Designing against phase and property heterogeneities in additively manufactured titanium alloys

Additive manufacturing (AM) creates digitally designed parts by successive addition of material. However, owing to intrinsic thermal cycling, metallic parts produced by AM almost inevitably suffer from spatially dependent heterogeneities in phases and mechanical properties, which may cause unpredictable service failures. Here, we demonstrate a synergistic alloy design approach to overcome this issue in titanium alloys manufactured by laser powder bed fusion. The key to our approach is in-situ alloying of Ti−6Al−4V (in weight per cent) with combined additions of pure titanium powders and iron oxide (Fe2O3) nanoparticles. This not only enables in-situ elimination of phase heterogeneity through diluting V concentration whilst introducing small amounts of Fe, but also compensates for the strength loss via oxygen solute strengthening. Our alloys achieve spatially uniform microstructures and mechanical properties which are superior to those of Ti−6Al−4V. This study may help to guide the design of other alloys, which not only overcomes the challenge inherent to the AM processes, but also takes advantage of the alloy design opportunities offered by AM.

In this manuscript, authors proposed a novel material designing strategy for additive manufacturing to achieve the phase and property uniformity in the AMed titanium alloy, using Ti-6Al-4V (Ti64) as an example. The critical and novel point in this strategy is to in-situ alloying of Ti64 with combined addition of pure titanium powers and iron oxide nanoparticle to tune the phase transformation kinetics during the additive manufacturing process. This strategy focusing on the uniformity of microstructure in the AMed titanium alloys is novel. In the past, the designing strategy of AMed titanium alloys mainly focused on the beta grain refinement (alloy design) or defect control (processing optimization). The current work on the other hand is focusing on the homogeneity of alpha+ beta microstructure in the AMed Ti-64 alloy. It is to avoid the alpha prime martensite phase formation during the AM process in the layers close to the top of the product, where limited thermal cycles occurred. Thus, the designing strategy is novel and the reported research in significantly beneficial to the understanding of processingmicrostructure-property in the AMed Ti64 alloy.
However, there are several questions authors are required to answer clearly: 1. The newly developed alloy is not Ti-64, but a new Ti-Al-V-Fe-O alloy. So why it is an important alloy to study?
Ti-64 alloy occupies the largest amount in the titanium market, and thus, significant amount of effort is spent on the low-cost manufacturing of this important Ti-64 alloy. However, in this work, different solutes (some is large amount) have been added into Ti-64 and thus significantly change the alloy composition. In the reported work, 25, 50 and 75wt% of CP-Ti was added into Ti-64 with 0.25-0.50wt% Fe2O3. Thus, the alloy manufactured is not Ti-6-4 anymore, but a new Ti-Al-V-Fe-O alloy. So the question is even if the microstructure produced in the newly developed alloy is full of alpha + beta microstructure, why it is an important alloy or why people need to study and manufacture this alloy?
Recently, different phase transformation mechanisms have been proposed in the field of titanium alloys to explain the formation of alpha phase. Whether or not partitioning is required to form alpha phase is being challenged: 1) Physical Review B 74, 134114 (2006). The concept of "bainitic alpha" was proposed in this work and it was claimed that "the growth of bainitic alpha plates is partitionless". 2) Acta Materialia 60 (2012) 6247-6256. The concept of "pseudo-spinodal decomposition" was proposed that the structure and composition change in the formation of alpha may not occur simultaneously.
If diffusion is not required to form alpha microstructure in the titanium alloys, is it still necessary to add the fast diffuser (like Fe in the current work) or to manipulate the partitioning of alloying element in phase decomposition, which is the "key to our approach" claimed in the manuscript?
Reviewer #2 (Remarks to the Author): This is an excellent contribution. The approach is novel, the methods and analysis is very well documented, the results on mechanical behavior quite interesting. My only recommendation for a minor modification is that the authors should point out also that while this approach is suitable for Ti6Al4V modified alloys for room temperature applications, it may not be suitable for creep applications to temperatures of 250 or 300C at which Ti6Al4V may be used, because Fe additions may lower creep resistance Reviewer #3 (Remarks to the Author): A new approach has been identified in this work to eliminate microstructural heterogeneity in Ti-6Al-4V, resulting from variations in thermal history during fabrication by laser powder bed fusion additive manufacturing, by modifying the alloy with cp-Ti and Fe2O3. The approach successfully eliminates heterogeneity and at the same time improves strength and ductility. The manuscript is well-written, but a few comments should be addressed before publication, as listed below: 1) It is explained that Fe addition favors the formation of beta phase owing to its beta stabilizing effect and higher diffusivity as compared to V, which rationalizes the addition of Fe2O3. However, the mechanism by which dilution of V through the addition of cp-Ti promotes beta phase formation is not clear. A follow-up question is, can addition of only Fe2O3, without any cp-Ti, eliminate the heterogeneity or not? This should be shown experimentally by printing Ti-6Al-4V + Fe2O3 alloy and performing the same microstructural characterization as done for other alloys. This result is also needed to support the authors' argument of synergistic effect of cp-Ti and Fe2O3 in eliminating the microstructural heterogeneity. 2) Provide the diffusivity values of V and Fe in alpha/martensite and beta phases at temperatures of interest.
3) A dedicated discussion on the sequence of phase transformation with and without the additives (cp-Ti and Fe2O3), possibly supported by a schematic, will give better insights into the mechanisms. It will help clarify questions such as: does the beta phase forms by martensitic decomposition or is it the retained beta from solidification; if beta forms due to martensitic decomposition, what is the contribution of accumulated heat in the sample, will the outcome change if the sample temperature of 500 °C assumed in the thermodynamic model is actually lower? 4) Although the SEM micrographs are able to differentiate between martensite and alpha + beta microstructures as a function of build height, these results can be supported by additional characterization using XRD, TEM, or both. 5) The best resolution for x-ray CT was 2 um as mentioned, but SEM micrographs in supplementary Fig.  10 show many smaller pores. These smaller pores should be characterized as a function of the build height to strengthen the argument that the variation in ductility with change in build height is not due to porosity. 6) For every figure with tensile curves, fractographs, or x-ray CT data, please mention the location in the as-build part from where the characterized sample are extracted. 7) What liquid is used to suspend Fe2O3 particles? 8) Explain the meander scanning strategy in more detail. 9) Corrosion is included in the methods section and supplementary information, but not referred to in the main text. 10) There are many typos in the manuscript, please proofread carefully.

