrate a wavelength tunable narrowband photomultiplication-type organic photodetector with external quantum efficiencies superior to those of pin-devices. The presented concept can further improve the performance of photodetectors based on the absorption of charge-transfer states, which were so far limited by the low external quantum efficiency provided by these devices.

The authors presented a method to enhance the external quantum efficiency (EQE) of near-infrared (NIR) photoresponse. This is realized by utilizing the charge-transfer states between organic electron donor molecule (D, ZnPC) and electron acceptor molecule (A, C60) with the photodiode having a photomultiplication-type structure, that is, reducing the acceptor composition to a few percent (e.g., 3 wt% here). The work mainly divided into two parts. The first part is to demonstrate the photomultiplication effect can be achieved with the active layer having 3 wt% C60 in ZnPC. A maximum EQE of almost 2000% at -10V and the optimal specific detectivity D* of 2.4 x 1012 Jones at about -2.5 V were demonstrated. The idea of on-purpose reducing acceptor composition in either polymer:fullerene or polymer:semiconductor nanoparticle active layer to achieve electron-trapping induced band bending and hole injection to realize photomultiplication has been reported before (e.g., refs. 10 and 14). Therefore, the first part of the work is an increment of the previously published work. The second part of the work is to utilize this photomultiplication-type photodiode (mainly the active layer with 3 wt% C60) but with semi-transparent Ag bottom electrode and opaque Ag top electrode to form nanocavity to realize the NIR photoresponse due to charge-transfer states. By varying the active layer thickness, the wavelength of NIR response varies so does EQE. The highest EQE was achieved ~80% at the peak position of 826 nm at -10V applied bias. Utilizing charge-transfer states of ZnPC:C60 and nanocavity to realize NIR detection was reported by the same group three years ago at Nature Communications (ref. 30). The difference between this work and the previous one is mainly the composition change (from roughly ZnPC:C60 of 1:1 ratio to 3 wt% C60) and active layer thickness change (from < 100 nm to 300-400nm). If compare the photoresponse at the same wavelength, e.g. ~870 nm, EQE is ~20% (ref. 30 Fig.  2a), and EQE is ~20% (this work, Fig. 4C). Therefore, the second part of the work is also an increment of the previous published work by the same group. More comments are provided below.
1. The PM-OPDs discussed in Introduction are mainly polymeric donors, few are small molecule donors (or not explicitly discussed). This work used small molecule as donor. So, it would be helpful a review of the PM-OPD based on small donors in Introduction. 2. Figure 1a and b should label the materials of two electrodes as well as anode and cathode. Symbols of electron and holes should be given to indicate what carriers are injected from anode and cathode under reverse bias, respectively. Figure caption should clearly state what bias is applied for Figure 1a and b. Energy levels of materials involved in the device should be provided. 3. The UV-Vis absorption spectra of ZnPc and C60 should be provided to show the absorption range of materials used in the active layer. These spectra could also help to explain feature exhibited in EQE, that is, a peak around 390 nm, a dip around 490 nm, and broad response between 500-870 nm (Figs.1d and 2a). 4. P. 4, lines 114-115. It should clearly state whether HATNA-Cl6 is also an ETL and whether it is doped or not. 5. Apparently, EQE varies with the incident light wavelength. When reporting the maximum EQE at difference reverse biases, the EQE at which wavelength should be indicated. What is the physical meaning of the ratio of EQE at different reverse biases to zero bias in Figure 1e? The calculation of specific detectivity will use the absolute value of EQE at each bias not the ratio. So, should the absolute EQE be a meaningful comparison? As author pointed out the EQE increases with the increase of applied reverse bias, why the reverse bias is capped at -10 V? Did authors try to further increase the bias? 6. P. 5, lines 139-140, please cite reference(s) in this sentence. "Recently, enhanced CT state absorption photodetectors (CT-OPDs) have been introduced, which could benefit from high EQEs provided by photomultiplication." 7. Dark current of the device containing 3 wt% C60 increases significantly with the applied reverse bias. Since Figure 1 displays the EQE up to -10 V reverse bias, J-V curves in Fig. 2a should also show the reverse bias to -10 V. 8. Figure 3 shows the J-V characteristics and D* of devices based on "Photomultiplicator, 3 wt%" and "pin-diode, 3 wt%". What is the device structure of pin-diode? It is suggested to show the J-V curves under forward bias, for example, to +2 V bias. There is no rectification shown in J-V curves under illumination for a "Photomultiplicator, 3 wt%" device ( Fig. 2a). It is interesting to show if the "pin-diode, 3 wt%" device shows a rectification in both dark and illuminated J-V characteristics. 9. The entire paragraph on pp. 10-11, lines 274-282 is very unclear. (1) Please provide the transfer matric optical simulations results. (2) Please clearly indicate which active layer thickness corresponds to which resonant wavelength. (3) What is the exact device structure to realize the so called "optical cavity" effect. 10. Different wavelength range of EQE spectra are shown, for example, 300-900 nm ( Fig. 1d and Fig. 2b), 300-1400 nm (Fig. 4a), and 600-1200 nm (Fig. 4c). Any explanations? 11. The concept of charge collection narrowing (CNN) was first proposed in the paper of ref. 42, not ref. 33. 12. P. 13, line 347, "The intensity is controlled by a Hamamatu S1337 silicon photodiode". The intensity is calibrated by this standard diode not controlled by this standard diode. 13. The active layers were prepared via vapor deposition, which gives an handle for controlling the distribution of C60 in the active layer relatively easier than using spin coating method, for example, a gradient composition distribution with more C60 in the active layer close to the cathode to facilitate electron trapping and band bending. Therefore, deposition conditions are critical and the details for organic molecules (ZnPc and C60) should be provided, such as how to control the composition of C60 to be 1 to 4 wt%, what are the deposition rates, what are the thickness of active layers, what is the background pressure, etc. 14. The ITO layer is quite thin. What is the conductivity of 90 nm ITO/glass? Reviewer #3 (Remarks to the Author): The manuscript describes a vacuum-processed organic photodetector, claimed to operate based on the photomultiplication principle, that leverages near-infrared absorption of a charge-transfer state between the two small molecular organic compounds comprising the absorption layer. The reported device performance is quite good on some of the relevant metrics, and the principle of operation potentially enhances the path to new and existing design and application possibilities.
The paper is generally well written and can be followed, albeit non-specialists may have to spend extra time parsing some of the specialized photodetector terminology. Figures are well formatted, presenting adequately the principle of operation, schematically showing the composition, and most of the key photoresponse characteristics of the device.
The concept itself appears to be a synthesis of prior work on photomultiplication detection in the solid state and organic photodetectors leveraging optical cavity-enhanced absorption, as well as some prior work showing absorption by the CT state. The specific choice of molecules is not new, although they are combined judiciously to produce a good photoresponse. Thus, the merit seems to be primarily in the practicality of the device in principle, and the path forward afforded by its processing, thinness, and low dark noise relative to alternatives.
It such light, it would be useful to better understand why the observed response is attributable to PMT action specifically, as opposed to say that of a Schottky diode. The discussion may also benefit from examining the temperature dependence of the response. Finally, the speed of operation of this device should be discussed, and whether it would place any practical constraints on its applications.

