Ultra-fast photodetectors based on high-mobility indium gallium antimonide nanowires

Because of tunable bandgap and high carrier mobility, ternary III-V nanowires (NWs) have demonstrated enormous potential for advanced applications. However, the synthesis of large-scale and highly-crystalline InxGa1−xSb NWs is still a challenge. Here, we achieve high-density and crystalline stoichiometric InxGa1−xSb (0.09 < x < 0.28) NWs on amorphous substrates with the uniform phase-purity and <110 >-orientation via chemical vapor deposition. The as-prepared NWs show excellent electrical and optoelectronic characteristics, including the high hole mobility (i.e. 463 cm2 V−1 s−1 for In0.09Ga0.91Sb NWs) as well as broadband and ultrafast photoresponse over the visible and infrared optical communication region (1550 nm). Specifically, the In0.28Ga0.72Sb NW device yields efficient rise and decay times down to 38 and 53 μs, respectively, along with the responsivity of 6000 A W−1 and external quantum efficiency of 4.8 × 106 % towards 1550 nm regime. High-performance NW parallel-arrayed devices can also be fabricated to illustrate their large-scale device integrability for next-generation, ultrafast, high-responsivity and broadband photodetectors.

growth". This is not clear and a more detailed explanation should be provided. In any case, this does not show a good and reproducible control of the NW growth/composition. Also, the paper claims that "… as shown in the typical SEM image of In0.28Ga0.72Sb NWs (Figure 1a), straight, smooth and dense NWs with the length greater than 10 um are non-epitaxially grown on the amorphous SiO2/Si substrate. When the In concentration of the NWs is increased, the NW morphology is maintained with only the slight change in their growth density". The NWs are clearly not straight and also there is an obvious density variation. Especially for the In0.15Ga0.85Sb NWs, the density is obviously lower. The description in the paper should be made more accurate. In fact some morphology change with different In concentration is expected and should be discussed with regard to the growth conditions.
2) The chemical compositions of the NWs are obtained by EDS. The NW crystal structures are studied by TEM with very similar results. Surprisingly there is no mentioning of any defects such as stacking faults, for any of the NWs. I wonder how thorough the NWs were examined? The authors should clarify this. To get a better understanding of the NW properties and their corresponding device performance, optical spectroscopy such as photoluminescence should also be performed to measure the bandgap as well as the optical quality of the NWs. 3) In this work, a very high hole mobility of 463 cm2/Vs has been obtained. However for mobility calculation, the authors didn't show how to calculate the gate capacitance of the NW and what are the various parameters used. The Reference [5] cited didn't show the calculation either. In Table I the work by other researchers normally use 300 nm SiO2 and metallic cylinder model to calculate gate capacitance, the work only use 50nm SiO2 -it is important that it provides sufficient information to confirm the calculation and results. 4) It is shown that the 9% InGaSb NWs have the highest mobility and the results are presented in Fig. 4 (c) and (d). However Fig. 4(a) and (b) present the measurements of NW FET made of single In0.28Ga0.72Sb. Then the inset of Fig. 4(a) shows the SEM of In0.09Ga0.91Sb NW. It is confusing and the paper should just provide all the mobility measurement results from In0.09Ga0.91Sb NW FET to avoid confusion. 5) For the NW photodetectors, despite that impressive performance has been achieved, there is not much comparison (among different type of NWs), discussion and explanation. For example, compared with visible, a much faster response has been obtained for IR detection. Why? Compared with the In0.28Ga0.72Sb NWs, the In0.09Ga0.91Sb NW seems to have much lower noise and larger photocurrent (Fig. S11). Why the paper claims that In0.28Ga0.72Sb is the optimized device especially when the In0.09Ga0.91Sb NWs have higher mobility? How about the other two types of NWs with intermediate In compositions? How do they behave (in terms of responsivity, response time and detectitivity) and why? To better understand the photodetector performance of the NWs with different In concentration (and thus bandgap), the dark current and spectral response curves should also be provided to fully understand the device performance. 6) For the photodetectors, the photocurrents sometimes are shown in nA and sometimes are shown in uA. I suppose for the former the dark current has been subtracted? For the comparison, it is better to show all the results (except for the response time measurements) under the same illumination intensity with dark current subtracted. 7) P. 22, it is mentioned that "the internal quantum efficiency is known to be far smaller than 1 because of the reflectance of light, insufficient absorption of light attributable to the finite thickness of NW and anisotropic absorption of light arising from the one-dimensional geometry". This is not correct. Internal quantum efficiency (IQE) is defined as the ratio of the number of charge carriers that are generated to the number of photons that are absorbed. It is mainly related to the material quality rather than the reflectance of the light etc. The description above is more relevant to EQE. I also suggest the following minor corrections: 1) Fig. 1(b) is not an EDS image, it is an SEM image. 2) Please give the full name of "TEM" in Page 6, "NP" in Page 10. In Page 3, please add (MBE) after "molecular beam epitaxy" since MBE is used in the text afterwards.
