Analysis of Compositional Modification of Commercial Aluminum Bronzes to Obtain Functional Shape Memory Properties

In copper-based shape memory alloys (SMAs), some exceptional phenomena, such as the shape memory effect (SME) or superelasticity (SE), are observable. However, commercial aluminum bronzes, Cu3Al-based alloys, do not present these functional properties (SME and/or SE) in their original state. Thus, since one of the main copper-based SMA systems is the Cu-Al-Ni alloy, this paper aims to analyze the modification of these commercial aluminum bronzes to SMA by the addition small amounts of Cu, Al and/or Ni. These modified bronzes were reprocessed by induction melting and injected by centrifugation into a ceramic coating mold. The modifications were made to determine the nominal composition for a Cu-13,0Al-4,0Ni (%wt) SMA. The effectiveness of the modifications was verified by differential scanning calorimetry (DSC) thermal analysis. All modified Cu-Al-Ni bronzes presented DSC peaks of the thermoelastic martensitic phase transformation, showing that SMA behavior was achieved, while the non-modified bronzes revealed no transformation. These results were supported by Vickers hardness (HV), X-ray diffraction (XRD), semi quantitative composition by EDS analysis and optical microscopy.


Introduction
Shape memory alloys (SMAs) are special metallic materials that present the functional phenomena of the shape memory effect (SME) and superelasticity (SE); thus, they are classified as active or smart metals.Both phenomena occur due to the thermoelastic martensitic phase transformation.The SMA undergoes a reversible martensitic transformation, allowing deformation through a twinning mechanism when temperatures are below the final martensite transformation limit (M f ) 1 .For the SME, this deformation is reversed when the detwinned structure (martensite) transforms to a parent phase above the initial austenite transformation temperature (A s ) through heating.Cu-Al binary alloys show a martensitic transformation with high transformation temperatures.However, these temperatures can be adjusted by adding a third element to the alloy.The addition of up to 4 %wt.nickel reduces the transformation temperatures of Cu-Al alloys 1,2 .The nickel acts to retard the diffusion of aluminum in annealed Cu-Al alloys, eliminating the stable γ 2 phase that does not undergo the martensitic transformation.It is also known that the transformation temperatures of these alloys are sensitive to their chemical composition [1][2][3][4] .For the Cu-XAl-4Ni alloys (% wt), it has been observed that the transformation temperatures abruptly decrease with a slight increase in Al content (X) 1,5 ; depending on the Al content, different martensites can be formed, denoted as γ 1 ' or β 1 ' 3,6 .
Copper-based SMA have steadily attracted the attention of researchers during recent decades due to their low cost and ease of manufacturing when compared with TiNi -based alloys 1,4 .From this knowledge and noting that commercial aluminum bronzes are basically Cu 3 Al -based alloys that have a number of other trace elements (Ni, Fe, Mn and others) in their composition 7 , the aim of this study was to investigate whether these bronzes can present a phase transformation in the as supplied state after quenching.If not, thus study sought to assess the possibility of minimally altering its chemical composition for the thermoelastic martensitic transformation to occur, which is the origin of the functional SME phenomena.
It is important to note that an attempt to modify commercial aluminum bronzes to obtain properties of the thermoelastic phase transformation and the shape memory effect has not been found in the international literature.

Materials
The materials used in this study were three commercial aluminum bronze alloys supplied by Metals Luna Company (São Paulo, Brazil): BRONZE AL NI CA 9581 C95800; BRONZE SAE 630 C63000 and BRONZE SAE 68D C95500.Due to the lack of identification, these bronzes were named A, B and C, respectively.