Dear reviewers,
We thank you for your careful peer-review of our manuscript and the very helpful criticism. In response to the reviewer's comments, we made substantial changes to the manuscript. In the document below, all the concerns and points raised are addressed and additional measurements and analyses are presented. For better readability, we visually structured our response as following: The comments and points of the referees are copied in black color and italic style.
Our response is colored in green.
Furthermore, we cite parts from the revised manuscript or the revised supporting information to afford a quick impression of the additions and improvements to the reviewers. These citations are indicated by blue text and are shifted to the right.
We hope that all points are clearly addressed. We thank the reviewer for the fruitful comments and for recommending our manuscript for publication. Below we address the question raised: 1. The authors claimed that "Narrowband PM-OPDs show peak EQEs from 20 to 80% under -10 V with full width at half maximum (FWHM) from 20 to 40 nm" is wrong. If the EQEs are lower than 100%, which is not a kind of PM-OPDs, should be photodiode type organic photodetectors.
While we understand the concerns raised regarding the definition of the photomultiplication (PM) effect, we believe that PM not necessarily has to lead to EQEs above 100%, since it also depends on the absorption characteristics of the targeted region. Especially for charge-transfer states, it is well documented that the absorption cross-section of these states is orders of magnitude lower than the main absorption region. Therefore, as a lower portion of the incoming photons is absorbed, less free charge carriers are generated, decreasing the magnitude of the EQE. A more robust characterization of the PM effect in the sub-gap absorption region can be achieved by analyzing the internal quantum efficiency (IQE), which takes into account only absorbed photons. Indeed, in Supplementary Figure 4b, we show that the calculated IQE reaches ~160% at -10 V for the PM-OPD with detection wavelength of 828 nm, even though the EQE is 80%. Nonetheless, it is also possible to demonstrate the PM simply by further increasing the applied voltage. This is summarized in Supplementary Figure 4