3) In Fig. 3a, better to use the same colors and order for Au, In, Sb, Ga as Fig. 3b. The current colors and order is confusing. 4) Please provide a reference for Equation 1 and also the values for Vgs, gm, Cox, and L. 5) For Equation 3, the definition of area "S" should be given. 6) In the manuscript, when discussing the photodetection performance, the single NW device is often referred as single NW FET. This could be confusing and I suggest to call it as single NW photodetector. 7) Page 19, "In addition, the good repeatability and fast response speed are also two important requirement for advanced photodetectors", "requirement" should be "requirements".
Reviewer #3 (Remarks to the Author): In this work, the authors investigated the high-performance transistors and photodetectors based on ternary III-V InGaSb nanowires. Indeed, III-V nanowires are promising optoelectronic materials; however, the reliable synthesis of InGaSb nanowires with controlled composition, crystallinity and morphology is still a big challenge till now. Here, using the enhanced chemical vapor deposition, high-density and crystalline stoichiometric InGaSb nanowires can be readily obtained on amorphous substrates, in particular with the excellent hole mobility and ultrafast photoresponse over both visible and infrared optical communication regions when these nanowires are configured into devices. Therefore, it is recommended to consider all these novel findings, which would contribute significant advancement to the scientific community. In any case, there are some questions needed to be addressed. 1. The authors mentioned that the mobility distribution can be well fitted by Gauss function, however, no Gauss fitting was made in the manuscript. Table S1 is confusing, especially the percentage combustion being inconsistent with the values given in the table. For example, for the ratio 20:1, the mass of source is 1 g, and source loss is 0.26 g; in this case, the combustion percentage should be 26%, instead of 25.1% given in the table. The authors should make all the calculation more explicit. Furthermore, it is not suitable to use the term "combustion percentage" because the source is not combusted during the synthesis. 3. Similarly, for Table S2, why doesn't the molar ratio of In in the nanowires simply increase with the increasing of mass ratio of InSb powder? This reason would be important for the controllable synthesis of ternary III-V nanowires. 4. Regarding the nanowire array photodetectors, the reduced R and D* should not arise from the higher parasitic capacitance. The capacitance can only affect the response speed of a photodetector. While the speed of the nanowire array photodetector is almost the same with single nanowire device as presented by the authors, this indicates that there is not any higher parasitic capacitance existed for the nanowire array devices.

Response to Reviewers' Comments and Suggestions on Manuscript NCOMMS-18-11449
We appreciate the referees for considering our manuscript and providing valuable comments. Accordingly, changes have been made in the manuscript, highlighted with red color. Below is our response to the reviewers' comments.

Response to the Reviewers' comments:
Reviewer #1: 1. The manuscript contains a detailed characterization and discussion about the growth mechanism and the nucleation. Although required for the completeness of the manuscript, the topic is far from new. The conclusions are very similar to the discussions for instance related to the VLS growth of InAs/GaSb nanowire heterostructures developed over the last decade. Unfortunately this reviewer did not find any new aspect in the presentation. Response:  We thank for the comment. As pointed out by the reviewer, it is essential to evaluate and understand the detailed growth mechanism of ternary InGaSb NWs for the completeness of this work. In fact, the growth mechanism presented here is very different from the typical VLS growth of InAs/GaSb NWs (B. M. Borg, et al, Nano Lett., 2010, 10, 4080). In specific, for the growth of GaSb with 5% In (i.e. InGaSb) NW segment on the InAs NW stem, the formation of In-rich AuInGa alloy catalyst particle was found to inhibit the growth of InGaSb NW segments because of the thermodynamic reasons; therefore, it is challenging to achieve the uniform growth of InGaSb NW segments utilizing the ternary AuInGa based catalyst particles. On the other hand, in our experiments, we found that the binary AuIn catalyst particle can effectively catalyze the growth of ternary InGaSb NWs due to the employment of a higher growth temperature. This observed phenomenon is somewhat similar with the growth of InGaAs nanowhiskers (D. Sudfeld, et al, Phase Transitions, 2006, 79, 727), where the AuIn catalyst particle was also used there. As a result, all these new findings would further the advance of uniform ternary NW growth for practical device applications.
