Study of Corrosion Characteristics of AlMg3.5 Alloy by Hydrogen-Induced Pressure and Mass Loss Evaluation Under Simulated Cementitious Repository Conditions

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

General comment – The data presented in this work is minimal. It cannot support assumptions or the creation of a model by itself. It can only be used as a part of a larger database for such uses.

Specific comments

Lines 50-52, please use the specific term "carbonation". Note that the pH of a fresh Portland cement paste is 13.4-13.6. This implies that the alkalinity results from alkaline hydroxide, not calcium hydroxide.

Lines 53-55, aluminum alloys do not have a passivation layer in an alkaline environment. Please refer to the Pourbaix diagram of aluminum.

Line 62, the 4- should be superscript.

Eq. 1-4, please indicate which process is relevant to which environment.

CW – its composition is far from pore-solution composition. Its pH is too low. There is no silicate. It is not clear why increasing ionic strength is needed if there are no electrical tests. The terms MQ and Eppendorf should be explicitly presented at their first use.

Line 115 – Do you mean that the number of specimens per treatment is n=2?

The same order of topics is expected in methods and results. That means the corrosion rate before X-ray. If X-ray is a subsection, so should be the corrosion rate.

The X-ray image (Figure 3) is not clear. There is no scale bar. The difference between the first two images is not clear. It is not clear if the third one is a section of the rod or concrete with a rod inside.

Line 158 – one cannot differentiate these things with the quality of the image supplied. Use an image with an enlargement of this region to support your words.

Figures 4 and 5 – It is not clear what the differences are between these figures. Why should concrete and CW be on different figures?

Line 168 (and on) – If n=2, then you should specify how uncertainties were calculated and what the confidence interval is, or replace the term with the term range.

Lines 182-189 look like a repetition of the last paragraph.

The discussion should follow the same order as the methods and results.

It is not clear from the images how the corrosion at 9 months is less visible. You should compare rod in concrete to rod in concrete and rod in CW to rod in CW. The usage of citation in this context is not clear. There is no data to support the assumptions. Darker spots or areas may result from bubbles.

There is a missing word on line 195: "…of the, in…"

Line 208 – Are these solids expected in the assumed pH? The corrosion rate may decrease due to other factors, like dehydration, precipitation of calcium hydroxide, and changes in the immediate environment pH. In case that references 1,6-8, and 19, all of them, found evidence for the proposed mechanism, you should state it implicitly and not present it as your deduction, since you have no evidence to support it. Nevertheless, several of the cited references reported investigations in a completely different environment. You cannot use them for this reasoning.

Lines 219-220 – not clear, please rephrase.

Line 244 – What is the confidence interval of the extrapolation?

Table 1. Where do the "duplicate in autoclave" come from? The standard deviation from a duplicate cannot be compared directly to that of a triplicate. See https://en.wikipedia.org/wiki/Student%27s_t-distribution#Occurrence_and_applications

With this range of confidence intervals, you should check what the probability is of having a real difference between aluminum and the R2 alloy corrosion rates in concrete.

The real (and important) conclusion of your work is that CW of the composition used cannot be used to simulate aluminum alloy corrosion in concrete.

 

Author Response

General comment – The data presented in this work is minimal. It cannot support assumptions or the creation of a model by itself. It can only be used as a part of a larger database for such uses.

Thank you for the general comment. We added another data point for the mass loss evaluation in concrete.

Lines 50-52, please use the specific term "carbonation". Note that the pH of a fresh Portland cement paste is 13.4-13.6. This implies that the alkalinity results from alkaline hydroxide, not calcium hydroxide.

Thank you for the important observation, the term carbonation has been added and the part has been changed.

 

Lines 53-55, aluminum alloys do not have a passivation layer in an alkaline environment. Please refer to the Pourbaix diagram of aluminum.

Thank you for the valuable feedback, the term “passivation layer” was incorrect and has been changed to oxide layer.

 

Line 62, the 4- should be superscript.

The 4 is supposed to be subscript while the – is supposed to be superscript. It has been changed to [Al(OH)₄]^- now since the original format could have led to misinterpretation.

 

Eq. 1-4, please indicate which process is relevant to which environment.