, where
EQEs over 100% are demonstrated for all the devices at -15 V leading to IQEs of 920% for the best performing device. We discuss this point in lines 46 to 49 of the revised manuscript: "If the transit time of injected charge carrier is lower than the lifetime of the accumulated, photo-generated charges, an EQE > 100% is observed. Here, we would like to stress that prior to the photomultiplication process the photon needs to be absorbed by the active layer. We therefore conclude that the minimum criteria for photomultiplication is that the internal quantum efficiency (IQE) is larger than unity." In lines 84 to 92: "Such narrowband OPDs could significantly benefit from the increased IQE, if photomultiplication would take place also by direct excitation of CT states However, it is unclear whether direct excitation of CT states can result in a photomultiplication process. Utilizing the intermolecular CT state absorption renders possible to detect NIR photons beyond 1700 nm meanwhile using rather small and sublimable organic semiconductors where the absorption profile can be easily tuned by the D-A system. On the other hand, the weak intermolecular absorption cross section 1,2 challenges the overall performance and here the photomultiplication could improve the electrical performance by increasing the gain for every absorbed photon." Also in lines 296 to 303: "While the PM effect is commonly accompanied by an EQE above 100%, it is the IQE which better defines the physical phenomenon behind this effect. In order to induce PM, free charge carriers must be firstly generated, requiring photons to be absorbed. As a means of quantifying whether absorbed photon induce enhanced injection in the sub-gap absorption region in our devices, we estimated their IQE, which accounts only for absorbed photons. To that end, we employ the transfer 2.The authors also claimed "narrowband photomultiplication type devices based on charge collection narrowing (CCN)33" The reference [33]  We thank the reviewer for pointing out this misunderstanding. The description of the effect demonstrated in ref. [33] (now 39) was improved. In lines 105 to 107 in the main text: "These results are comparable with narrowband organic pin-photodiodes based on cavities 5,6 , and higher than that of narrowband photomultiplication type devices based on charge injection narrowing (CIN) 7 ." In lines 332 to 334: "Moreover, it is superior than that of narrowband photomultiplication type devices based on CIN 7 , where, in addition, excessively thick devices are demanded, which increases the operation voltage." In lines 358 to 362: "In CT-OPDs, the thicknesses required are much smaller than those used in narrowband devices based on charge collection narrowing (CCN) 8,9 or on CIN 7,10 .
In the latter, for example, the thickness of the active layer must be much larger than the inverse of the absorption coefficient of the active layer, such that under illumination charges are generated close to the injecting contact, thereby causing the necessary band bending 10 .".
In lines 365 to 368: …, where active layers of 355 nm were used for spectral response at ~830 nm, compared to 2.5 μm reported for spectral response at 650 nm when using CIN combined with photomultiplication 7 .
The requested references are cited in lines 50 to 51: "The effect described above has been applied in organic and hybrid PDs, leading to outstanding external quantum efficiencies (EQEs) as high as 10 5 % [11][12][13] ." And lines 41 to 43: "…different strategies have been introduced to achieve such charge accumulation and thereby the required energy level bending through a lack of percolation path for one charge carrier type 14 As suggested in the conclusion, we believe that this approach can indeed lead to an improved performance, besides improving optical interference problems intrinsic to optical microcavity devices. However, this ongoing topic still requires a careful study, which is currently under consideration. Nonetheless, the results shown in our manuscript already describe the novel possibility of enhancing EQE in the CT absorption range, combined with microcavities, which had not been achieved so far.    .g., refs. 10 and 14).