2. One main argument for the study is that InGaSb nanowires are expected to have a higher hole mobility than GaSb. Indeed high values are reported. However, the data in figure 4d is contradictory, the mobility is the highest for the lowest In content. Unfortunately, the GaSb reference is not clear as it is taken from a different study. Reading the manuscript it is not clear if really the introduction of In is beneficial for the transport? Response:  We appreciate for the valuable input. We have grown the GaSb NWs using the similar method as presented in this work and measured their corresponding field-effect mobility values as shown in Figure S16. It is clear that the average NW diameter increases slightly due to the uncontrolled radial growth and their corresponding hole mobility becomes smaller with an average value of only 26 cm 2 V -1 s -1 , being consistent with the previous literature (Yang et al., ACS Appl. Mater. Interfaces, 2013, 5, 10946). Also, the obtained result agrees very well with the theoretical study on InGaSb films (Nainani et al., J. Appl. Phys., 2012, 111, 103706), in which the hole mobility increases with the introduction of In. After the mobility reaches a peak value, the mobility would then decrease accordingly. As a result, it is confirmed that adding In is beneficial for enhancing the transport properties of GaSb NWs.  In order to reinforce this viewpoint, we have added the following discussion: "For comparison, GaSb NWs were also synthesized with the similar method here and their corresponding mobility values were measured with an average value of only 26 cm 2 V -1 s -1 (Supporting Information Figure S16). This variation trend of the mobility as a function of In content is also consistent with the previous literature that the hole mobility first increases with the introduction of In, and then decreases after it reaches the peak value. 12 " to page 15 of the revised manuscript. We also modified Figure 4d to include the mobility data of GaSb NW devices.
3. It is claimed that the contacts are ohmic. However, no systematic study is presented. What is the access resistance in these devices and how does it affect the mobility determination? Response:  We thank for the valuable suggestion. We have measured the contact resistance of the NW devices by fabricating single NW devices with the multiple channel length. In specific, the contact resistance of In0.09Ga0.91Sb NWs is found to be 11.3 k, which is 10 times smaller than the resistance of the NW body. Combined with the linear relationship as observed in their output characteristics, we can infer that the contacts are Ohm-like. This relatively small contact resistance means that there is only a small voltage drop on the contact. In this case, the mobility estimated from the transfer curves can indeed represent the actual mobility values of the NW devices.  In order to illustrate all these arguments, the following discussion: "In order to evaluate the contact quality of NW devices, their contact resistance were measured by fabricating single NW devices with the multiple channel length (Supporting Information Figure S13). It is obvious that the contact resistance of the In0.09Ga0.91Sb NW device is found to be 11.3 k, which is 10 times smaller than the resistance of the NW body. Combined with the linear relationship as observed in their output characteristics, the Ohm-like contact between the NWs and the electrodes can be inferred. This relatively small contact resistance means that there is only a small voltage drop on the electrical contact. In this case, the mobility estimated from the transfer characteristics can indeed represent the actual mobility values of the NW devices." is added to page 13 of the revised manuscript.

The transistors show a hysteresis. How is this taken into account when determining the mobility?
Response:  We thank for the comment. Usually, for NW transistors, the hysteresis is originated from the surface charge trapping, which can reduce the mobility of NW devices. Also, the mobility values obtained from the positive-to-negative gate sweep is always smaller than that from the reverse sweep due to the surface nature of NWs. In this work, in order to get the appropriate estimation of the mobility values, we employed the positive-to-negative gate sweep for all the parameter calculation. For example, the peak mobility of the typical In0.09Ga0.91Sb NW FET is found to be 463 and 491 cm 2 V -1 s -1 , for the positive-to-negative and the negative-to-positive gate sweeps, respectively ( Figure R1). Although there is hysteresis observed, it is rather small, and we used the lower bound value of 463 cm 2 /Vs as its mobility in our analysis. In order to reinforce this argument, we have added Figure S15 to the revised Supporting Information.

It is claimed that the transistors have an Ion/Ioff ratio of 10^5 at Vds=0.4V. However no supporting data for this very bold claim is provided!
