Rational design of Al2O3/2D perovskite heterostructure dielectric for high performance MoS2 phototransistors

Two-dimensional (2D) Ruddlesden-Popper perovskites are currently drawing significant attention as highly-stable photoactive materials for optoelectronic applications. However, the insulating nature of organic ammonium layers in 2D perovskites results in poor charge transport and limited performance. Here, we demonstrate that Al2O3/2D perovskite heterostructure can be utilized as photoactive dielectric for high-performance MoS2 phototransistors. The type-II band alignment in 2D perovskites facilitates effective spatial separation of photo-generated carriers, thus achieving ultrahigh photoresponsivity of >108 A/W at 457 nm and >106 A/W at 1064 nm. Meanwhile, the hysteresis loops induced by ionic migration in perovskite and charge trapping in Al2O3 can neutralize with each other, leading to low-voltage phototransistors with negligible hysteresis and improved bias stress stability. More importantly, the recombination of photo-generated carriers in 2D perovskites depends on the external biasing field. With an appropriate gate bias, the devices exhibit wavelength-dependent constant photoresponsivity of 103–108 A/W regardless of incident light intensity.

(i) Accumulation of photogenerated holes at the Al2O3/perovskite interface leads to an electron current in the MoS2 through an electrostatic gating effect. If this is the correct understanding, then what is the motivation for choosing MoS2, why not Si or any other conventional semiconductor with much higher electron mobility than MoS2?
(ii) The photoresponse at 914 nm is beyond the absorption edge of the perovskite-the authors attribute this to excitation from valence band to trap states (lines 136-140). (iv) The authors claim that the hysteresis due to Al2O3 cancels that due to the 2D perovskite. However, voltage shifts due to dielectric charge need not be simply additive/subtractive, since the shift does not depend on charge density but the moment of charge density. Hence the contribution to hysteresis due to positive ionic charges in the perovskite will be less once it is placed on top of Al2O3. Comparing the hysteresis in Fig. S5 due to perovskite (n=3) alone and due to Al2O3 alone ( Fig. 2a), it seems unlikely that these will cancel out in the composite stack. Please clarify.
(v) Comparing the dark transfer curves (Ids-Vgs) in Fig. 2a for Al2O3/n=3, Al2O3 only and only n=3 in Fig. S5 one comes up with the following inconsistencies/questions: (i) The threshold voltage of n=3 is nearly 0 V, that of only Al2O3 is ~-2 V and that of Al2O3/n=3 is nearly -3 V. This seems to indicate net fixed positive charge in Al2O3. How did it increase with n=3 that seems to have no net fixed charge (VT ~ Vflatband)? (ii) n=3 has poor sub-threshold slope (visually) and on-current. How did these improve with the addition of a 9 nm dielectric in series? Please specify quantitative values of carrier mobility, subthreshold slope, VT, hysteresis, Ion and Ioff for all three key transistors-Al2O3/n=3, Al2O3 only and only n=3, maybe in a table in the supplementary information.
(vi) What could be the reason for a high gamma value of 2.3?
(vii) The energy band diagram in Fig. 4a under Vg < VRc does not seem to be correct. As per Fig.  3a the transistor is below threshold (close to being off) at large negative Vg. In this case the band bending should be in the opposite direction than what is shown, and MoS2 is depleted (not accumulated as shown). The electric field in the perovskite will be in the opposite direction leading to electron accumulation at Al2O3/perovskite interface.
(viii) The explanation for VRC crossover is not clear-if it is due to increasing electric field in the perovskite, what determines the critical electrical field-corresponding to VRC, in terms of perovskite material parameters, where the light intensity dependence of R flips over? Further, if a built-in electric field due to hole accumulation opposes further hole drift to the interface (Fig. 4f), shouldn't the R value saturate for increasing light intensity?
