In situ atomic-scale observation of oxidation and decomposition processes in nanocrystalline alloys

Oxygen contamination is a problem which inevitably occurs during severe plastic deformation of metallic powders by exposure to air. Although this contamination can change the morphology and properties of the consolidated materials, there is a lack of detailed information about the behavior of oxygen in nanocrystalline alloys. In this study, aberration-corrected high-resolution transmission electron microscopy and associated techniques are used to investigate the behavior of oxygen during in situ heating of highly strained Cu–Fe alloys. Contrary to expectations, oxide formation occurs prior to the decomposition of the metastable Cu–Fe solid solution. This oxide formation commences at relatively low temperatures, generating nanosized clusters of firstly CuO and later Fe2O3. The orientation relationship between these clusters and the matrix differs from that observed in conventional steels. These findings provide a direct observation of oxide formation in single-phase Cu–Fe composites and offer a pathway for the design of nanocrystalline materials strengthened by oxide dispersions.

The authors claim that oxides are formed in the interior of the grains. However, the TEM images projections of the TEM samples. The presented data do not exclude the possibility of the oxides forming at the TEM specimen surface. Further information is provided before the author can claim that the oxides are formed in the interior of the grains. This is crucial for the statements in the discussion and abstract.
The morphology of the oxides are not shown in the ex situ annealed samples. Why?
What was the partial oxygen pressure in the TEM during observation?
Did the authors determine the oxygen content of the initial raw powders? Did it correlate to the oxygen content in the HPT materials?  There are small areas with bright contrast at the grain boundaries in Figure 5 a in the supplementary material as well as the areas in the white circles. What is the difference between the small bright areas?
Reviewer #2: Remarks to the Author: In this manuscript oxidation in nanostructured FeC u alloys is studied using a wide arrays of experimental techniques and some electronic structure calculations are presented as support as well. It is an interesting manuscript for sure, but I do not deem this of high enough novelty nor of sufficient interest for the readership of Nature C ommunications. I therefore recommend rejection of the manuscript but I fully endorse the authors submitting this work to a more specialized journal.
For this reason, I provide some constructive feedback to the manuscript that the authors could take into account for future submissions:

Reply to Reviewer #1:
We really appreciate the reviewer for positive comments and helpful suggestions.
To address these comments and concerns, we have done comparative experiments to investigate the effects of altered layers on the TEM sample surfaces, and implemented systematical XPS studies regarding the oxygen contents in the initial and deformed materials.
HRTEM image and EELS elemental mapping of ex-situ annealed sample were supplemented. All concerns will be addressed in detail as follows.
1. There is no information about how the TEM specimens were prepared. Was there an altered surface layer of the TEM specimens due to ion beam thinning? How thick was the damaged layer? How thick were the TEM specimens in the areas that were studied? Did the TEM specimen thickness correspond to the grain size of the materials?
Our TEM samples were prepared using mechanical thinning method combined with ionmilling. The detailed procedure is described as follows: The TEM samples were cut from the HPT disks, and all microstructural investigations were undertaken at radius of 3.0 mm from the torsional axis of the HPT deformed disks as shown in the following schematic diagram of Fig. R1 (to avoid the possible confusion, all figures shown in this reply file are indicated with additions of letter "R", for example, Fig. R1). The TEM disk was then mechanically thinned and polished to a thickness of about 30 -40 µm, followed by mechanical dimpling with a remaining thickness of about 15 µm in the center of the dimple. Subsequently the samples were ion-milled using a Gatan Precision Ion Polishing System 691 (Gatan, Inc., P leasanton, USA) at -170 °C via liquid nitrogen cooling until perforation with voltage of 4 kV and angle of 4˚, and then gentle polishing with lower limit conditions of 1.5 kV and 1˚ -1.5˚ was implemented on the sample hole to 3 remove artefacts on the sample surfaces and make the thin area as flat as possible nearly without thickness gradient. The schematic diagram of ion-milling process is shown in Fig. R2.
Actually due to the sputtering of accelerated Ar ions on the sample surfaces during ion-milling, it seems to be inevitable to avoid the surface alteration. In addition, for TEM sample, the socalled surface relaxation phenomenon is existed at the thin area. Unfortunately, it is really hard to exactly know how thick of the surface altered layer, because it is related to the materials, internal strains and ion-milling parameters. However, as mentioned above, we have taken a measure by controlling the ion-milling parameters, with the lowest voltage of 1.5 kV and angle of 1˚ -1.5˚ as long as a weak ion current can be detected in the equipment, to gently polish the sample surfaces at the final ion-milling stage. By this way, the effect from surface altered layer due to ion-milling can be minimized. For all TEM samples we prepared in this way, we didn't observe any altered surface layer in HRTEM images even at the extremely thin edges. If the influence from such layer is pronounced, the changed areas should be detected by HRTEM as damaged regions.
