Integrated molecular diode as 10 MHz half-wave rectifier based on an organic nanostructure heterojunction

Considerable efforts have been made to realize nanoscale diodes based on single molecules or molecular ensembles for implementing the concept of molecular electronics. However, so far, functional molecular diodes have only been demonstrated in the very low alternating current frequency regime, which is partially due to their extremely low conductance and the poor degree of device integration. Here, we report about fully integrated rectifiers with microtubular soft-contacts, which are based on a molecularly thin organic heterojunction and are able to convert alternating current with a frequency of up to 10 MHz. The unidirectional current behavior of our devices originates mainly from the intrinsically different surfaces of the bottom planar and top microtubular Au electrodes while the excellent high frequency response benefits from the charge accumulation in the phthalocyanine molecular heterojunction, which not only improves the charge injection but also increases the carrier density.

molecularly thin organic heterojunction of F16CoPc/CuPC. They claim that the excellent frequency performance arises from (1) a soft top contact created by a rolled-up Au microtube treated by water and (2) charge transfer between Au electrode and the F16CoPc/CuPC heterojunction.
As this paper reports an interesting and important advance in developing molecular scale rectifiers, it is recommended to be published after addressing the following issues.
(1) In Figure S5a, I-V characteristics of devices with mesa width Wmesa = 10 µm are compared with those with mesa width Wmesa = 5 µm. Since mesa width difference is 2, the current difference should be around 2 times if the top contact created by a rolled-up Au microtube is similar. However, it appears that currents are at least 4 or 5 times higher for devices with mesa width Wmesa = 10 µm compared to Wmesa = 5 µm. This may suggest that there exists a lateral leakage current. Then it seems that better molecular scale rectifier can be achieved for devices with smaller mesa width. Analyze the effect of lateral leakage current on the rectifier performance.
(2) As the authors mentioned and depicted in Fig. S5b, the real electrode contact area for the device with mesa width Wmesa = 10 µm should be much less than 100 µm2. Then a forward current density can be much higher than ~ 200 A/cm2 at +2 V. Under such high current across very thin (~8nm) organic layer, one can expect quite high Joule heating, resulting in increased junction temperature. And the effect of Joule heating and junction-temperature rise should depend on the width of mesa. Since the charge carrier injection and conduction depend on the temperature, it is necessary to address the effect of Joule heating on the device I-V characteristics and the rectification ratio.
(3) The charge carrier conduction in the Au/CuPC/Au and Au/F16CoPC/CuPC/Au devices is described by the space-charge-limited current (SCLC) transport mechanism. However, one can expect the current conduction is contact-limited, rather than the bulk-limited, as the thicknesses of organic films (CuPc and F16CoPC/CuPC) are very small, a few nanometers.
(4) The devices show excellent stability of nearly one month in air, as shown in Fig. S6. The authors attributed to the stable properties of Au electrodes and organic molecules (F16CoPc and CuPc). But the Au electrodes are contaminated with H2O, NO2, CO2, and O2, as the authors mentioned. Then the contact properties are expected to change over time in the air.
(5) It would be better to show the yield of devices (ratio of successful rectifiers among devices in an array in a chip) and statistical analysis of rectification ratio and 3dB frequencies for devices. (6) In Fig. 2e "hole accumulation" and "electron accumulation" should be interchanged; i.e., "hole accumulation" in CuPC and "electron accumulation" in F16CoPc. (7) In line 108, he ◊ The (8) In Figure S2. d, Relative contents of Co(I), Co(I) and satellite peak … ◊ … of Co(I), Co(II) and satellite peak … (9) In line 207-208, "our diode" is redundantly used.

Response to the comments of Reviewer 1 Our response:
Thank you very much for your efforts on our manuscript. We greatly appreciate your advices and comments which are very helpful for improving the quality of this work. We have revised the manuscript according to your suggestions. The revised parts are marked by blue font. Our response is presented point-to-point in the following.