(ix) Can the authors show stability of optical photoresponse over a period of time? (x) There are a large number of writing errors, the manuscript needs thorough proof reading. A few examples: (a) line 206 -"stack" should be "stacked" (b) line 168-"charges" should "charge" (c) line 92 -"consisted" should be "consisting" Reviewer #2 (Remarks to the Author): In this work, the authors present a simple yet universal Al2O3/2D perovskite heterostructure dielectric for high-performance phototransistors. The type-II energy band alignment in 2D perovskite facilitates the effective spatial separation of photoexcited carriers, thus inducing the outstanding photoresponse properties. By adding a high-k Al2O3 layer between the 2D perovskite layer and the semiconductor layer, the two effects of charge trapping and remnant polarization under gate bias are found to be neutralized with each other, resulting in low-voltage transistors of negligible hysteresis and improved reliability. The devices also exhibit excellent linear dynamic characteristics.
Overall, I think this paper is sufficiently innovative and the experiments are nicely executed. I think the work will be of interest to diverse researchers working on perovskite materials and 2Dmaterial phototransistors and I would recommend its publication in Nature Communications after the following questions are clarified.
(1) Although there has been significant progress on the development of perovskite materials in recent years, the poor stability is still a grand challenge yet to be tackled. How's the stability of the devices reported in this paper and what are the possible approaches to improve the lifetime of perovskite-based devices? Some discussions would be helpful.
(2) The device is capable of broadband photodetection ranging from 457 nm to 1064 nm. However, the absorption edge of the upper perovskite layer is located at around 800 nm. Is the MoS2 layer responsible for the near-infrared response?
(3) Page 12, Line 5, the authors estimated the Ilight/Idark ratio to be 2.6 × 10^6 at Vgs = -3.5 V where the incident power density (Plight) is ~1 mW/cm2. The electrical measurements were carried out by varying the intensity from ~0.01 -~1 mW/cm2. In this case how can the author say that Plight is as low as 1 mW/cm2? (4) What are the channel length and width of the phototransistors? (5) First of all, Fig. 2b has not been mentioned in the manuscript. Furthermore, I assumed the arrows in Fig. 2b is for double sweep to show hysteresis (or lack thereof) in the Ids-Vds characteristics of the phototransistors. Usually people present hysteresis on transfer Ids-Vgs characteristics of the devices (e.g. Fig. 2a of the paper). It is not clear to me why it is necessary to also present it for the Ids-Vds characteristics.
Reviewer #3 (Remarks to the Author): Jiang et al. have demonstrated the fabrication of a high performance MoS2 phototransistor employing a novel Al2O3/2D perovskite hybrid dielectric configuration. The results highlight the achievement of both ultrahigh responsivity and excellent linear dynamic range in one device. In addition, this work on Al2O3/2D perovskite dielectric to achieve greatly improved operational stability is very crucial for the technology to be applied in practical fields. I appreciate the experimental work which is systematically performed and discussed. Considering that the overall quality and significance of this work, I would like to recommend this work for publication after the authors address the following questions: 1. In the Al2O3/2D perovskite hybrid dielectric, the thickness of Al2O3 layer is 9 nm. In my understanding, the Al2O3 layer is employed as a high-k dielectric, and the thinner thickness could help to enhance the gate capacitance. Does the Al2O3 thickness influence the performance of the device? 2. In Figure 3d, the photoresponsivity remains almost constant within the light power density ranging from 7.2 to 1232.2 μW/cm2. What is the R-squared (coefficient of determination) of the linear fitting for the device? 3. Why can the devices detect near-infrared light? The perovskites are usually employed for visible light detection, which is in accordance with the absorption spectrum shown in Figure S1. 4. Line 8, Page 3. the sentence "...is also become more eminent." should be corrected; Line 7, Page 10. "...and S denotes the active area of photodetector, ..." Compared to the illumination area, which is larger in the experiment setup? The authors should specify this value in the manuscript. 5. Line 16-18, Page 12. "In general, the γ value is in the range of 0 < γ <1 for photogating devices, ...photoconductive effect31. " The author should elaborate more on this point, as this value depends also on the surface defects and quality of film. 6. In Figure 2a and Figure 5a, what is the source-drain voltage used in the measurements?

Response to Reviewer # 1:
We thank the referee for careful reading the manuscript and providing a number of precious comments. And we have addressed the comments carefully with listed response below. Meanwhile we have revised our manuscript accordingly.

