Review Reports - Usage of Barkhausen Noise for Assessment of Corrosion Damage on Different Low Alloyed Steels

Round 1

Reviewer 1 Report

Title:

The MC1100 steel is not ferritic, but martensitic, MC700 – mixed. The title is wrong from the point of view of materials science.

Does "destructive" Barkhausen noise exist??? What is non-destructive Barkhausen noise? Non-destructive is the assessment, not the Barkhausen noise!

Abstract:

Abbreviations should be introduced in the abstract

General comments:

The physical reasons of the decline of the BN rms value in corroded steels are not discussed properly. It seems eventually, that this decline is a mere consequence of the formation of surface corroded layer which lowers the sensitivity of the method. The intro section refers to a variety of obstacles that impede the motion of domain walls like dislocations, precipitates, etc. These mechanisms totally disappear afterwards as potential candidates. The intro section should be focussed then on the corrosion mechanisms and detection method, not on microstructural aspects of domain wall pinning-depinning.

The description of steels used is not sufficient. Are these steels commercial? Then the manufacturer should be indicated. What is "thermo-mechanical rolling at elevated temperatures"? Do they mean hot rolling? What are the rolling temperatures?

Grain size is not given in the description of materials. Therefore, the effect of grain size in discussion of the results is purely speculative.

MBN data for all three steels are given in the manuscript for rolling and transverse directions, whereas the Materials and Method section states that only MC700 and MC 1100 steels were rolled??? What is correct?

It appears that the corroded samples were used in different experiments in different states: with and without removing the "loose corrosion residues". MBN was studied for samples with removed surface layer, microstructure - without such removal, albeit this has not been stated explicitly. The compatibility of the experiments and results should be discussed.

Many statements in the discussion are confirmed neither by the results, nor by appropriate references. For example:

- Why the rolling direction is the easy direction and transverse - hard? Does this statement imply the texture?

- Reference to difference in grain sizes

- Comparison of the critical stress to induce martensite with the yield stress

- "Remarkable alteration of surface layer during rolling takes major role and MBN is lower for MC700 as compared with S235 as well as MC1100"

Other examples are the lines 208-212, 216-218.

All these statements must be justified, otherwise they look as the author´ fantasy.

Figure 9 and its description is totally confusing. How does the shift of the detection threshold affect this figure? MBN is actually a canonical example of crackling noise with universal power law distributions of the amplitude, energy and other parameters of individual events. So, if the threshold is moved, the effect of this variation on total number of pulses can be estimated. Does variation of threshold in the software affect only the number of pulses, or other parameters shown as well? In any case, Fig. 9 in the present form is senseless.

Only the last point in the Conclusions section is a conclusion, others are just repetitions of experimental observations.

Technical comments

What kind of "correlation analysis" is performed in Fig. 10??? What do the authors mean by "correlation analysis", making just a plot???

What the "remarkable thermodynamical heterogeneity of the surface microareas" is, line 135? What the "thermodynamical heterogeneity" is and how it was evaluated: the word "remarkable" implies certain reference point.

The thickness of corrosion layer cannot be appreciated in Fig. 3. All is gray in these images.

The message behind the photos of microstructures in Fig. 2 is not clear.

Figure 10 does not show saturation of the MBN with the thickness of corrosion layer whatsoever. This part of discussion is inherited directly from Ref. [9], without appropriate reflection.

Lines 185-186: … MBN emission in RD drops down at the expense of TD…? Fig. 5 shows the MBN envelope drops in both directions, what does “at the expense” mean?

Table 2 is senseless, number of days as a function of number of days???

There are numerous misprints, misuses and senseless statements in the text.

"MBN emission is driven by the grain size", line 177

"MBN is more rich", line 178

"Steel descending region up to 18 days...", line 156

"grow, grows" instead of "growth" throughout the manuscript

.......................................

Author Response

Reviewer: The MC1100 steel is not ferritic, but martensitic, MC700 – mixed. The title is wrong from the point of view of materials science.

Does "destructive" Barkhausen noise exist??? What is non-destructive Barkhausen noise? Non-destructive is the assessment, not the Barkhausen noise!

Response: We agree with the reviewer. We altered the title.

Manuscript: Usage of Barkhausen noise for assessment of corrosion damage on different low alloyed steels

Reviewer: Abbreviations should be introduced in the abstract

Response: We agree with the reviewer. We added explanations of all abbreviations in the abstract.