Actually we have done comparative experiments using the same material as reported in the manuscript, HPT-deformed 75Cu-25Fe nanocrystalline alloy. We ion-milled the sample with parameters of voltage of 6 kV and angle of 8˚ for about 20 min until perforation, and then checked the sample in TEM. The BF and HRTEM images are shown in Fig. R3. It can be seen that the blurred spots with dimension of about 5 nm are damaged areas by ion hitting. Some spots are even changed to amorphous structures. However, a trial using ion-milling parameters of voltage of 5 kV and angle of 6˚ with liquid nitrogen cooling on as-deformed nanocrytalline pure Furthermore, the highest ion-milling parameters we used in this work are 4 kV and 4˚, which should make the sample surfaces more stable and consistent with the internal structures.
The TEM sample thickness close to the edge is about 15 -25 nm, and the average grain size of our investigated sample 75Cu-25Fe is about 56 nm which has already been mentioned in the manuscript. So the TEM sample thickness is much smaller than the average grain size, and it must be ensured that no grain overlaps exists in the area for in-situ HRTEM observation, otherwise Moiré fringes will be observed. But our HRTEM image recorded at 20 °C shown in the manuscript Fig. 2 is quite clear and doesn't display any Moiré fringe.  2. The authors claim that oxides are formed in the interior of the grains. However, the TEM images are projections of the TEM samples. The presented data do not exclude the possibility of the oxides forming at the TEM specimen surface. Further information is provided before the author can claim that the oxides are formed in the interior of the grains. This is crucial for the statements in the discussion and abstract. 7 We appreciate that this concern was raised by the reviewer. Actually from the HRTEM images shown in Fig. 2 in the manuscript, it can be determined that the oxides are formed inside the fcc matrix, rather than only on the surfaces. Probably the HRTEM images shown in the first version were too small to be distinguished clearly. When the sample is heated to 180 °C, some areas get blurred, and indeed Moiré fringes can be seen. The mappings shown in Fig. 2 in the manuscript confirm that these blurred areas are CuO and Fe 2 O 3 . The occurrence of Moiré fringes means the overlaps of oxides and the fcc matrix, which is due to that the oxides are nucleating and developing at the low temperatures, and their sizes are quite small compared to the sample thickness of the observed area. As the temperature increased to 260 °C, it can be seen that the blurred areas expand and their structures get more  Fig. R11). From this point of view, we can judge the oxides are inside the materials. We also checked the ex-situ annealed sample by HRTEM as shown in Fig.   R6, and for some area no Moiré fringe is observed, and morphology looks the same as already reported precipitates existed inside grains [1][2][3][4] . According to the morphologies of oxides shown in HRTEM images of in-situ and ex-situ annealed samples, we can draw the conclusion that oxides are formed in the interior of grains. Synchrotron X-ray diffraction results shown in Fig. R12 will confirm this point.
An additional evidence to prove the oxide particles are growing within the sample, instead on the surface of the sample is to carry out the line-scan analysis crossing the oxide particle using EDXS/EELS (as shown in the following Fig.R8). The core intensity/signal at the oxide particles will give difference when the oxide particles form on the surface or grow within the sample.   The HRTEM image of oxide of ex-situ annealed sample can be referred to Fig. R6. Meanwhile the EELS mappings of ex-situ annealed sample will be shown below as requested in seventh question, where the morphologies of oxides and Fe precipitates can be clearly seen. 4. What was the partial oxygen pressure in the TEM during observation?
The column vacuum in our TEM machine JEOL 2100F is a high vacuum system with vacuum of about 5 × 10 -7 Pa. If we regard the oxygen volume fraction is about 1/5 just like air, the partial oxygen pressure can be calculated to be 10 -7 Pa. 5. Did the authors determine the oxygen content of the initial raw powders? Did it correlate to the oxygen content in the HPT materials?