Lei Liao et al. report high optoelectronic performance MoS 2 transistors using a hybrid Al 2 O 3 /2D perovskite gate dielectric approach. Specifically, they attempt to show how net hysteresis close to zero can be achieved by choosing the perovskite layer such that the anti-clockwise hysteresis due to ionic movement in the perovskite cancels out the clockwise hysteresis due to negative charge trapping in the Al 2 O 3 dielectric. They also show stable dark current with the hybrid dielectric, high photoresponse over visible-near IR range, good LDR and detectivity, and switching speed of the order of 20-30 ms. Although the photoresponsivity values, LDR and broad spectral response metrics are impressive, not as much the switching speed, there are substantial concerns regarding the motivation, proposed mechanism and supporting data. The claims are novel but not scientifically convincing as outlined in the questions below. 1. Accumulation of photogenerated holes at the Al 2 O 3 /perovskite interface leads to an electron current in the MoS 2 through an electrostatic gating effect. If this is the correct understanding, then what is the motivation for choosing MoS 2 , why not Si or any other conventional semiconductor with much higher electron mobility than MoS 2 ?
We thank for the referee's detailed comments on these viewpoints. In the device architecture, a conducting channel with high electron mobility would lead to the improved photoresponsivity. But as the scale and diversity of application areas grow, the needs for a photodetection platform with higher integration and lower power consumption are becoming more eminent. In comparison with conventional semiconductors, the ability to control the MoS 2 thickness at the atomic level translates into improved gate control over the channel barrier and into reduced short-channel effects, which are beneficial for ultrascaled devices with low power consumption 1 . Fig. 1e? Figure 1. Schematic of the photo-generated carriers transfer process in spin-coated 2D perovskite films with type-II energy band alignment.

The photoresponse at 914 nm is beyond the absorption edge of the perovskite the authors attribute this to excitation from valence band to trap states (lines 136-140). (a) Isn't this inconsistent with Fig. 1e where electrons are shown in the conduction band? Where are the trap states in this picture? (b) If photogenerated electrons are captured by traps from the valence band leading to photogenerated holes in the valence band-how do these electrons move towards the gate electrode under external gate bias as shown in
(b) In this case, the photo-generated holes tend to move towards Al 2 O 3 /2D perovskite interface. The spatial separation of photo-generated carriers in 2D perovskite would lead to an increase in the electric field at the MoS 2 , which could increase the carrier density in the MoS 2 , thus enhancing the conductivity 2 . We have corrected the relevant discussion in the manuscript.

What are the mechanisms behind recombination of photogenerated holes in 2D perovskite and electrostatically induced "supplemental" electrons in MoS 2 -in their respective layers?
We are sorry for the confusion. The carrier recombination is probably associated with trap-assisted recombination through the traps or defect levels in Al 2 O 3 or perovskite layer 3 . We have added the relevant discussion in the manuscript. Fig. S5 due to perovskite (n=3) alone and due to Al 2 O 3 alone (Fig. 2a), it seems unlikely that these will cancel out in the composite stack. Please clarify.

The authors claim that the hysteresis due to Al 2 O 3 cancels that due to the 2D perovskite. However, voltage shifts due to dielectric charge need not be simply additive/subtractive, since the shift does not depend on charge density but the moment of charge density. Hence the contribution to hysteresis due to positive ionic charge in the perovskite will be less once it is placed on top of Al 2 O 3 . Comparing the hysteresis in
We thank for the referee's detailed comments and sorry for the confusion. In our experiments, a clockwise hysteresis loop is observed for the device using Al 2 O 3 gate dielectric, which is caused by the trapping of carriers from the gate bias-induced conduction channel into less mobile localized states. Meanwhile, the fabricated phototransistor using 2D perovskite gate dielectric presents hysteresis loop opposite to that of the device with Al 2 O 3 gate dielectric layer. The physics mechanism is considered to be that the electric field from ionic polarization in 2D perovskite induces additional carriers into the channel. Here, the hysteresis due to perovskite (n = 3) alone in the Fig. S5 is associated with the applied gate bias (the Fig. S5 has been merged into Supplementary Fig. 4b). In order to avoid the breakdown of the perovskite (n = 3) alone, the applied gate bias is limited. Accordingly, it is hard to compare the hysteresis due to perovskite (n = 3) alone and due to Al 2 O 3 alone. By adding a thin Al 2 O 3 layer between the 2D perovskite and the semiconductor layer, the two effects of charge trapping and ionic polarization under gate bias could be neutralized with each other, resulting in negligible hysteresis in MoS 2 phototransistor.