Thank you for the opportunity to clarify this point. The processes relevant for this research are mentioned in Line 58-64. Possible addition might be the limited dissolution of the solid products (alumina and aluminum hydroxide) at these pH values. This is partly mentioned through the dominating aqueous phase.

 

CW – its composition is far from pore-solution composition. Its pH is too low. There is no silicate. It is not clear why increasing ionic strength is needed if there are no electrical tests. The terms MQ and Eppendorf should be explicitly presented at their first use.

Thank you for this comment. The CW composition is very simplified and does not attempt to mimic actual conditions. Caused conditions are closer to old/weathered cement water conditions and therefore simulates conditions for longer-term repositories. Added salt is supposed to represent granitic ground water. Changes were made in text (103, 226, 242).

Terms MQ and Eppendorf are also presented now.

 

Line 115 – Do you mean that the number of specimens per treatment is n=2?

Thank you for the question. The experiments in autoclaves were performed in duplicates.

 

The same order of topics is expected in methods and results. That means the corrosion rate before X-ray. If X-ray is a subsection, so should be the corrosion rate.

Thank you for the observation. As suggested, the order of topics has been adjusted to ensure consistency between the methods and results.  

The X-ray image (Figure 3) is not clear. There is no scale bar. The difference between the first two images is not clear. It is not clear if the third one is a section of the rod or concrete with a rod inside.

That is an important observation. The difference between the first two is minor and explained in the text. It is difficult to add a scale bar since it was an exploratory analysis without definite measuring of the samples. The point of this figure is to show the optical differences between the different corrosion times. It is meant to be a data-supporting figure and therefore its placement was changed to the end of the results/discussion. It also highlights the difficulty of imaging an embedded sample.

 

Line 158 – one cannot differentiate these things with the quality of the image supplied. Use an image with an enlargement of this region to support your words.

Thank you for your assessment. These x-ray images are an exploratory analysis and are not used singularly to explain certain processes, but to support other evaluations. Also, it can point out that the imaging in concrete is difficult since densities of the oxide and the aluminum are similar.

 

Figures 4 and 5 – It is not clear what the differences are between these figures. Why should concrete and CW be on different figures?

Thank you for the question. The difference is that one is in concrete and one is in CW. It is displayed on different figures to point out differences in corrosion rates in both of them. Large scales and many data points would complicate understanding of the graphics if displayed in one.

 

Line 168 (and on) – If n=2, then you should specify how uncertainties were calculated and what the confidence interval is, or replace the term with the term range.

Thank you for the note, the method of calculation has been added in the text as suggested.

 

Lines 182-189 look like a repetition of the last paragraph.

Thank you for the comment. It sounds repetitive because the graphics show similar processes in different media.

 

The discussion should follow the same order as the methods and results.

Thank you for the observation. The order of the topics has been adjusted to ensure consistency between the methods and results.

 

It is not clear from the images how the corrosion at 9 months is less visible. You should compare rod in concrete to rod in concrete and rod in CW to rod in CW. The usage of citation in this context is not clear. There is no data to support the assumptions. Darker spots or areas may result from bubbles.

Thank you for your valuable feedback. We tried to explain the difference between the samples and the figure is meant to be supportive of the observed corrosion rates. Yes, the darker spots could also be a result of the bubbles, as briefly discussed in the manuscript.

 

There is a missing word on line 195: "…of the, in…"

Thank you for the observation. The sentence has been rephrased as suggested.

 

Line 208 – Are these solids expected in the assumed pH? The corrosion rate may decrease due to other factors, like dehydration, precipitation of calcium hydroxide, and changes in the immediate environment pH. In case that references 1,6-8, and 19, all of them, found evidence for the proposed mechanism, you should state it implicitly and not present it as your deduction, since you have no evidence to support it. Nevertheless, several of the cited references reported investigations in a completely different environment. You cannot use them for this reasoning.

Considering the speciation diagram of aluminum, they are not expected. However, they are known to be products of acute corrosion of aluminum alloys at alkaline conditions. Their dissolution is limited (e.g., J. Zhang, M. Klasky, and B. C. Letellier, “The aluminum chemistry and corrosion in alkaline solutions,” Journal of Nuclear materials, vol. 384, no. 2, pp. 175-189, 2009.)