We thank the reviewer for this suggestion. We have re-measured the oxygen contents in the initial raw powders, the as-deformed HPT sample and a piece of commercial pure Cu material (nominal purity of 99.99%, used as a reference) using X-ray photoelectron spectroscopy (XPS, ESCALAB 250Xi, Thermo Fisher Scientific, Waltham, USA). We have taken a series of special measures to remove the possible surface oxide layers before transfer the samples to the XPS chamber. All sample surfaces were fully polished in a media of ethyl alcohol and then transferred to XPS chamber immediately, which was followed by Ar ion sputtering with ion energy of 3 keV for 5 minutes to completely remove the possible surface oxide layers. We would emphasize that all samples were kept in ethyl alcohol after polishing, and the operation time of transfer from ethyl alcohol to XPS chamber was controlled to a minimum of about a few seconds. All three samples mentioned above were transferred into the chamber at the same time.  and 10.7 (Fe). It can be seen the powder samples contain a level of about 3 at.% oxygen inside material while for commercial pure Cu the oxygen content is less than 1 at.%. It should be emphasized that for the fine scan spectrum of commercial pure Cu, the signal-to-noise ratio of O 1s peak is quite poor and it is hard to calculate the integrated area of this peak accurately, so the given value here of 1 at.% is guaranteed to be overestimated.
The measured oxygen content in 75Cu-25Fe as-deformed disk here is 3.6 at.%. Actually we have carried out another two independent measurements on two separate as-deformed samples, the oxygen contents in these two samples are 3.4 at.% and 3.3 at.% respectively. So it can be assured that the oxygen content in the HPT as-deformed samples is about 3.5 at.%. As for the reason that the measured oxygen value in the compacted powders is 3.0 at.%, a little bit lower than the value in samples after HPT deformation, it may be attributed to the pores formed during compaction which is segmented and the oxygen has reacted with metal elements during continuous deformation.

Table R1
Constituents of 75Cu-25Fe as-deformed sample, commercial pure Cu rod and 75Cu-25Fe compacted raw powders calculated based on XPS spectra. 6. Figure 3 in the supplemental material illustrate the presence of Fe grains at 420 °C. It is not clear how the authors reach the conclusion that the two grains are pure iron grans.
Please clarify.  Figure 5 in the supplemental material shows concentration profiles after thermal anneal.
The presented data do not verify that the formation of iron occurs as a result of the thermal anneal since there are no data provided before thermal anneal. A figure corresponding to Figure 3 c, in the main manuscript, with data from EELS mapping at 20 °C for both in situ and ex situ would provide crucial information for this manuscript.
As we mentioned above in the sixth question, EDXS line scan was implemented on the annealed sample across the grains to determine the newly-generated Fe grains. Actually, from the morphologies of the annealed sample, the Fe grains grown due to the decomposition usually have rounded corner shapes rather than the irregular multi-angular grains of the matrix, just as shown in the Fig. 3 and Fig. 5 in the supplemental material of the first version. Based on this experience, we first select such newly-grown grains with rounded corners and then confirm the Fe grains by EDXS measurements. The data of Fig. 3 shown in the supplemental material of the first version has displayed the occurrence of Fe grains due to the thermal annealing.
We thank the reviewer for suggesting us to carry out the EELS mapping for the ex-situ annealed sample for confirmation of the EELS mapping results of in-situ heating sample. Fig. R8 shows the EELS elemental mapping of ex-situ annealed sample at 420 °C. As we mentioned above, the Fe grains grown due to the decomposition usually have rounded corner shapes which can be also revealed from the Fe_L mapping in

Reply to Reviewer #2:
We appreciate the reviewer's constructive comments very much.
One of the reviewer's main concerns is about the novelty of this work. To address this point and satisfy the referee, the main statements regarding the novelty are given as follows: Nowadays, production of applicable bulk nanocrystalline alloys using severe plastic deformation is of great importance for the next-generation high-performance structural materials, and it is drawing growing interests in the materials research field and industry applications. Unfortunately one prominent problem arising during the consolidation and straining processes is the unavoidable contamination from gaseous species for powders-processing technology, such as oxygen. Currently, people have realized that oxygen contamination can induce large discrepancies in the mechanical properties, microstructures and thermal stabilities. However, the exact state of oxygen contaminant inside the materials remains unclear, and how oxygen atoms behave during annealing has hardly been explored. Moreover, developing new nanostructured bulk materials for potential applications demands atomic-resolution insights into the structures, particularly, the understanding of oxygen behaviors in nanostructured materials. In this work, we employed the modern spherical aberration corrected high-resolution transmission electron microscopy to observe the oxygen behavior via in-situ annealing at the atomic scale. The results revealed that except decomposition process, the nanometer-sized oxide clusters could form inside the grains at specific temperatures. i) This is a first atomic-scale observation of the oxidation and decomposition process in bulk nanocrystalline alloys; ii) The finding in this study is of great significance in developing a new route for designing new bulk nanocrystalline alloys by intentionally introducing different oxygen content, which can produce dispersive nano-sized oxides in the matrix so that the mechanical properties can be tuned as desired; iii) Actually, in general, this study will assist understanding the oxygen atom behavior in other material systems, such as in nanostructured materials, oxides and steels, and clarifying its role in the field of surface science, and catalysts.