Comparing the dark transfer curves (I ds -V gs ) in Fig. 2a for Al 2 O 3 /n=3, Al 2 O 3 only and only n=3 in Fig. S5 one comes up with the following inconsistencies/questions: (i) The threshold voltage of n=3 is nearly 0 V, that of only Al 2 O 3 is ~-2 V and that of Al 2 O 3 /n=3 is nearly -3 V. This seems to indicate net fixed positive charge in Al 2 O 3 . How did it increase with n=3 that seems to have no net fixed charge (V T ~ V flatband )? (ii) n=3 has poor sub-threshold slope (visually) and on-current. How did these improve with the addition of a 9 nm dielectric in series? Please specify quantitative values of carrier mobility, subthreshold slope, V T , hysteresis, I on and I off for all three key transistors-Al 2 O 3 /n=3, Al 2 O 3 only and only n=3, maybe in a table in the supplementary information.
(i) We thank for the referee's detailed comments and sorry for the confusion. Here, in order to avoid the error caused by different MoS 2 flakes, the phototransistors are fabricated with the same MoS 2 flake (Figure 2a). The device with Al 2 O 3 dielectric exhibits a negative threshold voltage (Figure 2b). Meanwhile, the V th value with Al 2 O 3 dielectric is similar to that of Al 2 O 3 /2D perovskite dielectric. This result can be attributed to the giant dielectric constant phenomenon in 2D perovskite at low frequency (Figure 2c), which is probably caused by intrinsically polarizability of perovskite 4 . In typical series capacitance geometry, the total capacitance is given by: where t Al2O3 is the Al 2 O 3 thickness, and ρ(x) is the fixed positive charge density in Al 2 O 3 . In addition, the V th value for 2D perovskite dielectric is nearly 0 V, which is probably due to p-type doping effect in the MoS 2 channel and effective electrostatic control. (ii) We thank for the referee's valuable suggestion. The key performance parameters of the devices are summarized in Table S1. Here, in comparison with Al 2 O 3 and Al 2 O 3 /2D perovskite dielectrics, the device with 2D perovskite dielectric exhibits a much lower electrical performance. It is well known that the interface quality plays a crucial role in the carriers transport of 2D semiconductor devices 1 . This result indicates the severe carrier scattering involving surface roughness and Columbic impurity scattering at MoS 2 /2D perovskite interface 5 . Table S1. Comparison in device performance with different dielectric.

What could be the reason for a high gamma value of 2.3?
The high gamma value of 2.3 represents that the photoresponsivity increases with increased incident power, which is similar to other reported photodetector 6 . Under a strong light excitation, photon penetration depth increases, and part of photo-generated holes can drift to Al 2 O 3 /2D perovskite interface instead of recombination, leading to increased photoresponsivity. Fig. 4a under V g < V Rc does not seem to be correct. As per Fig. 3a the transistor is below threshold (close to being off)

at large negative Vg. In this case the band bending should be in the opposite direction than what is shown, and MoS 2 is depleted (not accumulated as shown). The electric field in the perovskite will be in the opposite direction leading to electron accumulation at Al2O3/perovskite interface.
We appreciate for the reminder. We have corrected the energy band diagram in Fig. 4a. In view of the natural energy band alignment resulting from the varying 2D perovskite phases (Fig. 1e in the manuscript), the 2D perovskite band could bend upward at Al 2 O 3 /2D perovskite interface under V g < V Rc . Thus, the photo-generated holes can accumulate at Al 2 O 3 /2D perovskite interface and lead to photogating effect.  (Fig. 4f)

, shouldn't the R value saturate for increasing light intensity?
We are sorry for the confusion. In view of the charge spatial separation throughout the vertical direction of the perovskite film, the V Rc crossover is probably determined by the band bending induced built-in electric field in 2D perovskite. In principle, the V Rc value can be modulated through controlled composition variation in 2D perovskite. In addition, because the hole accumulation opposes further hole drift to Al 2 O 3 /2D perovskite interface (Fig. 4f), the R value tends to saturate as the increase of the laser power density according to the equation, R = I ph /P light S ~ P light γ-1 (γ < 1).

Can the authors show stability of optical photoresponse over a period of time?
We appreciate for the reminder. The stability of the phototransistor with Al 2 O 3 /2D perovskite dielectric is investigated. The photoresponsivity retain 98% and 89% of its original value after exposing in nitrogen and air environment for more than 160 hours, respectively. It is known that perovskites usually exhibit poor stability. Here, the remarkable stability of the device can be attributed to the hydrophobicity of the organic spacer in 2D perovskite 7,8 , which shields the inner perovskite layers from moisture.

There are a large number of writing errors, the manuscript needs thorough proof reading. A few examples: (a) line 206 -"stack" should be "stacked" (b) line 168-"charges" should "charge" (c) line 92 -"consisted" should be "consisting"
We thank for the reminder and sorry for the mistakes. We have corrected them and revised the manuscript carefully.
Overall, we thank the referee for the detailed comments and suggestions, which help to greatly improve our manuscript. We believe the photoactive dielectric presented here will make significant contribute to high-performance perovskite-based photodetection for practical utilizations.

Response to Reviewer # 2:
We thank the referee for careful reading the manuscript and providing a number of precious comments. And we have addressed the comments carefully with listed response below. Meanwhile we have revised our manuscript accordingly.