Designing against phase and property heterogeneities in additively manufactured titanium alloys
We would like to thank the editor for giving us the opportunity to revise the manuscript and thank all reviewers for their careful review and valuable comments. We have revised our manuscript accordingly and have highlighted all revisions in red in the revised manuscript.
We have also added Prof. Gang Sha as one of the corresponding authors, due to his very important contribution to this work and the APT characterization. Please find below our point-by-point response to reviewer's comments, in which the reviewers' comments are in black and our responses are in blue. Additionally, the red texts indicate those added to the revised manuscript.

Reviewer #1:
In this manuscript, authors proposed a novel material designing strategy for additive manufacturing to achieve the phase and property uniformity in the AMed titanium alloy, using Ti-6Al-4V (Ti64) as an example. The critical and novel point in this strategy is to insitu alloying of Ti64 with combined addition of pure titanium powers and iron oxide nanoparticle to tune the phase transformation kinetics during the additive manufacturing process. This strategy focusing on the uniformity of microstructure in the AMed titanium alloys is novel. In the past, the designing strategy of AMed titanium alloys mainly focused on the beta grain refinement (alloy design) or defect control (processing optimization). The current work on the other hand is focusing on the homogeneity of alpha+ beta microstructure in the AMed Ti-64 alloy. It is to avoid the alpha prime martensite phase formation during the 2/25 AM process in the layers close to the top of the product, where limited thermal cycles occurred. Thus, the designing strategy is novel and the reported research in significantly beneficial to the understanding of processing-microstructure-property in the AMed Ti64 alloy.
However, there are several questions authors are required to answer clearly:

Response:
We really appreciate the reviewer for the careful review and positive comments. Ti-64 alloy occupies the largest amount in the titanium market, and thus, significant amount of effort is spent on the low-cost manufacturing of this important Ti-64 alloy. However, in this work, different solutes (some is large amount) have been added into Ti-64 and thus significantly change the alloy composition. In the reported work, 25, 50 and 75wt% of CP-Ti was added into Ti-64 with 0.25-0.50wt% Fe2O3. Thus, the alloy manufactured is not Ti-6-4 anymore, but a new Ti-Al-V-Fe-O alloy. So the question is even if the microstructure produced in the newly developed alloy is full of alpha + beta microstructure, why it is an important alloy or why people need to study and manufacture this alloy?

Response:
We thank the reviewer for the in-depth discussion. We agree with the reviewer that titanium alloys produced by L-PBF in this work are not Ti−6Al−4V anymore. As the "workhorse" alloy in the titanium industry, Ti−6Al−4V was designed and optimized for the conventional manufacturing routes. Despite being well fabricated by AM with a very high density, Ti−6Al−4V does not mechanically perform to the best of its capacities. For example, Therefore, the driving force behind this work is to design new alloys specifically for the AM process with exceptional mechanical performance. The newly developed titanium alloys can be candidate materials for applications where titanium alloys with unform mechanical properties are required.
To address this comment, we have revised the Conclusion (the last paragraph) of the main text on Page 11 to highlight: (1) the potential applications of the newly developed alloys, and (2), to a broader perspective, the influence of the alloy design strategy, as shown below.
"…We expect that the newly developed titanium alloys could be candidate materials for applications where titanium alloys with uniform mechanical properties are required. This requires a comprehensive evaluation of other mechanical properties (such as fatigue 4/25 properties and creep resistance) and corrosion resistance (Supplementary Note 2).
Furthermore, unlike previous studies which have mainly focused on grain refinement (through alloy design) and/or defect control (via processing optimization), our work demonstrates that addressing the phase heterogeneity is of equal, if not greater, importance to achieve the desired uniform mechanical properties. Since the phase heterogeneity due to the solid-state thermal cycling has been reported in a wide variety of metallic materials fabricated by different AM technologies 12,46-49 , we believe that our design strategy may help the development of other metallic alloys specifically for AM with uniform mechanical properties." Response: We thank the reviewer for the in-depth discussion. In this work, we utilized a FEI Scios Dual Beam SEM equipped with a concentric backscattered detector (CBS) to 5/25 differentiate between α′ and α under a very low accelerating voltage (3 kV) and a small working distance (5.5 mm). Unlike conventional backscattered electrons (BSE) imaging, the contrast in different phases through this technology stems partially from electron channelling [1]. It is capable of detecting twins within α′ (which are the fingerprint of α′ martensite [2]) and has been recently used by researchers to distinguish α′ and α [1, 3−5]. The main advantage of this technology over TEM is that the images can be obtained on bulk samples rather than thin foils. In addition, such SEM imaging allows for characterization of relatively large area on the sample. This is very important for this work to examine whether the microstructure is indeed homogeneous or not throughout the samples.
We didn't perform APT characterization of Ti−6Al−4V, because the previous work by on additively manufactured Ti−6Al−4V by L-PBF and EB-PBF, respectively.
The kinetics of element partitioning is also captured by the DICTRA simulation in our work.
We have plotted the composition profiles of Al and O across the α′/β interface at various times. It is evident that Al and O are accumulated within α′ phase close to the α′/β interface.
In the meantime, Fe and V diffuse out from α′ phase to β phase. This dynamic element partitioning is in line with the conclusion drawn from the APT characterization by Haubrich et al. [1].
To address the concerns raised, we have also provided the DICTRA simulation of composition profiles of Al and O in Supplementary Fig. 14  "…On the other hand, it is found that the α stabilizers Al and O are accumulating in the α′ phase as the cooling process proceeds (Supplementary Fig. 14a,b). The dynamic element 3. The claimed application of the approach to the beta titanium alloy needs further explanation.
Authors claim the proposed approach can be applied in the beta titanium alloys to trigger the omega phase and alpha phases in the building direction. However, Fe is a strong beta phase stabilizer and oxygen will impede the omega phase formation as well. Thus, I don't think the addition of Fe2O3 will promote the omega phase and alpha phase during the AM process in the building direction. So authors are required to introduce more details how the proposed strategy can be used in other titanium alloys.