Thank you for pointing out which other factors could lead to the decrease of corrosion rates. Following, we try to explain why we focused on the oxide layer to explain it. Dehydration will not be a factor for the decrease of the corrosion rate since the W/C ratio is chosen that even with full corrosion of the sample, the water would not be depleted in the concrete. Precipitation of calcium hydroxide is unlikely to cause a decrease as strong and as fast as detected. Changes in immediate environment pH could be a reason but not after that short amount of time, only with weathered cement. The citied literature is at alkaline conditions which could simulate concrete environment or at lower pH but with high salinity which causes similar effects in aluminum corrosion rates compared to high pH.

 

Lines 219-220 – not clear, please rephrase.

Thank you, the lines have been rephrased as suggested.

 

Line 244 – What is the confidence interval of the extrapolation?

The fit is a power fit calculated with Origin at a 95 % confidence interval.

Table 1. Where do the "duplicate in autoclave" come from? The standard deviation from a duplicate cannot be compared directly to that of a triplicate. See https://en.wikipedia.org/wiki/Student%27s_t-distribution#Occurrence_and_applications

Thank you for the question. The duplicates come from two separate replicates in autoclaves. This table shows the evolution of the R2 alloy over time with two different methods of measuring the corrosion rate, so it is not meant to be for direct comparison. The aluminum metal in concrete from literature is just for reference to see the similar rates between both materials.

 

With this range of confidence intervals, you should check what the probability is of having a real difference between aluminum and the R2 alloy corrosion rates in concrete.

Thank you for the important comment. It has been changed. There is no reason to believe that it is actually significantly higher after 12 months.

 

The real (and important) conclusion of your work is that CW of the composition used cannot be used to simulate aluminum alloy corrosion in concrete.

Thank you for your valuable comment. The CW is a reference and with the composition used, it can simulate weathered cement conditions with lower pH. The main conclusion, however, is that the R2 alloy has similar corrosion characteristics as aluminum metal.  

Reviewer 2 Report

Comments and Suggestions for Authors

This is an interesting work providing important data on corrosion of the Al-Mg3.5 alloy in contact with prepared cement paste and cement water. Authors embedded metallic samples in concrete and monitored the hydrogen-induced pressure build-up. They report for concrete corrosion rates initially as high 10 mm/y which gradually decreased to less than 500 μm/y after 2000 hours. Authors have also analysed corrosion in cement water where samples showed similar behaviour. The paper is recommended to publish after some minor amendments to account for the followings:

Lines 53-55: The statement “These materials are susceptible to accelerated initial corrosion under alkaline environments, owing to the rapid formation of a passivation layer through surface oxidation by water” is controversial, as it is a mixture of correct observation about “the rapid formation of a passivation layer through surface oxidation by water” and wrong observation that “materials are susceptible to accelerated initial corrosion” due to this. The situation in this case is just opposite, the susceptibility of these materials “to accelerate initial corrosion” is due not to the formation but to the corrosion of passivation layers in the high pH pore water of cements.

Lines 64-65: Authors might note that the formation of cracks was earlier detected, see for details the dedicated publications: L.M. Spasova et al., Acoustic emission detection of microcrack formation and development in cementitious wasteforms with immobilised Al. J. Hazard. Mat., A138, 423-432 (2006) and Characterisation of Al corrosion and its impact on the mechanical performance of composite cement wasteforms by the acoustic emission technique. J. Nucl. Mater., 375, 347-358 (2008).  

Line 72-73: Previous investigations have also included acoustic emission method, e.g. see the above noted references.

Figures 2 and 3 need scale bars for images shown.  

Figure 3 needs an explanatory text in its legend for the inset shown.

Author Response

Lines 53-55: The statement “These materials are susceptible to accelerated initial corrosion under alkaline environments, owing to the rapid formation of a passivation layer through surface oxidation by water” is controversial, as it is a mixture of correct observation about “the rapid formation of a passivation layer through surface oxidation by water” and wrong observation that “materials are susceptible to accelerated initial corrosion” due to this. The situation in this case is just opposite, the susceptibility of these materials “to accelerate initial corrosion” is due not to the formation but to the corrosion of passivation layers in the high pH pore water of cements.