In this work, the authors present a simple yet universal Al 2 O 3 /2D perovskite heterostructure dielectric for high-performance phototransistors. The type-II energy band alignment in 2D perovskite facilitates the effective spatial separation of photoexcited carriers, thus inducing the outstanding photoresponse properties. By adding a high-k Al 2 O 3 layer between the 2D perovskite layer and the semiconductor layer, the two effects of charge trapping and remnant polarization under gate bias are found to be neutralized with each other, resulting in low-voltage transistors of negligible hysteresis and improved reliability. The devices also exhibit excellent linear dynamic characteristics. Overall, I think this paper is sufficiently innovative and the experiments are nicely executed. I think the work will be of interest to diverse researchers working on perovskite materials and 2D-material phototransistors and I would recommend its publication in Nature Communications after the following questions are clarified. 1. Although there has been significant progress on the development of perovskite materials in recent years, the poor stability is still a grand challenge yet to be tackled. How's the stability of the devices reported in this paper and what are the possible approaches to improve the lifetime of perovskite-based devices? Some discussions would be helpful.
We thank for the referee's detailed comments on these viewpoints. Although a huge number of articles about perovskites have been reported, improving stability of perovskites is still a grand challenge yet to be tackled. There are three main intrinsic factors leading to perovskite instability: hygroscopicity, thermal instability, and ionic migration 1 . In previous studies, several strategies have been proposed to improve device stability. The hygroscopicity can be solved by encapsulation 2 . The thermal instability can be addressed by composition tuning to increase the decomposition energy or barrier 3 . Lastly, the issue of ionic migration is currently treated by A site alkali doping and replacement 4,5 , multiple dimensional perovskites engineering 6 , and organic molecular additives 7 . Therefore, we are optimistic that perovskites would perform extraordinarily well outside in an unshaded location for a long time. In our experiment, the photoresponsivity retain 98% and 89% of its original values after exposing in nitrogen and air environment for more than 160 hours, respectively. The remarkable stability of the device can be attributed to the hydrophobicity of the organic spacer 8,9 , which shields the inner perovskite layers from moisture. We have added the relevant discussion in the manuscript.

The device is capable of broadband photodetection ranging from 457 nm to 1064 nm. However, the absorption edge of the upper perovskite layer is located at around 800 nm. Is the MoS 2 layer responsible for the near-infrared response?
We thank for the reminder. As shown in Fig. 1f, the near-infrared photoresponsivity of the pristine MoS 2 device is several orders of magnitude less than the MoS 2 /Al 2 O 3 /2D perovskite device. Therefore, the impressive near-infrared response is mainly caused by the 2D perovskite layer. This is probably due to the efficient excitation of carriers from the valence band to the trap states within the perovskite bandgap, being similar to the extrinsic photoconductors based on conventional semiconductors 10 .

Page 12, Line 5, the authors estimated the I light /I dark ratio to be 2.6 × 10 6 at V gs = -3.5 V where the incident power density (P light ) is ~1 mW/cm 2 . The electrical measurements were carried out by varying the intensity from ~0.01 -~1 mW/cm2. In this case how can the author say that P light is as low as 1 mW/cm 2 ?
We thank for the referee's detailed comments. The I light /I dark ratio is a typical parameter employed to evaluate the degree of obtrusive noise. Generally, the I light /I dark ratio increases with increased incident power density. In comparison with previously reported photodetectors, the P light of 1 mW/cm 2 is a relatively low value 11 . The high I light /I dark ratio of 2.6 × 10 6 arises from the profound photogating effect in our device.

What are the channel length and width of the phototransistors?
We thank for the reminder. The channel length is 3 μm, and the channel width is 3-10 μm. We have added the relevant description in the manuscript. (e.g. Fig. 2a of the paper).

It is not clear to me why it is necessary to also present it for the I ds -V ds characteristics.
We thank for the reminder. We have added the relevant description in the manuscript. The output curves in Fig. 2b indicate the good ohmic contacts between the Cr/Au electrodes and the MoS 2 . In addition, the trapping of electrons at MoS 2 /Al 2 O 3 interface or migration of ionic vacancies in 2D perovskite can also induce the hysteresis phenomenon during I ds -V ds characteristic measurement. Here, the negligible hysteresis indicates the operational stability of the device.
Overall, we thank the referee for the detailed comments and suggestions, which help to greatly improve our manuscript. We believe the photoactive dielectric presented here will make significant contribute to high-performance perovskite-based photodetection for practical utilizations.