Manuscript: It can be reported that conventional rms value of Barkhausen noise signal as well as FWHM (Full Width at Half Maximum of Barkhausen noise envelope) can be linked with the corrosion extent, especially in the early phases of corrosion attack. The PP (Peak Position of Barkhausen noise envelope) values exhibit poor sensitivity.

Reviewer: The physical reasons of the decline of the BN rms value in corroded steels are not discussed properly. It seems eventually, that this decline is a mere consequence of the formation of surface corroded layer which lowers the sensitivity of the method. The intro section refers to a variety of obstacles that impede the motion of domain walls like dislocations, precipitates, etc. These mechanisms totally disappear afterwards as potential candidates. The intro section should be focused then on the corrosion mechanisms and detection method, not on microstructural aspects of domain wall pinning-depinning.

Response: The reason for the lower MBN along with increasing thickness of corroded layer is clearly indicated on page 6. But we agree that introduction should be re-addressed. Therefore, we added text.

Manuscript:

Presence of non-ferromagnetic or hard magnetic phase (for instance Fe oxides [10]) tends to attenuate MBN emission [9]. The increasing height of surface irregularities makes propagation of electromagnetic pulses more difficult [20]. Furthermore, a non-ferromagnetic layer on a ferromagnetic body makes weaker magnetising field which in turn decreases amplitude of acquired MBN pulses [21].

Reviewer: The description of steels used is not sufficient. Are these steels commercial? Then the manufacturer should be indicated. What is "thermo-mechanical rolling at elevated temperatures"? Do they mean hot rolling? What are the rolling temperatures?

Response: We added statement that the steels are commercially available. We replaced the term” "thermo-mechanical rolling at elevated temperatures" by the term “hot rolling”. We prefer to keep the manuscript without information about the sheet manufacturer. We do not know information about temperature of hot rolling.

Manuscript: Three different commercial low alloyed steels of nominal yield strength 235, 700 and 1100 MPa were employed in the study. 32 samples of size 25 x 50 mm of each steel were cut from sheets of thickness 5 mm (6 mm in the case of MC1100). The sheets of yield strength 700 and 1100 MPa were produced by hot rolling (further details are unknown).

Reviewer: Grain size is not given in the description of materials. Therefore, the effect of grain size in discussion of the results is purely speculative.

Response: We agree. Our cooperation institution carried out EBSD measurements a few weeks ago and the average grain size of the investigated steels was detected. Therefore, we added information about average grain size.

Manuscript: The average grain size obtained from the EBSD (electron back scattering diffraction) observations is 22 mm for the S235, 15 mm for the MC700 and 11 mm for MC1100 [22].

Reviewer: It appears that the corroded samples were used in different experiments in different states: with and without removing the "loose corrosion residues". MBN was studied for samples with removed surface layer, microstructure - without such removal, albeit this has not been stated explicitly. The compatibility of the experiments and results should be discussed.

Response: We disagree. The surface of all samples was cleaned in the same manner. Certain fraction of corroded layer was removed from the surface during cleaning. However, the certain volume of corrosion products (being firmly connected with the underlying matrix) retains on the surface and can be found on metallographic images. The cleaning of the surface was carried out in the manner as that potentially employed before MBN measurements on the real components.

Manuscript: We added explanation.

Certain fraction of the corroded layer was removed from the surface during the cleaning. However, the corrosion products firmly connected with the underlying matrix retain on the surface and can be found on metallographic images. The cleaning of the surface was carried out in the manner as that potentially employed before MBN measurements on the real components.

Reviewer: Why the rolling direction is the easy direction and transverse - hard? Does this statement imply the texture?

Response: The typical texture of the matrix was not observed in metallographic images (neither in EBSD images). The rolling direction (RD) was referred as easy direction due to higher MBN as compared with transversal one (TD).

Manuscript: RD direction represents the easy axis of magnetisation whereas TD represents the perpendicular hard axis (due to the lower MBN in TD as compared with RD).

Reviewer: Reference to difference in grain sizes

Manuscript: The average grain size obtained from the EBSD (electron back scattering diffraction) observations is 22 mm for the S235, 15 mm for the MC700 and 11 mm for MC1100 [22].