Response:  We thank for the input. We have added the corresponding log plot of the transfer curve of the NW device as shown in Figure 4a inset in page 12 of the revised manuscript. It is clear that the Ion/Ioff ratio of ~10 5 is observed at Vds = 0.4 V.

The optical data is interesting, but the response times are very long and the applied biases high.
The authors need to provide a much more detailed discussion on the possible surface effects and provided data for passivated nanowires, to make sure that this is not a material quality problem. Response:  We appreciate for this valuable comment. Actually, the applied biases are consistent with the ones widely utilized in the previous works, while the response times are very short as compared with the values reported in the literature (Table 1). In order to evaluate the surface effect, we also measured the photoresponse of passivated NW devices. In specific, the effect of surface passivation on the In0.28Ga0.72Sb NW photodetector with the channel passivated with (NH4)2S was thoroughly evaluated (Supporting Information Figure S21). It is found that the rise and decay time constants of the passivated InGaSb NW device were reduced from 38 and 53 s to 24 and 37 s, respectively, after surface passivation. These reduced response times can be explained by the minimized surface trap concentration owing to the effective surface passivation. Since the difference of the observed response times is relatively small as compared between with and without the surface passivation, all the discussion presented in this work would be based on the results without any surface passivation. In any case, the optical response can be confirmed due to the intrinsic properties of our InGaSb NWs, instead of the material quality issue.  In order to reinforce this viewpoint, the following discussion: "In the meanwhile, further reducing the surface trap concentration by surface passivation with (NH4)2S can decrease the response time down to 24 s as shown in Supporting Information Figures S21. Although there are abundant surface states in the nanostructured materials, in which these states should be first saturated by the photo-induced carriers making the response times slower, these small changes of the response time (i.e. with and without any surface passivation) can infer the relatively high NW crystal quality here." is added to page 21 of the revised manuscript.

Reviewer #2:
1. This work reports composition controllable InGaSb NWs achieved by mixing different ratio of binary material powder sources. However the results are not consistent. For powder source ratio of 30:1 (the third highest InSb case), the lowest In concentration (9%) was obtained. The supporting information mentions that it "is probably due to the lower substrate employed for the optimized growth". This is not clear and a more detailed explanation should be provided. In any case, this does not show a good and reproducible control of the NW growth/composition. Also, the paper claims that "… as shown in the typical SEM image of In0.28Ga0.72Sb NWs (Figure 1a), straight, smooth and dense NWs with the length greater than 10 um are non-epitaxially grown on the amorphous SiO2/Si substrate. When the In concentration of the NWs is increased, the NW morphology is maintained with only the slight change in their growth density". The NWs are clearly not straight and also there is an obvious density variation. Especially for the In0.15Ga0.85Sb NWs, the density is obviously lower. The description in the paper should be made more accurate. In fact some morphology change with different In concentration is expected and should be discussed with regard to the growth conditions. Response:  We thank for these valuable comments.  Regarding the result inconsistency, we have repeated the experiments for many times and obtained the similar, consistent results. In particular, for the optimal growth condition of InGaSb NWs with around 9 % In content, the growth temperature is optimized at 505 ℃, which is the lowest temperature required among all NWs with different In content. We have also tried the higher growth temperature (>505 ℃) with the same powder mixing ratio, but this combination of growth parameters is out of the optimized process window grown with defective NWs with lots of surface coating. Since the growth temperature can drastically affect the chemical ratio of between Au and In in the AuIn alloy catalyst particle, which can subsequently affect the chemical composition of the grown NWs, we believe that the different growth temperature employed in this particular condition would contribute to the inconsistency here, where the powder source ratio of 30:1 (the third highest InSb case) yields the lowest In concentration (9%).  In order to reinforce this standpoint, the following discussion: "Note: All associated values are extracted from more than 10 individual NWs for each sample group. It is noted that the powder source ratio of 30:1 would result in a relatively lower In concentration in InxGa1-xSb NWs. This inconsistency should not be an accidental result as we have repeated the experiments for many times and obtained the similar results. For the optimal growth condition of InGaSb NWs with around 9 % In content, the growth temperature is optimized at 505 ℃, which is the lowest temperature required among all NWs with different In content. We have also tried the higher growth temperature (>505 ℃) with the same powder mixing ratio, but this combination of growth parameters is out of the optimized process window grown with defective NWs with lots of surface coating ( Figure S2d). Since the growth temperature can drastically affect the chemical ratio of between Au and In in the AuIn alloy catalyst particle, which can subsequently affect the chemical composition of the grown NWs, we believe that the different growth temperature employed in this particular condition would contribute to the inconsistency here, where the powder source ratio of 30:1 (the third highest InSb case) yields the lowest In concentration (9%)." is added to page 2 of the revised Supporting Information.  