Response to Reviewer # 3:
We thank the referee for careful reading the manuscript and providing a number of precious comments. And we have addressed the comments carefully with listed response below. Meanwhile we have revised our manuscript accordingly.
Jiang et al. have demonstrated the fabrication of a high performance MoS 2 phototransistor employing a novel Al 2 O 3 /2D perovskite hybrid dielectric configuration. The results highlight the achievement of both ultrahigh responsivity and excellent linear dynamic range in one device. In addition, this work on Al 2 O 3 /2D perovskite dielectric to achieve greatly improved operational stability is very crucial for the technology to be applied in practical fields. I appreciate the experimental work which is systematically performed and discussed. Considering that the overall quality and significance of this work, I would like to recommend this work for publication after the authors address the following questions: 1. In the Al 2 O 3 /2D perovskite hybrid dielectric, the thickness of Al 2 O 3 layer is 9 nm. In my understanding, the Al 2 O 3 layer is employed as a high-k dielectric, and the thinner thickness could help to enhance the gate capacitance. Does the Al 2 O 3 thickness influence the performance of the device?
We thank for the referee's detailed comments on these viewpoints. In our device architecture, the 2D perovskite cover the Al 2 O 3 layer completely. Although the thinner Al 2 O 3 layer is beneficial for low-voltage phototransistor operation, it could induce an enlarged leakage current. Therefore, the thickness of Al 2 O 3 layer in our experiments is optimized to 9 nm. 3. Why can the devices detect near-infrared light? The perovskites are usually employed for visible light detection, which is in accordance with the absorption spectrum shown in Figure S1.
As shown in Fig. 1f, the near-infrared photoresponsivity of the pristine MoS 2 device is several orders of magnitude less than the MoS 2 /Al 2 O 3 /2D perovskite device. Therefore, the impressive near-infrared photoresponse is mainly caused by the 2D perovskite layer, which is probably due to the efficient excitation of carriers from the valence band to the traps states within the bandgap, being similar to the extrinsic photoconductors based on conventional semiconductors 1 . We thank for the reminder and sorry for the mistake. We have corrected the description in the manuscript. In addition, the active area of photodetectors represents the area of MoS 2 channel, which is slightly smaller than the illumination area. We have added the relevant description in the manuscript.

5.
Line 16-18, Page 12. "In general, the γ value is in the range of 0 < γ <1 for photogating devices, ...photoconductive effect." The author should elaborate more on this point, as this value depends also on the surface defects and quality of film.
We thank for the referee's reminder. We have corrected the description in the manuscript, "The nonunity exponent of 0 <γ<1 is often observed in photogating devices 2,3 , as a result of the complex process of carrier generation, trapping, and recombination within semiconductors. The photocurrent tends to saturate as the increase of the laser power, which is partly due to the gradually filled trap states. The smaller γ value represents the more prominent photogating effect, while γ = 1 represents the pure photoconductive effect." Figure 2a and Figure 5a, what is the source-drain voltage used in the measurements?

In
We appreciate for the reminder. The source-drain voltage is set to 1 V. We have added the relevant description in the manuscript.
Overall, we thank the referee for the detailed comments and suggestions, which help to greatly improve our manuscript. We believe the photoactive dielectric presented here will make significant contribute to high-performance perovskite-based photodetection for practical utilizations.  Table S1. Comparison in device performance with different dielectrics. Fig. 4a for MoS 2

based on my inputs. But, to stay consistent with their mechanisms, they had to keep the band bending direction the same in the 2D perovskite, leading to a discontinuity in electric field (hence new questions on interface charge etc.) at the Al 2 O 3 /perovskite interface. This has led to them to come up with highly speculative reasons (such as 2D perovskite phases) to justify the band bending in the perovskite.
We thank for the referee's detailed comments and sorry for this mistake. In the band diagram shown in Fig. 4a in the manuscript, the gate bias is below threshold but higher than the voltage where the device is in off state. At this point, the diffusion current in MoS 2 is caused by gradient of electron concentration. Therefore, the band diagram for MoS 2 in the previous version was correct. We have replaced it with the initial version ( Figure 2). In order to avoid confusion, we have changed the description "a Energy band diagram of the device under a gate bias lower than V Rc " into "a Energy band diagram of the device under a gate bias below V Rc but higher than the voltage where the device is in off state" in revised manuscript.

Figure 2.
Energy band diagram of the device under a gate bias below V Rc but higher than the voltage where the device is in off state.