Reviewer: Comparison of the critical stress to induce martensite with the yield stress

Response: We do not know the exact critical stress to induce martensite in the matrix during the manufacturing. We only can observe that localized islands of martensite can be found for MC 700. We are carrying out further experiments with the different steels of the different yield strength (about 9 different steels). It was found that martensite free matrix can be found for MC 500 (of the nominal yield strength 500 MPa) and all steels of lower yield strength. As soon as the yield strength attains 700 MPa, the small regions of martensite can be found in the matrix and increase in fraction along with increasing yield strength of low alloyed steel.

Manuscript: We prefer no change.

Reviewer: "Remarkable alteration of surface layer during rolling takes major role and MBN is lower for MC700 as compared with S235 as well as MC1100"

Other examples are the lines 208-212, 216-218. All these statements must be justified, otherwise they look as the author´ fantasy.

Response: We altered the explanation and we also added new Figure 6 to make it clearer.

Manuscript:

For this reason, MC1100 emits stronger MBN as compared with S235 (finer grain size of MC1100 [22]). One might expect that the MBN for MC700 would be richer as compared with S235 due to the finer grain size. However, microstructure of the near surface region (lying within the MBN reading depth) is rougher as contrasted against the deeper layer which in turn contributes to the lower MBN, see Figure 5 and Figure 6 (see also Figure 1b).

Reviewer: Figure 9 and its description is totally confusing. How does the shift of the detection threshold affect this figure? MBN is actually a canonical example of crackling noise with universal power law distributions of the amplitude, energy and other parameters of individual events. So, if the threshold is moved, the effect of this variation on total number of pulses can be estimated. Does variation of threshold in the software affect only the number of pulses, or other parameters shown as well? In any case, Fig. 9 in the present form is senseless.

Response: We agree with the reviewer. Reviewer note is very useful for us and we have to reconsider the contribution of floating threshold function influence with respect of MBN (rms, PP as well as FWHM). We carried out some tests and we found that the floating threshold valuably affects rms of MBN, PP and FWHM when the effective (rms) value of the MBN falls below 80mV. Being so, our published data (for MBN, PP and FWHM) retain untouched.

Manuscript: We decided to remove the part of the manuscript dealing with the number of MBN pulses (and all corresponding texts).

Reviewer: Only the last point in the Conclusions section is a conclusion, others are just repetitions of experimental observations.

Response: We altered the Conclusions.

Manuscript:

Results of the measurements prove that MBN technique could be potentially employed for monitoring of components exposed to the variable degree of corrosion attack. With respect of magnetic measurements such as MBN, corroded layer on the surface is the hindering barrier. Its role becomes stronger along with its increasing thickness. Effective value of MBN signal exhibits good sensitivity against of corrosion attack especially in its early phases followed by the less remarkable decrease. Reliable monitoring of corrosion damage through the thicker corroded layers can be carried out when the combination of FWHM and effective value of MBN is employed only. Moreover, information about the rolling direction of the body should be known or measured. The combination of the different MBN features extracted from the MBN signals becomes more important for the low alloyed steels of higher strength since the region in which MBN steep decreases occurs is shorter. This paper represents the pilot study since the combination of corrosion damage and the superimposing contribution of stresses should be expected in the real applications.

Reviewer: What kind of "correlation analysis" is performed in Fig. 10??? What do the authors mean by "correlation analysis", making just a plot???

Response: We agree. Statement about correlation analysis is not correct.

Manuscript: We removed the following statement “Correlation analysis as has been shown in” and the new sentence is “Figure 10 demonstrates that MBN (its rms values) exhibits good sensitivity against the thickness of corroded layer for all steels in RD as well as TD.”

Reviewer: What the "remarkable thermodynamical heterogeneity of the surface microareas" is, line 135? What the "thermodynamical heterogeneity" is and how it was evaluated: the word "remarkable" implies certain reference point.

Response: We agree. This statement can be misinterpreted. For this reason, we altered the text to better explain reason of the variable thickness of corroded layer.

Manuscript:

Figure 3 and Figure 4 also demonstrate that the thickness of corroded layer strongly varies and the regions, in which the corroded layer is quite thick neighbouring with those of limited thickness or/and free of the corroded layer as a result of complex nature of corrosion process in which small regions exposed to the corrosion attack become relatively anodic or cathodic. For this reason, the thickness of corroded layer in some regions is increasing at the expense of the neighbouring ones.

Reviewer: The thickness of corrosion layer cannot be appreciated in Fig. 3. All is grey in these images.