For the NIR
Therefore, the first part of the work is an increment of the previously published work.  . 2a), and EQE is ~20% (this work, Fig. 4C). Therefore, the second part of the work is also an increment of the previous published work by the same group. More comments are provided below.
We thank the reviewer for the critical review which helped us to improve the quality of the manuscript. Nonetheless, we would like to point out that there are fundamental differences between previous reported work 5 and what is being proposed in this manuscript. Previously, Siegmund et al. 5 reported a diode-type OPD embed in a cavity.
Such devices are limited to EQEs of 100%. The devices reported here combine the photomultiplicative gain with an optical cavity, allowing for EQEs higher than 100% even in the sub-band gap region. In order to explain this effect in more detail, we added Thank you for pointing out these deficiencies. The figure and the caption were updated to better explain the device working principle.

The UV-Vis absorption spectra of ZnPc and C 60 should be provided to show the
absorption range of materials used in the active layer. These spectra could also help to explain feature exhibited in EQE, that is, a peak around 390 nm, a dip around 490 nm, and broad response between 500-870 nm (Figs.1d and 2a).
The UV-Vis absorption spectrum of the blend ZnPc:C60 is now shown together with the voltage-dependent EQE in Fig. 1d.  HATNA-Cl6 is used intentionally undoped to achieve charge accumulation near the contact. While its low HOMO of around -7.1 eV blocks hole injection under reverse bias in the dark, its low conductivity helps the necessary electron accumulation. This is further clarified in the line 123: "Pristine HATNA-Cl6 is used as hole blocking layer (EBL)" And lines 131 to 134 in the main text: "…electrically undoped HATNA-Cl6 layer 35 . While n-doped HATNA-Cl6 has been already employed as an electron transport layer, in this device, we intentionally use a pristine layer such that the electron extraction is hindered and slowed down, which helps the electron accumulation at the cathode." Figure 1e? The calculation of specific detectivity will use the absolute value of EQE at each bias not the ratio. So, should the absolute EQE be a meaningful comparison? As author pointed out the EQE increases with the increase of applied reverse bias, why the reverse bias is capped at -10 V? Did authors try to further increase the bias?

Apparently, EQE varies with the incident light wavelength. When reporting the maximum EQE at difference reverse biases, the EQE at which wavelength should be indicated. What is the physical meaning of the ratio of EQE at different reverse biases to zero bias in
For clarity, we now specify the wavelength at which EQE is extracted, namely 670 nm, which represents the maximum EQE for all voltages applied. Indeed, in calculation of the specific detectivity D * , only the absolute value of EQE is needed as is shown in Fig.   3b, right axis. However, Fig. 1e represents the relative increase in EQE upon bias increase, as compared to the value achieved without enhanced charge injection.
As shown in Fig. 3b, increasing the reverse applied voltage does not increase the detectivity of our device, as the dark current increases concomitantly. Therefore, in spite of an expected increase in EQE, we capped the voltage at -10 V. We briefly explain that in line 148 to 149: "However, as it will be discussed below, an optimum operation regime exists in the range of -2.5 V, where the highest D* is achieved."  Fig. 2a should also show the reverse bias to -10 V.
The JV curve for 3 wt% device, which shows the best performance, is shown in Fig. 3a and was measured until -10 V. Fig. 2 instead aims to explain the difference between different acceptor concentrations, we believe therefore that the presented data is conclusive. Figure 3 shows the J-V characteristics and D* of devices based on "Photomultiplicator, 3 wt%" and "pin-diode, 3 wt%". What is the device structure of pindiode? It is suggested to show the J-V curves under forward bias, for example, to +2 V bias. There is no rectification shown in J-V curves under illumination for a "Photomultiplicator, 3 wt%" device (Fig. 2a). It is interesting to show if the "pin-diode, 3