Methods
Initially, the supplied aluminum bronzes were subjected to a heat treatment at 850°C for 30 minutes, followed by quenching in water at approximately 30°C.After this treatment, the aluminum bronzes were cut into samples of a mass between 20 and 100 mg to be investigated in the DSC thermal analysis using a Q20 model calorimeter (TA Instruments).As these first DSC analyzes did not reveal the occurrence of phase transformation peaks, the samples of each kind of bronze were subjected to a chemical analysis using energy dispersive spectroscopy (EDS) in a scanning electron microscope (SEM) VEGA3 model (Tescan) equipped with an EDS detector from Oxford Instruments (Oxford Instruments, X-act model).The magnification used was set at 2000x, and a total of 5 EDS measurements were taken at 5 random points for each sample.The compositions of the aluminum bronzes (as supplied) by weight (%wt.) from the EDS are summarized in Table 1.
According to this preliminary composition analysis for each bronze alloy, the new set of modified bronzes for an SMA nominal composition of Cu-13,0Al-4,0Ni (%wt.) was established based on the literature data 1,5 .For this, considering that the total mass to produce the new Cu-Al-Ni SMA specimens was 40 g, a mass correction of the elements for each of the as received bronze alloys (A, B and C) was performed, as shown in Table 2.The complementary elements for the production of modified bronzes (Cu-Al-Ni SMA) were cut according to the calculated masses shown in Table 2.
The modified bronzes were produced using the Powercast 1700 machine, from EDG equipment and controls (São Paulo, Brazil), which uses the induction melting technique with injection by centrifugation into a ceramic mold.At the end of the melting and injection process, the ceramic mold is removed from the machine and rapidly cooled in water.Once cleaned and prepared, samples were subjected to a heat treatment at 850°C for 30 minutes (betatization) followed by quenching in room temperature water.A specific nomenclature was given to identify the new modified aluminum bronze alloys, which can be seen in Table 3.
All modified aluminum bronzes were first characterized by DSC thermal analysis using the same equipment aforementioned (calorimeter Q20 model from TA Instruments).The sample masses for these analyses were between 20 and 100 mg, and the temperature range was from 0°C to 300°C with a heating/ cooling rate of 20°C/min under a nitrogen atmosphere.The starting and final transformation temperatures, both during cooling and heating, were obtained by the tangent intersection method of the DSC peaks 1 .
Vickers hardness (HV) measurements were carried out using a FM-700 equipment from Future-Tech.All indentations were performed at room temperature with a load of 300 gf and an application time of 15 s.One sample of each bronze (both the as supplied and for modified) was used, and each sample was subjected to 10 indentations in random points.
For the identification of the phases present in the aluminum bronzes, the XRD technique was employed using a diffractometer XRD-6000 Model from Shimadzu.The Cu-K α radiation was used with a 2Ө interval from 20 to 90° and scanning steps of 0.02° with 3 s of accumulation.
The microstructures were observed by optical microscopy (OM).The etching solution used for these observations was 10 g FeCl 3 + 25 mL HCl + 100 mL H 2 O.The images were captured using an Olympus microscope, model BX51M.