Response: We understand the reviewer statement. In some cases, it could be difficult to distinguish corrosion from mould or/and red tape. However, corrosion on the low alloyed steels appears on metallographic images as a grey region usually – our experience. Moreover, the different Fe oxides can appear in the different scales of grey colour. But from our point of view the corroded layer still can be reliably recognized.

Manuscript: We prefer no change.

Reviewer: The message behind the photos of microstructures in Fig. 2 is not clear.

Response: We agree, that the usage of these images from LM is debatable, but we wanted to show and point out, how confusing the usage of just visual inspection could be in terms of complexity of corrosion processes on steels. This should be the proof of relevance for usage of Barkhausen noise technique alternative for more precise non-destructive detection of real corrosion extent. We have rewritten all the paragraph, so that our motivation is clearer.

Manuscript: Visual observation of corroded samples showed that corrosion attack is not developed in the same rate with respect of the different steels. It is worth to mention that evolution of corrosion is strongly heterogeneous with respect of the different microanodic and microcathodic regions within one sample [1, 9]. For this reason, corroded region neighbouring with non-corroded one or/and extent of corrosion towards the deeper layers strongly varies. Figure 2 indicates the heterogeneity and coverage of corrosion products within the sample’s surface. All the surfaces of tested steels were completely covered by various types of iron oxides, hydroxides and oxohydroxides (proved by different colour and structure of corrosion product) even after 11 hours of exposure in salt spray atmosphere (Figures 2d-f) as a result of high humidity, temperature, presence of oxygen and chloride ions accelerating the corrosion process. Depending on local thermodynamic conditions changing over time within the corrosion process, different types of corrosion products are formed or transformed locally on the sample´s surface explaining the mentioned heterogeneity. However, as the heterogeneity of corrosion products is very similar within all the tested samples and all the surfaces were completely covered by corrosion products even after short exposure times, the testing method of surface evaluation by visual surface inspection cannot be used as a relevant non-destructive tool for assessing the extent of corrosion attack. Real differences in corrosion extent are visible from the cross section of individual samples (Figure 3) and interpretation of corrosion products thickness growth is presented in Figure 4. This different corrosion rate is related to various microstructure, chemical composition and performed hot rolling as all the other conditions and were the same for all the samples. However, it is not possible to use cross section analysis in real conditions, hence, another non-destructive technique, like Barkhausen noise emission should be employed to assess the real corrosion extent.

Reviewer: Figure 10 does not show saturation of the MBN with the thickness of corrosion layer whatsoever. This part of discussion is inherited directly from Ref. [9], without appropriate reflection.

Response: We agree with the reviewer. We removed this statement.

Manuscript: This statement “However, these evolutions saturate for the higher thickness of corroded layer” was removed.

Reviewer: Lines 185-186: … MBN emission in RD drops down at the expense of TD…? Fig. 5 shows the MBN envelope drops in both directions, what does “at the expense” mean?

Response: It means that domain walls tend to rotate to TD. For this reason, MBN is decreasing whereas the TD is increasing. However, such evolution can be found in the deeper layers since the microstructure in the near surface region can differ as Figure 6 indicates. We altered the text and added one figure (new Figure 6).

Manuscript: We removed this paragraph due to possible reader misinterpretation and we added new Figure 6 as well as the corresponding text.

Reviewer: Table 2 is senseless, number of days as a function of number of days???

Response: We agree that the content of the Table 2 can be misunderstanding. For this reason, we altered appearance of this table.

Manuscript: Please check the appearance of Table 2.

Reviewer: "MBN emission is driven by the grain size", line 177

Response: We altered the text.

Manuscript: MBN emission can be also affected by grain size and phase transformation (strain induced martensite transformation).

Reviewer: "MBN is more rich", line 178

Response: We altered the text.

Manuscript: For this reason, in this particular case MBN originating from ferrite of finer grain size is richer as a result of…

Reviewer: "Steel descending region up to 18 days...", line 156

Response: Of course typo mistake – “steel” replaced by “steep”.

Manuscript: Steep descending region up to 18 days is followed by less remarkable, but still valuable decrease in MBN.

Reviewer: "grow, grows" instead of "growth" throughout the manuscript

Response: We altered the requited text.

Manuscript: However, there are real differences in corrosion extent visible from the cross section of individual samples (Figure 3) and interpretation of corrosion products thickness grow is presented in Figure 4.