As for the other technical concerns, we have implemented systematic XPS studies regarding the oxygen impurity in the initial materials (powder material and arc-melted bulk materials), and recalculated the formation enthalpies using DFT + U model with consideration of the strong correlation effects between transition metal elements and oxygen. In addition, we quantified the electron density under the HRTEM imaging condition. The concerns will be addressed in detail below.
1. The authors have missed out on a large part of the rather rich literature of oxidation and oxidation kinetics in oxide dispersion strengthened (ODS) steels. These are mentioned very briefly in the introduction but no reference is made to significant studies of kinetics, phase formation, effect of ion beams (or electron beams) on the evolution of these nanostructured alloys.
We thank the reviewer for helpful suggestion. Some typical literature related to ODS materials are included in the revised version of the manuscript, and some descriptions related to ODS alloys are modified. The reason that we mentioned the ODS steels in the introduction part is that we tend to extend the idea on our observed intragranular formed oxides in nanocrystalline materials for the other potential application, similar like the ODS steels.
2. The authors give too little information on the impurity levels and characteristics in the asreceived materials. It is not trivial to ascertain if low-T oxidation is abnormal or curious, as the authors claim, when we have no clear information on the impurity contents, other 23 than the balance numbers. This information should be available to the authors and will be very useful when drawing conclusions.
We appreciate the reviewer's comments very much on determining the impurity level in the raw materials. We have done systematic investigations on the impurity contents in raw powders and as-deformed HPT samples using XPS. As we replied to the fifth question (question # 5) raised by the first reviewer about the impurity, in the light of prudent measurements using XPS, we can determine that in our HPT deformed Cu-Fe samples from powders, the oxygen impurity level is 3.5 at%. We know the nanocrystalline alloys deformed by high pressure torsion by extremely large strains are in high non-equilibrium status. A large number of defects and boundaries are existed inside the materials, resulting in very high interfacial energy, elastic energy and so on, which is the reason for the thermally instability of the high-strained nanostructured alloys. As we discussed in the manuscript, the high stored energy in nonequilibrium conditions should be responsible for the low-temperature oxidation and decomposition behaviors in the severely deformed materials. Synchrotron X-ray diffraction results shown in Fig. R12 and the result of in-situ heating experiment on the arc-melted bulk sample (as a reference to show the oxygen content difference) displayed in Fig. R13 will testify the oxygen existence in the HPT deformed 75Cu-25Fe powder sample.
3. The language of the supplementary materials has to be improved significantly. There is a large discrepancy between the manuscript and the suppl.mater. There are also some small 24 issues with the manuscript itself, but these are minor and should be caught by a third party that proofs it.
We are grateful to reviewer's suggestions. The language in the revised version of manuscript has been checked by the native-English speaker, so it has been largely improved. It should be noted that all the mentioned samples were transferred into the XPS chamber at the same time and kept at the same condition. In addition, we have implemented three independent measurements on separate samples using the same procedures mentioned above, and the oxygen contents are 3.4 at.%, 3.3 at.% and 3.6 at% respectively. Therefore, these data can sufficiently support the conclusion of that oxygen content in the as-deformed sample is (3.43 ± 0.15) at% indeed. For simplification, it is written as 3.5 at.% in the as-deformed materials.
As for the concern of that such relatively high oxygen content of 3.5 at.% in Cu-Fe solid is far from the equilibrium solution amount, the explanation is that our materials were generated under 25 the high pressure torsion and Cu-Fe systems formed the single phase supersaturated solid solutions which were far from the conventional equilibrium states. Meanwhile, the oxygen atoms were dissolved into the fcc matrix, being assumed to occupy the octahedral interstices, which were also in high non-equilibrium states. Therefore, here, the oxygen existing in deformed alloys is different from the oxygen present in the conventional equilibrium solid solutions.
5. Vanilla DFT applied on transition metal oxides is a well-known problem area. Strong correlation effects can be quite important. The use of the enthalpies here thus calculated is not so significant in the current manuscript but still, one should do this properly or not at all.
In the first version of manuscript, the reason why we use DFT is to calculate the formation energies of likely oxides. We attempt to explain the formation of CuO and Fe 2 O 3 rather than other kind of oxides from the viewpoint of formation energies.
Due to the strong correlation effects in transition metal oxides, PBE 5 exchange-correlation function with LDA and GGA pseudo-potentials and Hubbard U model 6,7 were considered. Here, for Cu and Fe, U = 3 eV was employed. The results are shown in Table R2, which are in reasonable agreement with literature 8 . Although the strong correlation of transition metal oxides, combining our calculated results with the values given in literature 8 , we can draw the conclusion that CuO and Fe 2 O 3 possess the lowest formation energies in Cu and Fe oxides.