Verification of phase transformation by Differential Scanning Calorimetry (DSC)
DSC analyses were performed for the aluminum bronze alloys under the following conditions: supplied after quenching (AT, BT and CT), modified with no heat treatment (AMB, BMB and CMB) and modified after quenching (AMT, BMT and CMT).
It can be observed from the DSC results shown in Figure 1 that all supplied bronzes showed no phase transformation, even after the heat treatment (quenching).On the other hand, all modified alloys in both conditions (no quenching and after quenching) provided phase transformation peaks.The occurrence of the thermoelastic martensitic phase transformation in all modified bronzes reveal a strong potential for the shape memory effect to be achieved.
Figure 1 also shows that the temperatures of the forward (cooling, M s and M f ) and reverse (heating, A s and A f ) transformations have been reduced for all modified bronze alloys after quenching, which is now denoted Cu-Al-Ni SMA.The forward (during cooling) and reverse (during heating) phase transformation temperatures were measured with the method of tangents in the transformation peaks 1 .Table 4 summarizes these phase transformation temperatures of the new Cu-Al-Ni SMA with no heat treatment and after quenching.It is verified that the transformation temperatures decrease with the increase of the aluminum content in the modified bronzes (Cu-Al-Ni SMA), according to the literature 1,5 .By checking Tables 4 and 5, it is noted that the M s temperature is reduced from 183.8°C in the CMT alloy (9.7 %wt.Al) to only 41.4°C in the AMT alloy (17.0 %wt.Al).According to Chang 5 , thermal transformation of the Cu-13,0Al-4,0Ni (%wt.)SMA occurs between 150°C and 210°C, with a M s temperature on the order of 181°C.
Ten thermal cycles were performed to verify the stability of the phase transformations observed in the new Cu-Al-Ni SMA manufactured in this study.Figure 2 shows the DSC curves for alloys A, B and C (modified and before and after quenching) during the thermal cycling.For the three alloys, well-defined peaks are again observed, confirming the forward (cooling) and reverse (heating) martensitic transformation.The transformation temperatures shifted at each cycle to higher temperatures for alloys A and B. It is well known that thermal cycling up to 200°C may promote not only modifications in the alloy structure and composition but also changes in its physical and mechanical properties 1,6 .Araujo et al. 8 detected an increase in the critical transformation temperatures associated with a change of γ 1 ' into β 1 ' martensite after a sufficiently long aging time at 200°C.In our case, and only for alloy C, after an increase during the first seven thermal cycles, the transformation temperatures shifted to lower values.The enthalpies of transformation also decreased, indicating that the phase transformation for this particular alloy tends to disappear with thermal cycling between 0°C and 300°C.This is caused by the fact that the modified alloy C presents the highest transformation temperatures A s and A f between 200°C and 250°C.This behavior was also observed for the Cu-Al-Nb SMA, where after the fifth thermal cycle, at temperatures higher than 250°C, the phase transformation tends to disappear completely.The low thermal stability observed in the alloys of the Cu-Al-Nb system is due to intense development of the diffusion processes at relatively high temperatures of the cycled heating 9 .This is probably the case for our modified alloy C (CMT).However, after disappearance after a few cycles to a maximum temperature of 300°C, the phase transformation can be recovered by a new heat treatement (quenching), so that the material (Cu-Al based SMA) can be reused for a few more thermal cycles.
The aforementioned variation in the transformation temperatures for alloys A, B and C in the modified and quenched condition can be best seen in Figure 3.An increase of the peak temperatures was observed for both the forward (M p ) and reverse transformation (A p ), as well as the stability of the thermal hysteresis behavior (H t = A p -M p ) during the cycling, except for alloy C beginning at the eighth cycle, as previously described.
The behavior of the forward (ΔH A-M ) and reverse (ΔH M-A ) transformation enthalpies (in J/g) can be seen in Figure 4.For the modified alloys AMT and BMT, a slight difference between the forward and reverse transformations was observed.Moreover, neither enthalpies are practically affected by the number of cycles.This is an apparent indication that the alloy thermodynamic states are not influenced by either the direction of transformation or the number of thermal cycles, as also reported by Pereira et al. 6 .Then, a better stability of transformation energies was verified for the modified alloys A and B and a gradual reduction for alloy C due to the thermal degradation associated with the diffusion processes Materials Research   at relatively high temperatures of cycled heating 9 .For all the modified alloys, it was found that these energies agree with the transformation enthalpies reported in the literature for Cu-Al-Ni SMA with 13.5 to 14.0 %wt.Al in the range of 5 to 10 J/g 1,5,6 .