Regarding the changes on the NW density, we have revised the description: "Furthermore, as shown in the SEM image of typical In0.28Ga0.72Sb NWs (Figure 1a), smooth and clean NWs with the length greater than 10 m are grown on the amorphous SiO2/Si substrate. When the In concentration of the NWs is increased, the NW morphology is maintained, but there is a slight change on their growth density due to the increasing amount of InSb powder in the precursor source mixture. (Supporting Information Figure S2)." in order to clearly illustrate our argument in page 6 of the revised manuscript.  (Figure 2, Figure S4, S5 and S6). We observed that there is not any noticeable 1D and 2D crystal defects such as stacking faults, twin boundaries and others, indicating the high crystal quality of the obtained NWs.  For the additional characterization, we have measured the reflectance spectra of the NWs with different compositions, in which these data are added to Figure S11 and S12 in the revised Supporting Information. Also, more discussion: "Furthermore, reflectance spectra were taken to evaluate the optical properties of the NWs (Supporting Information Figure S11). As anticipated, the band gap of the NWs is found to decrease with the increasing In content (Supporting Information Figure S12), indicating the effective incorporation of In into the GaSb lattice as well as the good composition uniformity of the NWs." is added to page 9 of the revised manuscript.  Figure S14. In order to clearly illustrate this argument, the following discussion: "Cox is the gate capacitance and L is the channel length. Cox can be obtained from the finite element method by using COMSOL MultiPhysics software (Supporting Information Figure S14). For a typical In0.09Ga0.91Sb NW transistor, when L is 4.2 m, NW diameter is 41 nm, gate capacitance is determined to be 0.26 fF for 50 nm SiO2 dielectric layer from COMSOL and peak transconductance is found to be 7 × 10 -8 S for Vds = 0.1 V (Supporting Information Figure S15), then the peak  of the NW device can be calculated as high as 463 cm 2 /Vs (Figure 4c). This mobility value is higher than the one of pure GaSb NWs……" is added to page 14 of the revised manuscript. Fig. 4 (c) and (d). However Fig. 4(a) and (b) present the measurements of NW FET made of single In0.28Ga0.72Sb. Then the inset of Fig. 4(a) shows the SEM of In0.09Ga0.91Sb NW. It is confusing and the paper should just provide all the mobility measurement results from In0.09Ga0.91Sb NW FET to avoid confusion. Response:  We thank for pointing out our mistakes. All the results given in Figure 4a, b and c were for the In0.09Ga0.91Sb NW device. We are sorry for the mistakes and we have corrected them in the revised manuscript.

For the NW photodetectors, despite that impressive performance has been achieved, there is not much comparison (among different type of NWs), discussion and explanation. For example, compared with visible, a much faster response has been obtained for IR detection. Why?
Compared with the In0.28Ga0.72Sb NWs, the In0.09Ga0.91Sb NW seems to have much lower noise and larger photocurrent (Fig. S11) concentration (and thus bandgap), the dark current and spectral response curves should also be provided to fully understand the device performance. Response:  We appreciate for these valuable comments.  Since the response speed of a photodetector is mostly related to the carrier lifetime and carrier trap concentration, that is τ0 = (1 + pt/p)τ, where τ is the lifetime of photogenerated carriers, pt is the trapped carrier density and p is the photo-generated free carrier density (A. Rose, Concepts in Photoconductivity and Allied Problems, Roberte Krieger Publishing Co., New York, 1978), the response speed should have an insignificant dependence on the detection wavelength. In this case, we have repeated the photoresponse measurement of the photodetector with a new 405 nm laser and a 532 nm laser used previously, which can now be modulated in a higher frequency (1 kHz) by directly tuning the power source of the laser, instead of using a mechanical shutter. The measurement result is then given in the revised Figure 5d. It is found that the rise and decay time constants are 39 and 46 μs, respectively for the 532 illumination (instead of in the range of milliseconds), which are consistent with the response observed in the IR region as well as being consistent with the theory discussed above.  It should be pointed out that in our previous measurements in the visible range, the modulation frequency cannot be modulated in high frequency so that low frequency (e.g. 20 Hz) was used. However, the low-frequency modulation is not the reason for obtaining the inaccurate response time if the mechanical shutter is fast enough. After a careful consideration, we have identified the problem, which is due to the inappropriate use of the low noise current amplifier. For example, if we employed a low-frequency low pass filter (30 Hz), the response curve can be affected drastically. However, if a higher frequency filer is used, the curve would become very noise. In order to solve the problem, a higher modulation frequency should be utilized if the device response is fast enough. In this manner, we resolved the noise issue and obtained the accurate photoresponse of all devices here. More discussion of "This way, the rise and decay time constants are impressively found to be 39 and 46 μs, respectively, indicating the fast response of the device." is then added to page 17 of the revised manuscript in order to clarify the confusion.  