Question (1.1)
The Title is misleading as 10 MHz is usually considered as relatively low-frequency by the high-frequency electronics community. Therefore, I suggest that "10 MHz" should replace "High-frequency" as for example: "All printed diode operating at 1.6 GHz", PNAS 111, 11943 (2014)

Response (1.1)
Thank you very much for this suggestion. We agree that, although this is the first time to achieve a nanometer-thick integrated organic rectifier with an operation frequency beyond 1 MHz, the obtained maximum 3db operation frequency (~10 MHz) is still in a frequency range considered as "low" by the electronics community. Hence, according to your suggestion, we changed the title into "Integrated Molecular Diode as 10 MHz Half-Wave Rectifier Based on an Organic Nanostructure Heterojunction".  Fig.5d.

Response (1.2)
We appreciate this comment. It should be pointed out that the claimed 17 GHz molecular rectifier in Ref.11 is a different type of rectifier compared to our work. In Ref. 11, a Pt tip was biased to the superposition of direct current (DC) and radio frequency (RF) excitation simultaneously (shown in Figure R1a). The performance index is the ratio of reflected RF signal power to the incident RF signal power during the DC scan (i.e., S-parameter S11) 1 . In other words, this kind of rectifier is to rectify the reflected RF signal power which is related to the DC-affected impedance. But the rectifier in our work is a kind of half-wave rectifier, which aims at converting single alternating current (AC) input into the DC output. Please see the comparison  Figure R1b). This is the reported state-of-the-art molecular scale rectifier device which

Response (1.3)
Thank you very much for these comments. In our work, there exists a thickness-related tradeoff between current density and rectification ratio as shown in Figure R2  are in the same range of the molecular film thickness (shown in Figure R3 or Supplementary Fig. 6 in the Supplementary Information). The thinner the molecular layer is, the more likely the device gets shorted or breakdown because the effective distance between the two electrodes reduces from the roughness of the Au finger and tube electrodes, leading to a very low yield of successful rectifier devices. On the other hand, for the devices with thicker molecular layer (for instance, 15 nm), the rectification ratio is still maintained as high as about 2 orders, however the forward current is too low to act as an efficient rectifier. Therefore, we chose an 8 nm thick F 16 CoPc/CuPc hybrid layer to investigate the electrical characteristic, which has not only high rectification ratio but also high forward current density. We have supplemented the corresponding discussions on  Average currents (at 2 V) of devices based on W design = 5 and 10 μm.
In order to investigate the capacitive network, we tried to obtain the detailed equivalent circuit by using impedance spectroscopy (partial data are shown in Figure R5a and b). However, the impedance data from different measurements on the same device are not consistent, and they could not be fitted well due to rather randomly distributed data points. We think it is caused by the robust but imperfect mechanical contact between the rolled-up tubular electrode and ultrathin molecular ensembles, as illustrated in Figure R5c. Owing to the relatively rough surfaces of the CuPc film and Au tube, there are many local tiny contacting areas with ill connection or air gap, resulting in the superposition of many individual parallel RC circuits placed between the CuPc and Au tube electrode (shown in Figure R5d). Furthermore, when a DC or AC crosses the device, the involved local high current density and charging/discharging processes might result in local tiny deformation and mechanical vibrations of the tube electrode. As a result, the complicated physical scene and electric process make it difficult to obtain a constant and accurate equivalent circuit in detail. Therefore, we thank you for providing the clue of analysis, and we plan to carry out more detailed investigations and improvements in a future study.

Response to the comments of Reviewer 3
Our response: Thank you very much for your efforts on our manuscript. We greatly appreciate your advices and comments which are very helpful for improving the quality of this work. We have revised the manuscript according to your suggestions. The revised parts are marked by blue font. Our response is presented point-to-point in the following. Figure S5a However, it appears that currents are at least 4 or 5 times higher for devices with mesa width Wmesa = 10 μm compared to Wmesa = 5 μm. This may suggest that there exists a lateral leakage current.