Mechanical Behavior -Hardness HV
Figure 5 shows the hardness HV values for the aluminum bronze alloys in all studied conditions (as supplied, as supplied then quenched, as modified, and as modified then quenched).It was observed that for all as supplied bronzes, heat treatment (quenching) resulted in an increase in the hardness from 200 HV to 350 HV, in agreement with literature values for nickelaluminum bronzes 10 .On the other hand, it was found for the modified bronze alloys that the samples before quenching, by the melting process, presented a higher hardness value than the one for the alloys after quenching.This behavior can be explained by the occurrence of a larger amount of the twinned martensite phase after quenching (γ 1 ' and β 1 ' martensites).This level of hardness for the thermoelastic martensitic structure, from approximately 270 HV to 300 HV, is in agreement with values found earlier in the literature for the same Cu-Al-Ni SMA 11 .

X-Ray Diffraction (XRD)
Figure 6 shows XRD patterns obtained at room temperature for the as supplied (Fig. 6a) and modified (Fig. 6b) bronzes, both annealed and quenched.The identification of the XRD peaks was performed based on the literature values for Cu-Al-Ni SMA 3,12 .In Figure 6(a) for the three alloys, the XRD revealed the presence of the α phase (Cu 4 Al) with DO 3 ordering and other peaks that were not identifiable.The alloy CT also revealed peaks of the high temperature β phase (Cu 3 Al) ordered with a DO 3 structure.On the other hand, for the three Cu-Al-Ni SMA (Fig. 6b), the presence of two types of martensite phases were observed, β 1 ' (Orthorhombic -18R) and γ 1 ' (Monoclinic -2H), peaks of the high temperature β phase (Cu 3 Al) ordered with a DO 3 structure and the α phase (Cu 4 Al) with DO 3 ordering.For the AMT and BMT alloys, the presence of a solid phase rich in aluminum was still detected, γ 2 (Cu 9 Al 4 ), with DO 3 ordering, which is responsible for the brittleness of the Cu-Al-Ni SMA.The deformation of the two-phase domain is impractical due the fragility that the γ 2 phase produces, caused by its inconsistency with the matrix 1 .There is also evidence of another possible precipitate that these alloys may present: the AlNi, which has a B2 ordered type structure 12 .As the martensitic phase type β 1 ' is less stable than γ 1 ', the thermal cycling may cause the alloy that contains a higher volume of this phase to lose its SME properties.This occurs due to structural changes and dislocation arrangements at quenching sites that function as seeds for nucleation of the orthorhombic martensite phase β 1 ' 3 .

Energy Dispersive Spectroscopy (EDS)
The semi-quantitative evaluation of the chemical composition of the modified and quenched aluminum bronze alloys (AMT, BMT and CMT) was obtained by EDS microanalysis carried out in dots and lines.The measurements were statistically treated to obtain the average values, as summarized in Table 5.These values show that the alloys are not identical in composition, as they present different contents of Cu and Al.From these results, it is possible to understand the behavior of the phase transformation temperatures obtained by the DSC for each modified bronze, considering that with the increasing aluminum content, the transformation temperatures decrease 1,8 .This also may explain the occurrence of the γ 2 brittle phase in the alloys with a higher aluminum content (AMT e BMT), as well as its absence in the CMT alloy (Fig. 6b).Comparing the compositions summarized in Table 1 with those in Table 5, it can be seen that after the modification, while the Ni content values maintained a correlation to the designed alloy (nominal composition of Cu-13,0Al-4,0Ni SMA), the aluminum content values presented the opposite behavior.Additionally, it should be noted that the reprocessing with the investment casting technique alters the content of the residual elements (Fe, Mn, Si and Zn).The EDS results of the modified and quenched bronzes were statistically compared by a one-way ANOVA.P-values of less than α=0.05, as shown in Table 6, indicate significant differences between the content values of the main elements (Al, Cu and Ni).
It is important to note that these aluminum bronzes may have other residual elements, such as Fe and Si 7,10 .In our case, a significant Zn content was also detected (Table 1).In the case of bronze B, the analysis by EDS did not detect the presence of Si, perhaps because it is a semi quantitative technique.After modifying the bronzes by remelting with addition of new amounts of Cu, Al and/or Ni, the compositions undergo a complete rearrangement, including the residual elements Fe, Mn, Si and Zn, as indicated in Table 5.The Zn content is reduced and even disappears in the case of alloy B. This may occur by evaporation of Zn during the modification process which includes a further melting of the bronze.Si appearing in alloy B after modification may be associated with the limitation of the EDS technique on detecting it in the original bronze (alloy B, Table 1).