Table R2
Calculated formation enthalpies for possible different oxides (eV/atom). 26 6. The study in the suppl.mater motivating the claim in the main manuscript that the e-beam has no effect is far from conclusive. I would recommend the authors to improve the quantitative analysis of this effect (or lack thereof).
We appreciate very much that the reviewer commented on beam effect. As we mentioned in the manuscript, actually it is well known that in TEM studies the electron beam effect is nonnegligible, and should be considered because it may generate extra heat and facilitate the chemical reaction process. Especially in in-situ experiments, electron beam effects are unavoidable and should be taken into consideration.
During our experiments, we have taken some measures to minimize the potential influences of the electron beam. I) During image recording, the electron beam was spread into a specific size fittin completely to the fluorescent screen every time, and the beam was switched off during the heating process and the imaging was done immediately within 10 s after heating was finished. As pointed out by the reviewer, we have quantified the electron density on two images with magnifications of 600K and 800K respectively, recorded on the same area as shown in Fig. R10.

Reply to Reviewer #3:
We appreciate the reviewer's constructive comments very much.
One of the reviewer's main concerns is about the novelty of this work. In the revised version, we made this clear by rewriting the Abstract and Introduction parts to sort out the main information, and emphasize the importance of his study. To address this point and satisfy the referee, the main statements regarding the novelty are given as follows: Nowadays, production of applicable bulk nanocrystalline alloys using severe plastic deformation is of great importance for the next-generation high-performance structural materials, and it is drawing growing interests in the materials research field and industry applications. Unfortunately one prominent problem arising during the consolidation and straining processes is the unavoidable contamination from gaseous species for powders-processing technology, such as oxygen. Currently, people have realized that oxygen contamination can induce large discrepancies in the mechanical properties, microstructures and thermal stabilities. However, the exact state of oxygen contaminant inside the materials remains unclear, and how oxygen atoms behave during annealing has hardly been explored. Moreover, developing new nanostructured bulk materials for potential applications demands atomic-resolution insights into the structures, particularly, the understanding of oxygen behaviors in nanostructured materials. In this work, we employed the modern spherical aberration corrected high-resolution transmission electron microscopy to observe the oxygen behavior via in-situ annealing at the atomic scale. The results revealed that except decomposition process, the nanometer-sized oxide clusters could form inside the grains at specific temperatures. i) This is a first atomic-scale observation of the oxidation and decomposition process in bulk nanocrystalline alloys; ii) The finding in this study is of great significance in developing a new route for designing new bulk nanocrystalline alloys by 30 intentionally introducing different oxygen content, which can produce dispersive nano-sized oxides in the matrix so that the mechanical properties can be tuned as desired; iii) Actually, in general, this study will assist understanding the oxygen atom behavior in other material systems, such as in nanostructured materials, oxides and steels, and clarifying its role in the field of surface science, and catalysts.
To address all the technical concerns and consolidate in-situ HRTEM observations, we have performed systematic XPS studies regarding the oxygen contents in the initial and deformed materials, and synchrotron X-ray diffraction measurements on the powder sample and reference sample (arc-melted bulk material, supposed to be less oxygen content) with the same Cu-Fe composition under the identical measurement conditions. Most importantly, the in-situ heating experiment was carried out also on the reference sample (arc-melted bulk sample) with recording HRTEM images, being used as a comparative atomic-resolution experiment to address the effect difference when the different oxygen contents are involved in the two materials (powder sample and reference sample).
All other concerns will be addressed in detail as follows.
1. The oxygen analysis by XPS is not very convincing. I would be much more convinced if a piece of pure copper had been ion cleaned and analysed at the same time. Why was a quantitative bulk analysis technique not used to determine the oxygen content of the samples?
We thank the reviewer to point out the comparative method of the XPS measurement. As we replied for the fifth concern raised by the first reviewer, we have done a series of investigations to 31 systematically study the oxygen content in the as-deformed Cu-Fe alloys. The re-measured spectra and calculated constituents are shown in Fig. R7 and Table R1. Based on the different measurement results, it can be determined that the oxygen content in the HPT deformed Cu-Fe alloy is about 3.5 at.%.
As we mentioned before, for the XPS measurements, the most important point is to exclude the influence from the possible surface oxide layers. We have taken a series of measures to keep the sample surfaces unoxidized. We polished all the samples in the ethyl alcohol and then transferred to XPS chamber immediately at the same time. The sample surfaces were then severely sputtered by Ar ions with energy of 3 keV for 5 minutes to completely remove the possible surface oxide layers. By such measures, we can check all the samples at the same conditions and the results are accurate by our comparative study. It is very ideal to measure the oxygen content using quantitative bulk analysis techniques. Unfortunately, we could not find a proper bulk analysis technique which can fully exclude the possible influence of the surface oxide layers. Another advantage for XPS measurement is that we can obtain not only the constituents but also the valence states of the elements, by which their chemical environments corresponding to deformation behaviors can be judged.