Question (3.1) In
Then it seems that better molecular scale rectifier can be achieved for devices with smaller mesa width. Analyze the effect of lateral leakage current on the rectifier performance.

Response (3.1)
Thank you very much for your comments. In the device fabrication, the designed mesa width (W design ) was defined by the lithography patterns. However, the real width (W real ) is influenced by the isotropic under-etching by HF solution during the mesa formation (shown in Figure R6a  Average currents (at 2 V) of devices based on W design = 5 and 10 μm. Fig. S5b

Response (3.2)
Thank you very much for the comment. It is challenging to avoid the critical Joule heating effect of molecular-scale electronics. As shown in Figure R7a-b (Supplementary Fig. 9a-

Response (3.3)
Thank you very much for this comment. The CuPc/Au (tube) interface formed by the robust mechanical contact can block the hole injection from the Au tube into the molecular layer, however, the hole crossing from the molecular layer to the Au tube is not blocked. This interface is mainly responsible for the function of rectification, as demonstrated by the I-V characteristics (Figure 4a).
In other words, under forward bias condition the charge transport (from Au finger to Au tube electrode) is not limited when compared to the reverse bias condition (from Au tube to Au finger electrode). To further clarify the charge transport process in forward direction, apart from the SCLC model 3 , the forward-direction currents are also fitted with two possible contact-limited conduction models (i.e., Fowler-Nordheim tunneling and Schottky emission 4 ), as shown in Figure R8 ( Supplementary Fig. 10 in the Supplementary Information). As we can see, only the SCLC model fits well, which exhibits the typical three transport regions with noticeable different slopes (shown in

Question (3.4)
The devices show excellent stability of nearly one month in air, as shown in Fig. S6.

Response (3.4)
Thank you very much for this comment. Considering the practical scenario of the device formation, the contamination with H 2 O, NO 2 , CO 2 , and O 2 , happened to the Au electrodes and molecular layers during the multi-step fabrication process, which was revealed by the UPS and XPS results. We think once the devices are formed, a certain influence of further contamination in the stable environment atmosphere will still take place, but will not be significant enough to dramatically change the device performance over a short time. To demonstrate the change of contact properties over a long time in air, after 420 days since the first measurement, the I-V characteristics of a device is measured again.
It is found that both the forward and reverse currents increased slightly, while the rectification ratio decreased a little, as shown in Figure R9 (Supplementary Fig. 9c in the Supplementary   Information). This may be ascribed to the oxygen doping 7 and/or further contamination happening to the device 8 . A discussion is supplemented in the revised manuscript on Page 9 in Lines 223-224 and on Page 16 in the Supplementary Information (Supplementary Note 3).

Question (3.5) It would be better to show the yield of devices (ratio of successful rectifiers among devices in an array in a chip) and statistical analysis of rectification ratio and 3dB frequencies for
devices.