Response:
We thank the reviewer for this valuable comment. In fact, we took metastable β titanium alloy as an example to highlight "The phase heterogeneity due to the solid-state thermal cycling has been reported in a wide variety of metallic materials fabricated by different AM technologies." Currently, we have been working on additively manufactured 9/25 metastable β titanium alloys and found that solid-state thermal cycling can also produce spatially dependent microstructures and hence undesirable non-uniform mechanical properties. We agree that the addition of Fe2O3 cannot address the phase heterogeneity in metastable β titanium alloys.
In this paragraph, we stated that "we believe that our approach is not restricted to the selected titanium alloy presented here and could be applied to other metallic alloys" in the initially submitted manuscript. It means that our alloy design strategy that aims at eliminating phase heterogeneity may help the design of other alloys which suffer from the same issue.
The previous studies on achieving uniform mechanical properties have mainly focused on grain refinement and/or defect control. In fact, addressing the phase heterogeneity is of equal, if not greater, importance to grain refinement or defect control to eliminate the property heterogeneity.
To address the concerns raised, we have revised the Conclusion (the last paragraph) of the main text. Please see our response to Comment 1.
4. The addition of Fe, the claimed key to the proposed approach, needs further discussion.
Recently, different phase transformation mechanisms have been proposed in the field of titanium alloys to explain the formation of alpha phase. Whether or not partitioning is required to form alpha phase is being challenged: 1) Physical Review B 74, 134114 (2006). The concept of "bainitic alpha" was proposed in this work and it was claimed that "the growth of bainitic alpha plates is partitionless". 10/25 2) Acta Materialia 60 (2012) 6247-6256. The concept of "pseudo-spinodal decomposition" was proposed that the structure and composition change in the formation of alpha may not occur simultaneously.
If diffusion is not required to form alpha microstructure in the titanium alloys, is it still necessary to add the fast diffuser (like Fe in the current work) or to manipulate the partitioning of alloying element in phase decomposition, which is the "key to our approach" claimed in the manuscript?

Response:
We thank the reviewer for the helpful comment and the interesting references. We have read the above references carefully and still believe that elemental partitioning is critical to promote the formation of lamellar (α+β) microstructure in our study. As mentioned in our response to Comment 2, the addition of Fe -which is a fast diffuser and a β stabilizer -is essential to promote the formation of β phase and hence the resulting lamellar (α+β) microstructures. This is supported by previous studies on AM of Ti−6Al−4V [1][2][3][4]. For example, in the APT work by Haubrich et al. [1], it is found that increasing the volume energy density from 77 to 145 J/mm 3 results in stronger intrinsic heat treatment effect, which leads to significant elemental partitioning of Fe, V, Al and O. Additional heat treatment of the part fabricated at 145 J/mm 3 results in further elemental partitioning. Previous work on Ti−6Al−4V inspires us to find a pathway to address the phase heterogeneity through alloy design.

Reviewer #2 (Remarks to the Author):
This is an excellent contribution. The approach is novel, the methods and analysis is very well documented, the results on mechanical behavior quite interesting. My only recommendation for a minor modification is that the authors should point out also that while this approach is suitable for Ti6Al4V modified alloys for room temperature applications, it may not be suitable for creep applications to temperatures of 250 or 300C at which Ti6Al4V may be used, because Fe additions may lower creep resistance.