2. What is the evidence that the oxide particles are growing within the sample during the insitu heating and not on the surface of the sample. Indeed the micrograph shown in Fig. 4a looks like a section through the edge of a thin sample with copper oxide growing between the amorphous surface layer on the left and the fcc substrate on the right. The ex-situ XRD data shown in Fig3a only demonstrate the growth of iron particles, not oxide, within the deformed matrix.

32
The same concern was raised by the first reviewer in the second question (question #2), and we have addressed it clearly from the point of view of formation of Moiré fringes. In addition, we have done a comparative study to show that if any oxides form only on the sample surfaces, Moiré fringes will be definitely observed. Fig. R11 shows HRTEM and FFT images of pure Cu foils exposed to air for a certain time and TEM sample of as-deformed 75Cu-25Fe alloy exposed to air for 2 days. From the FFT images, we can know that CuO layers formed due to the exposure to the air. The Moiré fringes indicate that the oxide layers have adhered to the sample surfaces.
The morphologies of oxides are absolutely different from those oxides grown within the sample during in-situ heating as shown in Fig. R5. An additional evidence to prove the oxide particles are growing within the sample, instead on the surface of the sample is to carry out the line-scan analysis crossing the oxide particle using EDXS/EELS (for instance, Fig.R8). The core intensity/signal at the oxide particles will give difference when the oxide particles form on the surface or grow within the sample. Fig. 4a in the first version of manuscript actually was taken from the in-situ annealed sample at the highest temperature of 420 ˚C. As we mentioned in the manuscript, oxygen will move inside the matrix lattice during heating, and it is possible for the oxygen to gather at the edge area via diffusion where the surface energy is high. From the morphology of the CuO shown in HRTEM image, it is more likely that a small edge part of the [011] grain underwent a chemical reaction from Cu to CuO during heating. Because no Fe and Fe 2 O 3 precipitates were mixed together at the thin area, we got opportunity to observe the atomic structures and the matching relationship between CuO and fcc matrix.
As for the XRD profiles shown in the Fig. 3a for the ex-situ annealed samples in the first version of manuscript, it was mainly for displaying the process of Fe decomposition. Actually, 33 we have tried to use the normal X-ray diffraction to detect the oxide formation process. But, in fact it is impossible because the signal-to-noise ratio was not high enough to indicate the trace amount of oxides present inside the material. Instead, we measured the ex-situ annealed samples using synchrotron method which has an extremely powerful energy to generate peaks even for the tiny amount components (as shown below). Synchrotron experiments were performed at the PETRA III P07 beamline at the DESY Photon Science facility (Hamburg, Germany).
First, we fabricated a reference sample by arc-melting with the composition of 75Cu-25Fe using the high purity commercial Cu and Fe rods (nominal purity: 99.99%). Because large pieces of Cu and Fe rods were used in arc-melting process, it could effectively reduce the influences of the surface oxides. Second, the powder sample and arc-melted sample with composition of 75Cu-25Fe were deformed by HPT to the same strains, followed by ex-situ annealing in Ar atmosphere at 420 °C at the same time. Then these two samples were cut and polished to the same shape and thickness. Synchrotron measurements were implemented on these two samples at the same conditions. Fig. R12a shows the synchrotron profiles of ex-situ annealed 75Cu-25Fe samples at 420 °C, which were deformed from powders and arc-melted bulk, respectively. (124) Fe2O3 respectively. We used the arc-melted sample with the same composition as a reference, and the oxides can be only detected for the powder sample. So the synchrotron measurements 34 provide a strong evidence to prove that the oxides formed inside the sample after annealing. The detailed results and discussion of the difference between powder sample and arc-melted bulk sample will be shown in our next paper.
35 Fig. R11 HRTEM and FFT images of (a) pure Cu foils exposed to air for more than 3 months and (b, c) TEM sample of as-deformed 75Cu-25Fe alloy exposed to air for 2 days. 3. I am also concerned about the way the (S)TEM samples have been prepared. Focussed ion beam thinning leaves the surface of the samples in a very reactive state, so that even removing a sample of iron from the vacuum can lead to the nucleation of surface oxides before it can be transferred to the microscope. I appreciate that the authors say that they have minimised the transfer time but, in my experience, this may not be sufficient.