Thank you for this important comment. The mentioned paragraph has been rephrased now as suggested.

 

Lines 64-65: Authors might note that the formation of cracks was earlier detected, see for details the dedicated publications: L.M. Spasova et al., Acoustic emission detection of microcrack formation and development in cementitious wasteforms with immobilised Al. J. Hazard. Mat., A138, 423-432 (2006) and Characterisation of Al corrosion and its impact on the mechanical performance of composite cement wasteforms by the acoustic emission technique. J. Nucl. Mater., 375, 347-358 (2008).  

Line 72-73: Previous investigations have also included acoustic emission method, e.g. see the above noted references.

Thank you for the note, we have added a paragraph about the observation of cracks in the discussion and mentioned the previous detection by acoustic emission.

 

Figures 2 and 3 need scale bars for images shown.  

 

Thank you for the advice. A scale bar has been added to Figure 2. Figure 3 is hard to scale since it is hard to scale the sample in concrete and might lead to confusion.

 

Figure 3 needs an explanatory text in its legend for the inset shown.

Thank you for your feedback. We assumed the observation refers to Figure 4, the legend has been revised to included explanatory description of the inset as suggested.

(In new version Figure 3 = Figure 5, Figure 4 = Figure 3 and Figure 5 = Figure 4).

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Corrosion characteristics of AlMg3.5 alloy by hydrogen-induced pressure and mass loss evaluation under simulated cementitious repository conditions R2

General comment—The data presented in this work is minimal. It cannot support assumptions or the creation of a model by itself. It can only be used as part of a larger database for such uses. Adding additional data points does not resolve the problem that one cannot infer a mechanism from corrosion rate alone.

Give a reference to some papers with compositions of CW

It is unclear from the changes that the CW is intended to mimic an old concrete in a granite environment. Please be more explicit and give a reference to support your intention.

It should be stated explicitly what is the number of replications is for each experiment.

Figure 3 – it is not clear what the intention of the graph in the inner frame is. Why don't you use a logarithmic scale?

It will be interesting to show the corrosion rate in concrete versus in CW with a phase of about 250 hr. If it yields a linear line, it will hint that the CW method can be used to predict the corrosion rate of this alloy in concrete. (let's remember that the concrete pH is much higher, and may be reduced in the vicinity of the alloy due to the corrosion products)

The method to create a scale for the x-ray images is to measure the distance of the bar width in pixels. Since the bar thickness change is known and much lower than its thickness, you will get a scale of pixels per millimeter, which can be used to draw a scale on the image.

The discussion regarding the decrease in corrosion rate, including possibilities that can be eliminated, should be in the text, not only in your answer. Nevertheless, since your research cannot confirm any mechanism, the question should remain open. Notice that the alkaline corrosion process of aluminum reduces the pH and consumes water. As a consequence, a pH and a water potential gradient may occur. As the process continues, the thickness of the affected zone may increase and cause the logarithmic decay. However, your discussion should state everything explicitly. Writing only "the oxide layer" is not good enough. You can adopt J.R. Galvelle model for pH polarization in pit corrosion, with the necessary changes.

Please choose one of the options: 1) draw the 95% confidence interval envelope on the graphs; or 2) when you write a predicted number, add the limits of the confidence interval (and write it explicitly).

Author Response

Give a reference to some papers with compositions of CW

Thank you for the advice. References are given now in the discussion of the corrosion rates.

It is unclear from the changes that the CW is intended to mimic an old concrete in a granite environment. Please be more explicit and give a reference to support your intention.

Thank you for the valuable comment. It is now mentioned that crystalline bedrock conditions in a repository are mimicked by the water composition.

It should be stated explicitly what is the number of replications is for each experiment.

Thank you for the comment. For pressure measurements, it was mentioned in Line 120. For mass loss evaluation, it is also mentioned now in Line 122.

Figure 3 – it is not clear what the intention of the graph in the inner frame is. Why don't you use a logarithmic scale?

Thank you for this question. We decided to switch to a logarithmic scale now and leave out the inner frame to simplify.