Response:
We really appreciate the reviewer for the positive comments and constructive suggestions. The creep resistance of titanium alloys depends strongly on the volume fractions of α and β phases. In general, α phase shows superior creep resistance to β phase, due to the relatively limited ability for atoms to diffuse and HCP crystal structure [1]. The creep resistance of titanium alloys often deteriorates with increasing volume fraction of β phase. In this work, titanium alloys were developed based on Ti−6Al−4V through combined additions of CP−Ti and Fe2O3. As can be found in Supplementary Fig. 20  14/25

Reviewer #3 (Remarks to the Author):
A new approach has been identified in this work to eliminate microstructural heterogeneity in Ti-6Al-4V, resulting from variations in thermal history during fabrication by laser powder bed fusion additive manufacturing, by modifying the alloy with cp-Ti and Fe2O3. The approach successfully eliminates heterogeneity and at the same time improves strength and ductility. The manuscript is well-written, but a few comments should be addressed before publication, as listed below: Response: We greatly thank the reviewer for the careful review and positive comments.
1) It is explained that Fe addition favors the formation of beta phase owing to its beta stabilizing effect and higher diffusivity as compared to V, which rationalizes the addition of

Response:
We thank the reviewer for pointing out this important omission. We fabricated (Ti−6Al−4V + Fe2O3) alloys according to the research plan. However, we did not include the associated result in the initially submitted manuscript, because phase homogeneity cannot be achieved through the sole addition of Fe2O3 and we thought that it was not directly related to our alloy design strategy (that is, the combined addition approach). This comment makes us 15/25 realize that it is very important to include this result to demonstrate the synergistic effect of CP−Ti and Fe2O3.
To address this comment, we have provided the microstructural characterization of Ti−6Al−4V + 0.25 wt % Fe2O3 in Supplementary Fig. 21  " Supplementary Fig. 21 shows the microstructural analysis of (Ti−6Al−4V + 0.25 wt % Fe2O3) along the building direction. Overall, the addition of 0.25 wt % Fe2O3 to Ti−6Al−4V results in a significant reduction in the phase width throughout the part compared with Ti−6Al−4V (Supplementary Fig. 1a,b). It is apparent that the internally twinned martensite "It is evident that Fe shows a much stronger partitioning tendency in the β phase than V ( Supplementary Fig. 12 c-e), due to its significantly high diffusivity (Supplementary Fig. 13). 4) Although the SEM micrographs are able to differentiate between martensite and alpha + beta microstructures as a function of build height, these results can be supported by additional characterization using XRD, TEM, or both.

Response:
We thank the reviewer for the valuable suggestions. We have provided XRD and TEM characterization to support the phase analysis by SEM. As shown in Supplementary Fig.   2, the XRD result of Ti−6Al−4V samples shows the phase heterogeneity along the building direction and is consistent with SEM analysis (Fig. 1b, Supplementary Figs. 1a and 1b). The 19/25 XRD result is mentioned in the main text of the revised manuscript on Page 3, as shown below.
"…Such a graded phase distribution is also confirmed by scanning electron microscope (SEM) (Fig. 1b, Supplementary Figs. 1a and 1b)  We have also added the experimental procedure on XRD in Methods of the revised manuscript on Page 24, as shown below.
"X-ray diffraction 20/25 X-ray diffraction (XRD) analysis was conducted on a D8 ADVANCE X-ray diffractometer (Bruker, Germany) with Cu radiation source operated at 40 kV and 40 mA with a step size of 0.02°." Besides, we have provided the TEM images of the selected newly developed alloy, as shown in Supplementary Fig. 15. This figure is mentioned in the main text of the revised manuscript on Page 9, as shown below.
The STEM-EDS images clearly reveal that the β phase is enriched in both Fe and V while it is depleted in Al (Supplementary Fig. 15c)." The experimental information on TEM sample preparation and TEM characterization is also provided in Methods of the revised manuscript on Page 25. 5) The best resolution for x-ray CT was 2 um as mentioned, but SEM micrographs in supplementary Fig. 10 show many smaller pores. These smaller pores should be characterized as a function of the build height to strengthen the argument that the variation in ductility with change in build height is not due to porosity.

Response:
We thank the reviewer for pointing out this important issue. The small "pores" in curves (like what we did in Fig. 1c, that is, H1, H2, H3…), because this will make the figures look somewhat messy. However, we have provided this information in the text and other figures.