At the same time, the repeated measurement results of the InGaSb NW photodetector under 405 nm illumination with different Indium concentration are added to revised Supporting Information Figure S17, S19 and Table S3). Importantly, we compared the photocurrent ( Figure S17 and S19), responsivity, detectivity and response time of the devices (Table S3), in which there is not any noticeable variation among all parameters as the In concentration changes, indicating the potency of these NWs for broadband photodetection in the visible regime. More discussion of "Photodetection performance of the NWs with other In contents were also measured and the results are shown in Supporting Information Figure S19 and Table  S3. It is clear that there is not any significant difference in their performances among different NW compositions. Typically, the photocurrent is proportional to the product of carrier mobility and carrier lifetime ( p I   ); 43 however, the density of different types of recombination centers can drastically change the value of carrier lifetimes here. 43 As a result, it is anticipated that those NWs would exhibit the similar photodetection performance although their carrier mobilities are different." is added to page 17 and 18 of the revised manuscript.  In additional, we have also carried out more thorough photoresponse measurements of the InGaSb NW devices under 1550 illumination and compiled all the results presented in Figure  S20 and Table S4 of the revised Supporting Information. For example, we have added all the dark current characteristics in Figure S20 inset and there is not any noticeable difference on the dark current among different NW compositions. Similarly, as anticipated, NW devices with different In content exhibit the similar response time (in the order of tens of s), which agrees well with the theory discussed above. On the other hand, it is interesting that the NWs with the highest In content (i.e. In0.28Ga0.72Sb with the smallest bandgap here) yield the highest photoresponsivity, external quantum efficiency (EQE) and detectivity among all NWs studied. In contrast to the hot carriers effect observed for the photodetection in the visible region, the In0.28Ga0.72Sb NWs probably have the largest absorption coefficient towards 1550 nm irradiation although it has the relatively low mobility as compared to other InxGa1-xSb NWs. Combined with the highest NW density obtained on the growth substrates (Figure 1 and Figure  S2), the In0.28Ga0.72Sb NWs are concluded to be the optimized device channel material for the IR photodetection in this work.  In order to clearly illustrate this argument, the following discussion of "the IR response of those NW devices would enhance substantially when the In concentration of the NW device channel is increased accordingly, which can probably be attributed to the reduced bandgap as well as the enhanced absorption coefficient towards 1550 nm irradiation (Supporting Information Figure S11, S20 and Table S4)" is added to page 19 of the revised manuscript.

For the photodetectors, the photocurrents sometimes are shown in nA and sometimes are shown in A. I suppose for the former the dark current has been subtracted? For the comparison, it is
better to show all the results (except for the response time measurements) under the same illumination intensity with dark current subtracted. Response:  We thank you for the good suggestion. We have subtracted the dark current in the revised manuscript as shown in the revised Figures 5 and 6, and all units of photocurrents are now changed to nA. 7. P. 22, it is mentioned that "the internal quantum efficiency is known to be far smaller than 1 because of the reflectance of light, insufficient absorption of light attributable to the finite thickness of NW and anisotropic absorption of light arising from the one-dimensional geometry". This is not correct. Internal quantum efficiency (IQE) is defined as the ratio of the number of charge carriers that are generated to the number of photons that are absorbed. It is mainly related to the material quality rather than the reflectance of the light etc. The description above is more relevant to EQE. Response:  We appreciate for the comments and we are sorry for the inappropriate use of term of "internal quantum efficiency". Here, it is suitable to refer it as "quantum efficiency". The definition of quantum efficiency is the fraction of the incident optical power that contributes to electronhole pair generation, which is widely used in photoconductive detectors (B.  York, chapter 17, P649). The definition is different from internal quantum efficiency (IQE) and external quantum efficiency (EQE). The ICE considers the fraction of light absorbed by the detector, while here it considers all the light incident on the detector. EQE is the ratio of the number of charge carriers collected by electrodes to the number of photons incident on the detector. In fact, the EQE is the product of quantum efficiency and gain. In order to clearly clarify this argument, we have changed the term of "internal quantum efficiency" to "quantum efficiency", and the definition of the quantum efficiency is also given in page 23 of the revised manuscript.