It will be interesting to show the corrosion rate in concrete versus in CW with a phase of about 250 hr. If it yields a linear line, it will hint that the CW method can be used to predict the corrosion rate of this alloy in concrete. (let's remember that the concrete pH is much higher, and may be reduced in the vicinity of the alloy due to the corrosion products)

Thank you for this interesting idea. We have now added a graph showing the time span from 1250 to 1500 hours of exposure time. There is no significant difference between the decrease in corrosion rates of both media.

The method to create a scale for the x-ray images is to measure the distance of the bar width in pixels. Since the bar thickness change is known and much lower than its thickness, you will get a scale of pixels per millimeter, which can be used to draw a scale on the image.

Thank you for this valuable advice. A scale was added to the figure. It was mentioned that the exact orientation of the sample in the concrete cannot be determined, but the scale could be used to determine an approximate thickness of the (assumed) oxide layer.

The discussion regarding the decrease in corrosion rate, including possibilities that can be eliminated, should be in the text, not only in your answer. Nevertheless, since your research cannot confirm any mechanism, the question should remain open. Notice that the alkaline corrosion process of aluminum reduces the pH and consumes water. As a consequence, a pH and a water potential gradient may occur. As the process continues, the thickness of the affected zone may increase and cause the logarithmic decay. However, your discussion should state everything explicitly. Writing only "the oxide layer" is not good enough. You can adopt J.R. Galvelle model for pH polarization in pit corrosion, with the necessary changes.

Thank you for the note. We mentioned other processes now and explained why we assume that the oxide layer is causing the decrease in corrosion.

Please choose one of the options: 1) draw the 95% confidence interval envelope on the graphs; or 2) when you write a predicted number, add the limits of the confidence interval (and write it explicitly).

Thank you for providing these options. A 95 % confidence interval envelope has been added to the graph.

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

It seems you misinterpreted one of my comments. The Idea is to plot corrosion rate in concrete vs. corrosion rate in CW (Fig. 5), so if you get a linear curve, you can support your statement in lines 249-252 and 315-317.

It is unclear how the 95% confidence interval envelope does not include the measured points. The meaning of 95% confidence interval is that the probability of falling out of it is 5%. There are more than 5% of the results out of the envelope. That means something went wrong. 

A logarithmic scale is appropriate for Fig. 4 as it is for Fig. 3

Lines 315-317 need rephrasing

As aluminum and its alloys are clearly unsuitable for use with concrete, it is not clear if there is a practical reason to verify if CW can be used to asses the corrosion rate in concrete.

Author Response

It seems you misinterpreted one of my comments. The Idea is to plot corrosion rate in concrete vs. corrosion rate in CW (Fig. 5), so if you get a linear curve, you can support your statement in lines 249-252 and 315-317.

 

Thank you for the comment. The graph has been reworked as suggested and a linear dependency was determined. The statements have been rephrased.

 

It is unclear how the 95% confidence interval envelope does not include the measured points. The meaning of 95% confidence interval is that the probability of falling out of it is 5%. There are more than 5% of the results out of the envelope. That means something went wrong. 

 

Thank you for the comment. The asymmetry of the confidence interval looks compressed on a logarithmic scale which causes the missing of more than 5 % of the data points since we are fitting heteroscedastic data due to the high uncertainties in the beginning caused by sample preparation time.

 

A logarithmic scale is appropriate for Fig. 4 as it is for Fig. 3

Thank you for your valuable advice. We discussed to use a linear scale because values in corrosion rates are only going up to 1000 and a logarithmic scale would stretch the data in an unbeneficial way since it would not highlight the delayed corrosion anymore.

 

Lines 315-317 need rephrasing

Thank you for your advice. Line 315-317 have been rephrased as suggested.

 

As aluminum and its alloys are clearly unsuitable for use with concrete, it is not clear if there is a practical reason to verify if CW can be used to asses the corrosion rate in concrete.

Thank you for your comment. The Swedish repository plan includes encapsulation in concrete before disposal. This has not been clearly mentioned in the text. It has been rephrased in Line 47. Concrete will also be used for structural material and will therefore be abundant in repositories. Therefore, in different timescales (depending on degradation barriers), the material will always come into contact with CW,  Results of this project will be used in safety assessments.