Author Response File: Author Response.doc

Reviewer 2 Report

The manuscript is interesting and well presented. It needs extension on the test methodology (even references to proper test standards), the reader should be able to reproduce. Furthermore, the discussion on test results and analysis should be extended and improved providing more information. Please add quantitative results in your conclusions.

English language is generally ok but it can be improved in some phrases and terms. For example the "salt environment" could be written somehow as Cl- attack.

Author Response

Reviewer: The manuscript is interesting and well presented. It needs extension on the test methodology (even references to proper test standards), the reader should be able to reproduce.

Response: We added reference n. 23 for test standard.

Manuscript: The samples were corroded in the neutral salt spray atmosphere with respect of STN EN ISO 9227 standard [23] in the VSC KWT 1000 chamber (temperature of 35 °C, pH in the range from 6.5 to 7.2, pressure in the chamber 120 kPa).

STN EN ISO 9227: Corrosion tests in artificial atmospheres — Salt spray tests, Slovak technical standard. SK version of ISO 9227: 2017. Release date: 1.7.2018, 18 p.

Reviewer: Furthermore, the discussion on test results and analysis should be extended and improved providing more information. Please add quantitative results in your conclusions.

Response: We added new explanation in discussion part. We also altered the Conclusions, but we prefer to edit the Conclusion part in more general way without exact data.

Manuscript:

Corroded layer on the underlying ferromagnetic matrix is a barrier for magnetisation of the sample and transfer of MBN pulses from the untouched surface towards the sensing coil. Corroded layer can contain non-corroded ferromagnetic particles. However, their size as well as the corresponding amplitude of produced MBN pulses drops down along with the increasing degree of corrosion damage due to the increasing fraction of 90° DWs at the expense of 180° DWs and the reduced free path of their motion [9]. For this reason, these non-corroded particles cannot fully compensate zero MBN emission originated form the non-ferromagnetic or hard ferromagnetic oxides [10] and MBN emission drops down along with more developed corrosion damage.

MBN emission originating from feritic steels is usually higher as compared with martensitic steel [27] as a result of higher dislocation density and high pinning strength of carbides [16, 17] after quenching. However, nearly no precipitates in the form of carbides (as very strong pinning sites) can be found in the low alloyed steels due to the missing elements creating the simple or/and complex carbides (see Table 1). Therefore, the fine grained structure and the corresponding higher density of DWs in motion can be reported as the main reason for stronger MBN [28] originating from the martensitic MC1100 as contrasted against the feritic S235. The higher hardness and the corresponding higher dislocation density of MC1100 (as well as MC700) [22] only increases the magnetic field in which DWs are unpinned (PP for MC1100 and MC700 are more as compared with S235, see also Figure 8) but takes only the minor role in the amplitude and density of produced MBN pulses.

Reviewer: English language is generally ok but it can be improved in some phrases and terms. For example, the "salt environment" could be written somehow as Cl- attack.

Response: We made some alterations of the text (as in was indicated by reviewers). Please check the manuscript.

Manuscript: Three different steels of nominal yield strength 235, 700 and 1100 MPa were subjected to the variable degree of corrosion attack developed in the corrosion chamber under neutral salt spray (NSS) atmosphere.

Reviewer 3 Report

The corrosion damage of low alloyed feritic steels of variable strength was investigated by the use of conventional metallographic observations and Magnetic Barkhausen noise in this study.

The findings are interesting and can be published on Appl. Sci.

Also, there have some points must be clarified.

It is confused that the manuscript is mainly deal with the different low alloyed ferritic steels, but MC1100 steel is totally martensitic steel given in Fig. 1.
The method of MBN measurement is not clearly described. Line 75-77, the authors stated that “Loose corrosion residues were manually removed from the surface of each sample and subsequently the surface of samples was water cleaned and dried”. So is the measurement of MBN carried out after the loose corrosion residues is removed? If so, how to evaluate the degree of corrosion attack?
What is the relationship between the microstructure (ferrite, martensite) and the MBN signal?

Author Response

Reviewer: It is confused that the manuscript is mainly deal with the different low alloyed ferritic steels, but MC1100 steel is totally martensitic steel given in Fig. 1.

Response: We agree with the reviewer. We altered the title.

Manuscript: Usage of Barkhausen noise for assessment of corrosion damage on different low alloyed steels

Reviewer: The method of MBN measurement is not clearly described. Line 75-77, the authors stated that “Loose corrosion residues were manually removed from the surface of each sample and subsequently the surface of samples was water cleaned and dried”. So is the measurement of MBN carried out after the loose corrosion residues is removed? If so, how to evaluate the degree of corrosion attack?