8.
wt%" device shows a rectification in both dark and illuminated J-V characteristics. "In order to test whether such devices could be achieved, we embedded the best performing PM-OPD, i.e. 3 wt%, into an optical microcavity, see inset in Figure 4 for the device structure. Due to the higher Ag work function as compared to that of ITO, we inserted a 10 nm thick MeO-TPD layer to hinder hole injection in 10. Different wavelength range of EQE spectra are shown, for example, 300-900 nm ( Fig. 1d and Fig. 2b), 300-1400 nm (Fig. 4a), and 600-1200 nm (Fig. 4c). Any

explanations?
The sensitivity of the setup used in the measurement of Fig. 1d and Fig. 2b is much lower than the one used for the measurements shown in Fig. 4a and Fig. 4c, so that no signal can be measured for wavelength above 900 nm. This setup also uses only a Si diode, in contrast to the more sensitive one, which also uses an InGaAs diode as reference. In Fig. 4a, the entire measurable spectra are shown in order to convince the reader that the amplification occurs similarly from 300 to 1400 nm. In Fig. 4c, we focus on the region for which the cavity effect is observed. Thank you for pointing out this error. The reference has been fixed in line 358.
"In CT-OPDs, the thicknesses required are much smaller than those used in narrowband devices based on charge collection narrowing (CCN) 8,9 or on CIN 7,10 ." 12. P. 13, line 347, "The intensity is controlled by a Hamamatsu S1337 silicon photodiode". The intensity is calibrated by this standard diode not controlled by this standard diode.
The corresponding sentence has been rewritten in line 405: "The intensity is calibrated by a Hamamatsu S1337 silicon photodiode." 13. The active layers were prepared via vapor deposition, which gives a handle for controlling the distribution of C 60 in the active layer relatively easier than using spin coating method, for example, a gradient composition distribution with more C60 in the active layer close to the cathode to facilitate electron trapping and band bending. "Additionally, the method presented here allows placing the active layer in different positions within the device or using gradients of D-A mixing ratio, thereby enhancing injection and diminishing the effect of optical overtones, a critical problem in CT-OPDs."

The ITO layer is quite thin. What is the conductivity of 90 nm ITO/glass?
The sheet resistance of ITO is 32 Ω□ -1 . This information is now provided in line 394: The concept itself appears to be a synthesis of prior work on photomultiplication detection in the solid state and organic photodetectors leveraging optical cavityenhanced absorption, as well as some prior work showing absorption by the CT state.
The specific choice of molecules is not new, although they are combined judiciously to produce a good photoresponse. Thus, the merit seems to be primarily in the practicality of the device in principle, and the path forward afforded by its processing, thinness, and low dark noise relative to alternatives.
It such light, it would be useful to better understand why the observed response is attributable to PMT action specifically, as opposed to say that of a Schottky diode. The discussion may also benefit from examining the temperature dependence of the response. Finally, the speed of operation of this device should be discussed, and whether it would place any practical constraints on its applications.
The authors thank the reviewer for the positive feedback provided. Below, we discuss the points raised by the reviewer: The term photomultiplication in organic devices is commonly attributed to devices that present an increased injection under applied bias as a result of a charge accumulation close to the injecting electrodes. In our devices in dark conditions, the barrier imposed for holes under reverse bias can be interpreted as a Schottky barrier, where the activation energy is related to the barrier height. However, under illumination, the injection process is dominated by the bending of the energy level, allowing holes to efficiently tunnel through the barrier. This process is characterized by an increased EQE, i.e., a photo-gain, which is not expected for a Schottky diode. Therefore, in such condition, the interpretation of a Schottky diode no longer describes the working mechanism of these devices. This can be seen from the activation energy Ea under illumination as compared to Ea under dark conditions. While in the dark Ea decreases with the reverse applied voltage, as a result of a decreased barrier height, under illumination very week voltage dependence is observed. In the latter, the band bending allows charges to tunnel the barrier such that the barrier does not play a role in the injection. The authors have conducted careful revision and addressed the questions/comments raised by reviewers. The authors also stated the novelty and the difference between this work and the previous work from the group. It is recommended for publication before making the correction as described below. Figure 1d, it is suggested to use the right y-axis for "Absorption" because Absorption is not in the same order of percentage (if authors not choosing to use (abu) for the unit of Absorbance) as EQE.

Dear reviewers,
We thank you for your careful peer-review of our manuscript and the very helpful criticism.
In the document below, all the concerns and points raised are. For better readability, we visually structured our response as following: The comments and points of the referees are copied in black color and italic style.
Our response is colored in green.
We hope that all points are clearly addressed.
Sincerely, Jonas Kublitski on behalf of all authors