Response (3.5)
Thank you very much for this comment. The construction of our rectifiers is based on the rolled-up nanomembrane which contacts the ultrathin molecular ensembles from the top, thus providing a damage-free and self-adjusted electrode. Any defects in sacrificial layer or strain layer could lead to the failure of rolling. For a 96 device array, the yield of initial devices with successful rolled-up contacts is about 80%, as shown in Figure R10a   I appreciate the efforts made by the authors that include a clearer title, abstract, additional datas/discussion on the film thickness and the fair comments on the capacitive network. I believe that the paper deserves to be published in Nature Communications as it can be used as a clear performance reference for future studies from both molecular electronics and organic electronics fields.
Overall, it is a nice piece of work.
I still suggest some minor modifications: * The Intro is still not clear enough, at least from a broad audience perspective.
-For example, it is mentioned that such devices could operate up to THz range, but it is not clearly mentioned what demonstrations have been obtained so far towards this target, and their limitations. One important reference is "Quantum Plasmon Resonances Controlled by Molecular Tunnel Junctions" Science, 343, 6178 (2014), as it is shown that quantum tunneling through molecules is clearly operational in the range 140-240 THz. This demonstration does not include the "rectification fucntion" though.
Ref 11 demonstrates that the rectification function is still working until 17 GHz, but not in an integrated device configuration. [comment to the difference in current and S11: authors suggest in their answer to referee that current and S11 are very different, but in fact the experimental setup using alternating current in this study is limited to relatively low frequency. If the authors want to make their device operate in the GHz range or higher in the future, they will likely have to consider the S parameters approach. One ref from P.S Weiss with is AC-STM on molecules (GHz) would have been fair as well.
-Related: In the intro, the sentence "However, these techniques are unable to exploit the intrinsic nanoscale properties ..."; this sentence is too aggressive or ambiguous. As a matter of fact, the Nature communication paper "Controlling the direction of rectification in a molecular diode" by Nijuis et al. nicely exploits intrinsic nanoscale properties, with and EGaIn approach. Maybe the authors wanted to mention: Large area techniques including GaIn tend to reduce the electronic coupling to the upper part of the junction, which reduces frequency range operation. Therefore, while a GHz range molecular diode has been demonstrated using an AFM based technique, one important goal is to demonstrate functional molecular rectifiers operating at high frequency using an integrated top electrode.
* The capacitive network => I understand that precise estimation is difficult, but just a // plate configuration, taking a dielectric constant of 2.2, the thickness and junction area would give a rough approximation for the capacitance, and the theoretical cutt-off frequency taking into consideration the capacitances.
* In the discussion, just a prediction of the theoretical limitation for this current device and future perspective/suggestions for the future devices is still missing. For example, in "Molecular diodes, Breaking the Landauer limit", Nature Nano 12, 725(2017), it is mentioned that we have to consider some tradeoff between rectifictaion and high-frequency operation (if the electronic coupling is large, frequency operation is large, but rectification ratio is degraded; this is the opposite for weak coupling). The device proposed by the authors is a good example of nice tradeoff (good rectification ratio and high frequency of operation) => it is worth mentioning.

Response to the comments of Reviewer 1
Our response: Thank you very much for your efforts on our manuscript. We greatly appreciate your affirmation and encouragement on our work, and your advice is very helpful for improving the quality of this work.
We have revised the manuscript according to your suggestions. The revised parts are marked by blue font. Our response is presented point-to-point in the following. -Thank you very much for these comments, especially for the suggestion about the S parameter approach which provides us with a valuable clue in future. In order to improve the introduction part and include the previous efforts made to address the THz target in the molecular electronics field, we revised the manuscript (on Page 2 in Lines 43-50).
-Thank you for pointing out the impropriety of the sentence "However, these techniques are unable to exploit the intrinsic nanoscale properties ...", we revised the manuscript (on Page 3 in Lines 59-65).

Question (1.2)
The capacitive network => I understand that precise estimation is difficult, but just a // plate configuration, taking a dielectric constant of 2.2, the thickness and junction area would give a rough approximation for the capacitance, and the theoretical cut-off frequency taking into consideration the capacitances.

Response (1.2)
We appreciate this comment. According to your suggestion, we calculated the approximate capacitance and the theoretical cut-off frequency of our molecular diode, and revised the manuscript (on Page 15 in Lines 366-374). Here, we assume a dielectric constant of 4 for phthalocyanine materials, as this has been reported in previous studies, for instance, Ref. 52 and53.

Question (1.3)
In the discussion, just a prediction of the theoretical limitation for this current device and future perspective/suggestions for the future devices is still missing.
For example, in "Molecular diodes, Breaking the Landauer limit", Nature Nano 12, 725(2017), it is mentioned that we have to consider some tradeoff between rectifictaion and high-frequency 5 operation (if the electronic coupling is large, frequency operation is large, but rectification ratio is degraded; this is the opposite for weak coupling).

Response (1.3)
Thank you very much for this suggestion. In the Discussion part, through analyzing the related parameters, we provide some opinions and perspectives for the future devices, and revised the manuscript (on Page 15 in Lines 365-378 and Page 16 in Lines 389-392).