Thanks for the concern raised by the reviewer. The same question was also asked by the first reviewer in the first question (Question #1). Our TEM samples were prepared using mechanical thinning method, and combined with low voltage ion-milling as a final step. As we mentioned in the reply to the first reviewer, we have done a series of experiments to minimize the influence from the ion-milling process. After the final gentle polishing by ion-milling with the parameters of lower limit conditions of 1.5 kV and 1˚ -1.5˚, we didn't see any clear ion damage from the HRTEM image, as shown in Fig. R4 (c, d).
As for the concern of surface nucleation of oxides or adsorption of oxygen, we agree the reviewer's opinion of that oxygen may somewhat adsorb on the surfaces of the sample as long as the sample is exposed to the air once, and it seems that it is inevitable during the entire experiments. However, what we are certain is that the influence from the surface oxygen adsorption can be very tiny on the in-situ heating experiment, and it doesn't affect the conclusion drawn from the current experiment. Our sample may be exposed to air only for 3 minutes (sometimes, less time 3 mins) during the transfer. Meanwhile, we did a test experiment to leave the sample inside the microscope for more than 12 hours, and then checked the sample with HRTEM, we didn't see any nucleated oxides or Moiré fringes as shown in Fig. R11. In addition, to confirm our conclusion, we carried out another comparative experiment like synchrotron measurements using above-mentioned 75Cu-25Fe arc-melted bulk material. We did the in-situ heating experiment on the 75Cu-25Fe alloy deformed from arc-melted bulk materials (which has less oxygen contents). Here we should emphasize that the TEM sample preparation method is 38 strictly the same as previous powder sample, and transfer time is controlled to be almost the same as previous experiment with about 3 minutes. Fig. R13 shows the HRTEM and corresponding  Table R2.
From another point, the synchrotron measurement also confirmed the formation of CuO and  5. The English used in the paper needs to be improved. In some places it even masks the message that the authors are trying to convey. For example, p3 l39. What does "no matter during powders consolidation" mean?
We really appreciate the reviewer's suggestion. The language of the resubmitted version of manuscript has been checked by the native-English speaker. We hope that our meanings can be conveyed clearly in the revised version.
Therefore, thickness change causes the slight contrast variations. II) Local distortion (originated from the defects, e.g. dislocations, point defects, stacking faults etc...) leads to contrast variations and local orientation change. In addition, heating enhance the defects mobility, which can significantly affect the local contrast variations and fringe visibility in different regions within one grain.
During the in-situ heating experiments, we usually tracked several grains, and selected the orientation-unchanged grains to record HRTEM images for analyzing the oxidation and decomposition behavior. Once RTEM images were taken over a large region, it may still contain some small areas slightly off the zone-axis, showing contrast variations and differences in fringe visibility. Thank the reviewer for very helpful comments.
It is true that the presence of Moiré fringes hardly provide information on whether the oxide is formed on the surface or within the TEM sample. Here, with the new APT data available, two approaches were proposed to identify the oxides location. 1) We supplemented atom probe tomography (APT) experimental data to show the oxygen distributions after ex-situ annealing at 300 °C. The APT results are included in the revised version of manuscript (Fig. 3d, also shown below). From the upper overall image, some Fe-rich areas (green color) embedded with oxygen atoms are 3 observed with dimensions of 20 -50 nm which accords with the results obtained from EELS mapping displayed in Fig. 1c. The bottom image shows that oxygen atoms distribute almost homogeneously except that some O-rich clusters with sizes of 3 -8 nm form within the TEM sample after ex-situ annealing. The APT provides a direct evidence to display the oxygen dissolution and oxygen clusters.  2) The second evidence is from the STEM-EELS mappings which can help to indirectly attest the oxides forming inside TEM sample. One line-profile is made crossing Fe-O particles (as indicated by white line), supplementary Fig. 6f (also shown below), Fe, O and Cu atom distributions are clearly shown. Note that the Cu concentration across the Fe-O particle decreases down to less than 10 at.%, approaching to ~ 0 when at the middle of particle location. If an oxide particle with a certain thickness is formed on the surface of TEM sample, the Cu concentration on the particle will not reduce to be ~ 0 at the middle location simple because of Cu substrate contribution. The line profiles in supplementary Fig.6 also confirm that the oxides formed within TEM sample. In addition, one may recognize that some big particles are formed at the grain boundary , and some small particles are nucleated in the interior of grains Thank the reviewer for pointing out the issues.