Optical Microscopy (OM)
Figure 7 shows the optical micrographs of the studied bronze alloys.The A, B and C alloys present large quantities of the α phase, which is characteristic of aluminum bronzes.This phase is a solid copper-based solution having a CFC structure and formed when the cooling time is sufficient for the occurrence of the phases, according to the phase diagram of the alloy 13 .It can also be observed for the AT alloy that  the heat treatment of quenching promoted a thin plate shape for the α phase.Whereas for the BT and CT alloys, the microstructures presented coarser grains in the α phase domains.According to Rodrigues 13 , the α phase is ductile, and its reduction tends to increase the overall hardness.This fact was also observed through the microhardness analysis in this study.For the other alloys, the presence of the α phase and discrete needles of the martensite β 1 ' (with structure 18R) was observed on the matrix of the austenite β (with the ordinate structure DO 3 ).Grain boundaries were observed for the quenched alloys, where possible acicular precipitates still appear scattered.According to Chang 5 , these precipitates can correspond to the γ 2 phase, which has a Cu 9 Al 4 structure.

Conclusions
The possibility of commercial aluminum bronzes to present the thermoelastic martensitic transformation and its origin in shape memory phenomena was investigated in this work.It was observed that a small compositional modification in these commercial materials can promote the occurrence of thermoelastic transformation, which was confirmed through a DSC thermal analysis.Modified bronze alloys, now classified as Cu-Al-Ni SMA, present M s temperatures in the range of 40°C to 190°C, depending on the Al content.The hardness HV results showed that the modified samples, before and after quenching, exhibit a lower hardness value than the as supplied bronzes (after quenching) due to the occurrence of the twinned martensite phases.The XRD results confirmed the presence of the twinned martensite phases, which is the phase that promotes the shape memory phenomena.Images from the optical microscopy allowed the observation of the studied alloys microstructure, confirming the presence of the martensite phases for the modified and quenched alloys (Cu-Al-Ni SMA).In conclusion, this work has shown that commercial bronze alloys can be modified to be used as SMA, and the BMT alloy presented the best results concerning the transformation temperatures (between 30°C and 180°C) and thermal stability of the martensitic thermoelastic transformation.

Figure 1 .
Figure 1.DSC thermal analysis of the as provided and modified bronze alloys (Cu-Al-Ni SMA).

Figure 2 .
Figure 2. Effects of thermal cycling on the Cu-Al-Ni SMA (modified and quenched).

Figure 3 .
Figure 3. Evolution of the peak temperatures of the transformations and hysteresis during thermal cycling of the Cu-Al-Ni alloys (modified and quenched).

Figure 4 .
Figure 4. Evolution of the transformation enthalpies during thermal cycling of the of the Cu-Al-Ni alloys (modified and quenched).

Figure 7 .
Figure 7. Microstructures of all the studied non-modified and modified bronze alloys (Cu-Al-Ni SMA).

Table 1 .
Chemical composition (%wt.) of the as received commercial aluminum bronzes obtained by EDS.

Table 2 .
Correction for the compositional aluminum bronzes A, B and C with a total mass of 40 g and the Cu-13.0Al-4.0NiSMA.

Table 3 .
Nomenclature used for bronze alloy modified during execution of the work.

Table 5 .
Chemical composition (%wt.)obtained by EDS for the modified and quenched aluminum bronzes.

Table 6 .
Statistical comparison of the EDS results by a one-way ANOVA of the modified and quenched aluminum bronzes (Cu-Al-Ni SMA).