8. I also suggest the following minor corrections: 1) Fig. 1(b) is not an EDS image, it is an SEM image. 2) Please give the full name of "TEM" in Page 6, "NP" in Page 10. In Page 3, please add (MBE) after "molecular beam epitaxy" since MBE is used in the text afterwards. 3) In Fig. 3a, better to use the same colors and order for Au, In, Sb, Ga as Fig. 3b. The current colors and order is confusing. 4) Please provide a reference for Equation 1 and also the values for Vgs, gm, Cox, and L. 5) For Equation 3, the definition of area "S" should be given. 6) In the manuscript, when discussing the photodetection performance, the single NW device is often referred as single NW FET. This could be confusing and I suggest to call it as single NW photodetector. 7) Page 19, "In addition, the good repeatability and fast response speed are also two important requirement for advanced photodetectors", "requirement" should be "requirements". Response:  We thank you for pointing out all those mistakes. We have corrected all of them throughout the entire revised manuscript.
Reviewer #3: 1. The authors mentioned that the mobility distribution can be well fitted by Gauss function, however, no Gauss fitting was made in the manuscript. Response:  We thank for the comment. Gauss fittings were made and the fitting lines are now shown in Figure 4d of the revised manuscript. Figure R2. Statistics of the extracted mobility of several NW FETs with different NW stoichiometry as the device channel. Table S1 is confusing, especially the percentage combustion being inconsistent with the values given in the table. For example, for the ratio 20:1, the mass of source is 1 g, and source loss is 0.26 g; in this case, the combustion percentage should be 26%, instead of 25.1% given in the table. The authors should make all the calculation more explicit. Furthermore, it is not suitable to use the term "combustion percentage" because the source is not combusted during the synthesis. Response:  We thank for pointing out the mistakes. For clarity, as the term of "combustion percentage" is not an important parameter in this study, we decide to delete it in Table S1 of the revised supporting information.

Similarly, for Table S2, why doesn't the molar ratio of In in the nanowires simply increase with the increasing of mass ratio of InSb powder? This reason would be important for the controllable synthesis of ternary III-V nanowires.
Response:  We appreciate for the valuable input. In fact, as responded to reviewer#2 (question 1), we have repeated the experiments for many times and obtained the similar, consistent results. In particular, for the optimal growth condition of InGaSb NWs with around 9 % In content, the growth temperature is optimized at 505 ℃, which is the lowest temperature required among all NWs with different In content. We have also tried the higher growth temperature (>505 ℃) with the same powder mixing ratio, but this combination of growth parameters is out of the optimized process window grown with defective NWs with lots of surface coating. Since the growth temperature can drastically affect the chemical ratio of between Au and In in the AuIn alloy catalyst particle, which can subsequently affect the chemical composition of the grown NWs, we believe that the different growth temperature employed in this particular condition would contribute to the inconsistency here, where the powder source ratio of 30:1 (the third highest InSb case) yields the lowest In concentration (9%).  In order to reinforce this standpoint, the following discussion: "Note: All associated values are extracted from more than 10 individual NWs for each sample group. It is noted that the powder source ratio of 30:1 would result in a relatively lower In concentration in InxGa1-xSb NWs. This inconsistency should not be an accidental result as we have repeated the experiments for many times and obtained the similar results. For the optimal growth condition of InGaSb NWs with around 9 % In content, the growth temperature is optimized at 505 ℃, which is the lowest temperature required among all NWs with different In content. We have also tried the higher growth temperature (>505 ℃) with the same powder mixing ratio, but this combination of growth parameters is out of the optimized process window grown with defective NWs with lots of surface coating ( Figure S2d). Since the growth temperature can drastically affect the chemical ratio of between Au and In in the AuIn alloy catalyst particle, which can subsequently affect the chemical composition of the grown NWs, we believe that the different growth temperature employed in this particular condition would contribute to the inconsistency here, where the powder source ratio of 30:1 (the third highest InSb case) yields the lowest In concentration (9%)." is added to page 2 of the revised Supporting Information.