Response: The surface of all samples was cleaner in the same manner. Certain fraction of corroded layer was removed from the surface during cleaning. However, the certain volume of corrosion products (being firmly connected with the underlying matrix) retains on the surface and can be found on metallographic images. The cleaning of the surface was carried out in the manner as that potentially employed before MBN measurements on the real components.

Manuscript: We added further explanation.

Reviewer: What is the relationship between the microstructure (ferrite, martensite) and the MBN signal?

Response: We added further explanation.

Manuscript:

Reviewer 4 Report

In this manuscript, conventional metallographic approach and Barkhausen noise-based non-destructive magnetic technique were combined to investigate the temporal evolution of corrosion damage in three steels (different chemical compositions, microstructures, etc.) The measurements are systematic, and the results are interesting and decently presented.

My biggest concern with this manuscript is that it lacks sufficient discussions about the Materials Science behind these measurements. I have seen the authors mentioned surface, microstructure, and yield strength, but the detailed discussions and key messages of these are either not included or highlighted enough. Look at the conclusions on Page 11, the listed findings are simply reports of measurements. There aren't extra flavors in the underlying mechanisms, e.g. how the surface or microstructure actually changes during corrosion

All in all, while the results are good, this manuscript lacks sufficient discussions especially the underlying Materials Science. The present form of this manuscript seems to be plain report. Major editing and rewording are expected in revisions, and some measurements of microstructural evolution may be helpful.

Author Response

Reviewer: My biggest concern with this manuscript is that it lacks sufficient discussions about the Materials Science behind these measurements. I have seen the authors mentioned surface, microstructure, and yield strength, but the detailed discussions and key messages of these are either not included or highlighted enough.

Reviewer: All in all, while the results are good, this manuscript lacks sufficient discussions especially the underlying Materials Science. The present form of this manuscript seems to be plain report. Major editing and rewording are expected in revisions, and some measurements of microstructural evolution may be helpful.

Response: We added further texts.

Manuscript:

Reviewer: Look at the conclusions on Page 11, the listed findings are simply reports of measurements. There aren't extra flavours in the underlying mechanisms, e.g. how the surface or microstructure actually changes during corrosion

Response: We altered the Conclusions.

Manuscript:

Round 2

Reviewer 1 Report

The reviewer upholds the initial decision to reject the manuscript. The authors have addressed only easy to deal obvious mistakes pointed out by the reviewer. Important issues remain unanswered.

The authors make fundamental mistakes not only in materials science, but in magnetism as well. No texture in polycrystals means no easy direction. Moreover, except for MC 1100 there is no difference whatsoever in the RMS value of MBN for TD and RD, which is used by the authors to claim this “easy direction”. The only difference is the scatter of data.
The issue of strain induced martensite remains open. The reviewer asked to clarify this point. The authors eliminated the main part of the associated passage but still insist on strain induced martensite. The reviewer is not metallurgist, but, according to his knowledge, martensite in steels is induced by heat treatment, not during hot rolling. Since the Ms temperature is some 500 C below the eutectoid isotherm, it is impossible to induce martensite by stresses lower than the yield stress at elevated temperatures of hot rolling.
Most importantly, the reviewer pointed the important issue of the effect of reducing the detection threshold on MBN parameters presented. He asled to clarify what would be the effect of reducing the threshold on all other parameters. The authors eliminated Fig. 9 as not appropriate and ignored the question: all other dependences remain the same. However, if the authors report the effect of corrosion, they should keep the conditions of measurements unchanged, otherwise the result is a kind of a convolution of the parameters of the sample and of the software employed. Then, the results reported are valid only for these specific experimental conditions, software and hardware used.
The authors insist on keeping Fig. 3 as is. The reviewer insists that the quality of the microstructure image is poor, everything is gray in these images. Moreover, the authors should have taken the opportunity, using these images, to indicate how they measured the thickness of the corrosion layer.
Addressing the issue of microstructure, the authors added Ref. [22] in the revised version. This reference is not accessible, however.

Reviewer 2 Report

Reviewer 4 Report

I thank the authors' responses. I believe that the added paragraphs have increased the depth of this manuscript, and the underlying Materials Science has been discussed sufficiently. Therefore, I would like to recommend acceptance to this manuscript.