As mentioned earlier, the presence of Moiré fringes hardly provide information on whether the oxide is formed on the surface or within the TEM sample, as well as inside the grains or at the grain boundary. Therefore, we used the newly-acquired APT data and STEM-EELS maps to justify the location of oxides. We modified the text accordingly. We appreciate the reviewer for raising the questions. I) For nanocrystalline alloys, the small bright dots inside grains with different contrasts in STEM images could be due to many reasons, such as defects (dislocations and stacking faults), precipitates, ion-milling damages, and surface oxides etc... For Cu-Fe, here, these bright dot contrasts are due to the numerous dislocations and stacking faults. Figure R9 mentioned by the reviewer, actually it is difficult to judge which reasons cause the bright contrasts at grain boundaries only based on one ADF-STEM image. To address this query, we tracked this TEM sample and cleaned the sample completely by re-ion-milling for 1.0 h. After ion-milling the sample was transferred to the microscope within 3 minutes to minimize the possible surface oxidation. We recorded a series of BF, ADF, HAADF images (as shown below) at several thin areas (grain overlapping less likely). From these images, contrast variations at the grain boundaries were not observed. The following BF, ADF, HAADF images were recorded from two different positions after heavily cleaning of TEM sample (from the ex-situ 420 °C-annealed specimen. Via EDXS measurements, the dark areas marked by white arrows in HAADF images can be identified as Fe-rich particles.

II) As for the prior
Although these, it cannot be excluded that there are oxide areas at the grain boundaries.
Fortunately, with the APT data and STEM-EELS maps available, as shown in Question 2, it becomes clear. In short, we concluded that some of oxides are nucleated in the interior of grains, and some of them are formed at the grain boundary. 8

III)
As for the Figure R3d in prior 'reply to review comments', it is taken from the asdeformed 75Cu-25Fe sample without any heat treatment, and the TEM sample was intentionally subject to heavily ion-milling with a voltage of 6 kV and an angle of 8˚ for 20 min. So, there are a lot of damaged areas introduced, like amorphous intergranular regions embedded with 5 nm-sized crystallites, also including partially surface oxides. Because this sample was solely used for demonstrating the influence of ion-milling process, no special measures were taken to avoid the surface oxidation. 9 Reply to Reviewer #2: The authors have made a good effort at answering the questions I and the other referees raised, but I still do not agree on this being novel or innovative enough to merit publication in Nature Communications.
The justification given by the authors as reponse to this point are not sufficiently convincing. There is no doubt this merits publication, just not in this journal, in my honest opinion. I leave it to the editors to decide the matter.
Side note: In case of publication here or elsewhere, the re-written introduction would need some touching up language-wise. Phrases like "For decades, people usually think..." are not really up to standard for scientific literature.
We are really grateful to the reviewer #2 for positive assessment on the revised manuscript.
Regarding to the scientific expression suggested by the reviewer, we have carefully read through the Introduction part and corrected according descriptions. Since the reviewer still concerned about the novelty of this study, we add new arguments on the novelty of this study.
Because oxygen has a high electronegativity, its presence can have a marked effect on the properties of materials. This can occur through the binding of free electrons, which leads to a loss of electrical conductivity (e.g., in metal oxides) through the presence of vacancies in the oxygen sublattice. The presence of oxygen atom has a strongly effect on the mechanical properties of nanostructured materials via forming oxides or interstitial atoms, leading the distortion of crystal lattice as very recently shown for carbon interstitial atoms in steel [Ref.1]. Up to now, it is still unclear on the exact role of interstitial atoms behaves in materials, e.g. in nanostructured materials. This is urgently demanded for developing the novel nanostructured materials and for potential applications of such nanocrystalline materials in industry. 10 The work presented here is a first study on oxygen behavior in nanostructrued materials prepared by severely defromation. Our results revealed that in addition to the decomposition process, for the non-equilibrium nanostructured materials, the nanometer-sized oxide clusters could unexpectedly form at specific temperatures. Moreover, our recently obtained new APT (atom probe tomography) experimental data, together with STEM-EELS mappings, can clearly strengthen our novel observations. The main points for novelty of this work: i) This is the first atomic-scale observation of the oxidation and decomposition process in bulk nanocrystalline alloys, which shows an unexpectedly behavior, and the first study on oxygen impurity effects on the structure and properties of bulk nanostructure materials; ii) First experiments to demonstrate the thermal stability of such Cu-Fe nanostructured materials, providing the guide for application and study the stability of other relevant nanostructured alloys.
iii) The finding is of great significance in developing a new route for designing new bulk nanocrystalline alloys with tunable mechanical properties by intentionally introducing different oxygen contents, which can create nano-sized oxide dispersions in the matrix.
iv) This study will assist in understanding the dynamic behavior of oxygen in other material systems, i.e. oxides and steels, and clarifying the role of oxygen in the field of surface science and catalysis.