4.
Regarding the nanowire array photodetectors, the reduced R and D* should not arise from the higher parasitic capacitance. The capacitance can only affect the response speed of a photodetector. While the speed of the nanowire array photodetector is almost the same with single nanowire device as presented by the authors, this indicates that there is not any higher parasitic capacitance existed for the nanowire array devices. Response:  We thank for pointing out the mistakes. The quality of the printed NW arrays may contribute

Response to Referees' Comments and Suggestions on manuscript NCOMMS-18-11449A
We appreciate the referees for considering our manuscript and providing valuable comments. Accordingly, changes have been made in the manuscript, highlighted with red color. Below is our response to reviewers' comments.

Reviewer #1
The  Figure S16e and f, it is obvious that there are many lattice defects (e.g. stacking faults and inversion domains, etc) existed in the GaSb nanowire, which is in a distinct contrast to the ones of InxGa1-xSb nanowires as presented in Figure 2 and Supplementary Information Figure S4 to S6. These defects could result from the notorious surfactant effect of Sb constituents during the nanowire growth (Nat. Commun., 2014, 5, 5249). These large amounts of defects could contribute to the relatively low mobility of GaSb nanowires due to the severe carrier scattering there. Furthermore, these defects can also provide a large amount of free carriers that cause the down shift of the Fermi level of GaSb to its valance band maximum. As a result, it is difficult to deplete the free carriers of the nanowire to achieve the device OFF state by electrical back-gating. It is also a common phenomenon observed for back-gated GaSb nanowire transistors that cannot be turned off at room temperature (Appl. Phys. Lett., 2011, 99, 262104; RSC Adv., 2013, 3, 19834). On the other hand, the minimized lattice defect concentration of InxGa1-xSb nanowires could contribute to the correspondingly enhanced device carrier mobility. Moreover, even utilizing sulfur surfactants during the growth to obtain highly crystalline GaSb nanowires, the maximum device mobility value of 400 cm 2 V -1 s -1 (ACS Nano, 2015, 9, 9268) is still lower than the one of In0.09Ga0.91Sb nanowire devices with the comparable nanowire diameter, which suggests the introduction of In in GaSb nanowires not only reduces the defect concentration but also leads to the favorable changes in their electrical device performance. As a result, the device mobility values of InxGa1-xSb nanowires are much improved as compared to the pure GaSb nanowire ones with the nanowire channel grown by the similar technique.
-Furthermore, the mobility values of nanowire devices can also be affected by the channel surface condition, such as the surface roughness, passivation and adsorbents, etc. In this work, since both GaSb and InGaSb nanowires have the smooth surface as revealed from the TEM characterization in Figure 2, Supplementary Information Figure S4 to S6 and S16, the surface roughness effect should be the same among all these nanowires. To further exclude the effect of any surface modification (e.g. adsorbents) on the device mobility of the nanowires, we measured the transfer curves of a typical In0.09Ga0.91Sb nanowire transistor in both air and vacuum (3.5×10 -4 Pa) as shown in Supplementary Information Figure S17.
Explicitly, there is not any noticeable difference observed for the measurement result in both air and vacuum, which suggests that the adsorbents have insignificant effect on the electrical properties of nanowire devices. Then, surface passivation with ammonia sulfide (NH4)2S were also performed on the In0.09Ga0.91Sb nanowire transistor. As presented in Supplementary Information Figure S18, the surface passivation can slightly reduce the subthreshold swing and move the subthreshold voltage towards the negative voltage direction by minimizing the surface trap concentration on the nanowire device channel, which can further enhance the corresponding device mobility to some extent. This phenomenon indicates that the initial high device mobility values of our In0.09Ga0.91Sb nanowires are indeed attributed to their intrinsic material properties, instead of relating to any surface passivation effect. In this case, the effective electrical back-gating to efficiently turning the device ON and OFF can also be attributed to the intrinsic nanowire properties, rather than any other extrinsic effect (e.g. surface modification).
-In this regard, in order to reinforce all the above argument that the enhanced device mobility of In0.09Ga0.91Sb nanowires is mainly attributed to both the effective incorporation of In into the nanowire lattice and the minimized lattice defect concentration, instead of relating to any surface modification effect, we have included all these discussion in page 13, 14 and 31 of the revised manuscript as well as added Figure S15c, S16e, S16f, S17 and S18 in the revised Supplementary Information.