Corrosion–microstructure interrelations in new low-lead and lead-free brass alloys

ABSTRACT Stress-relieving heat treatment has been reported to deplete the corrosion resistance of new low-lead and lead-free brass alloys. How the heat treatment, processing and material composition relates to the microstructure and corrosion performance is not well understood. The present study aims to fill this knowledge gap by mapping stress-relieving annealing conditions and different standardised compositions to their respective microstructures and dezincification performance. It was found that loss of corrosion resistance was only the most severe for alloys with higher aluminium and iron content. These alloys displayed significant precipitation of intermetallic aluminium arsenide particles on grain boundaries, twins and lead particles, as well as the formation of β-phase along grain boundaries. This paper is part of a Thematic Issue on Copper and its Alloys.

Furthermore, investigations have reported that heat treatment at certain annealing temperatures, generally in the range 300-400°C [14], results in the arsenic forming intermetallic arsenide particles with other elements in the alloy, including aluminium and iron [1,17,18], which are added to enhance castability and grain refining [1,11,18]. The formation of arsenide particles depletes the active arsenic content, so deteriorating the dezincification protection of the α-phase and sensitising it to dezincification attack [17][18][19]. This transformation is a diffusion controlled process [1,3,4] and it is often observed that intermetallic arsenide particles precipitate adjacent to grain boundaries due to the higher lattice disorder there compared to the grain interiors. Thus, grain boundaries are subject to dezincification attack, also referred to as grain boundary attack or intergranular attack [13-1517-20]. With careful heat treatment, precipitation of both β-phase and arsenide particles can be avoided [1]. Brasses with arsenic successfully protecting the material from dezincification are appropriately referred to as dezincification resistant brass alloys [13].
Lead has traditionally been added to brass in contents of a few per cent to improve its hot workability and machinability in many types of water fittings [2,21]. However, it has been found that lead leaches into the drinking water at too high concentrations, thus promoting several countries and institutes, including the 4 Member State Joint Committee, to propose more strict regulations regarding the maximum allowed amount of lead in the material composition . Since lead has been a crucial element for enhanced machinability [1,2,21], a paradigm shift is now taking place within the brass manufacturing industry in terms of compositional and material design to adjust to the new regulations.
Among some of the new commercially available low-lead and lead-free brass alloys, aluminium content has been increased and the copper/zinc ratio has changed. How this affects the microstructure and the  dezincification properties is not well understood. One reported issue with the new alloys is that their corrosion protection is depleted during stress-relieving heat treatment, which is performed after all cold deformation treatments, including machining [23]. For manufacturers of water fittings and plumbings, minimising residual stresses is crucial for avoiding stress corrosion cracking [1,3,5,14]; thus, this is problematic since manufacturing these products requires cold deformation and machining.
To enable the industry to make more informed decisions in choosing annealing conditions for the new lowlead and lead-free brass alloys, the present study aims to better understand the influence of stress-relieving heat treatment on the dezincification resistance as well as the microstructure of new brass alloys. To accomplish this, three different new low-lead and lead-free brass alloys have been heat-treated using different annealing conditions, and their resulting microstructures are mapped in relation to their dezincification performance. The experimental results are also related to thermodynamic simulations.

Material
The samples used in this study were obtained from three extruded bars produced by Nordic Brass Gusum AB. Each bar represented a different alloy, referred to as CW511L, CW625N and CW626N, respectively. Their chemical compositions are given in Table 1.
Owing to the geometrical differences of the three bars, the composition of CW626 was measured using optical emission spectroscopy (OES) while X-ray fluorescence (XRF) was used for CW511 and CW625. The As content was further specified using inductively coupled plasma optical emission spectrometry (ICP-OES).
CW511L is commercially considered to be lead-free since it contains less than 0.3% lead. The only active elemental addition beyond copper, zinc and lead is arsenic. The remaining elements are considered as impurities which arise from recycled scrap. In addition, the chemical analysis of the CW511L revealed the lowest Cu/Zn ratio of the three alloys: ∼ 1.78.
CW625N and CW626N are considered as low-lead alloys since they contain 1.2-1.3% lead. In addition, they have been actively alloyed with additional aluminium and contained slightly higher iron concentrations as compared to the CW511L alloy. The main differences between CW625N and CW626N were the higher Cu/Zn ratio in the latter, as illustrated in Table 1.
It should be noted that the amount of aluminium in CW626N is 0.1 percentage points below the standard value of this alloy; however, this will not affect the overall analysis of this study.

Method
Samples cut from the extruded bars were heat-treated in muffle furnaces in two steps. First, they were preheated at 550°C for 2 h in order to minimise the volume fraction of β-phase and to ensure a common thermodynamic starting point. Thereafter, the samples were annealed at either 250, 300, 350 or 400°C for 2, 10, 100 or 1000 h. The temperatures in both heat treatment steps were monitored using two thermocouples; one measured the ambient temperature and the other the temperature of the largest sample in the furnace. After each heat treatment, the samples were cooled in air to 25°C. Samples without any heat treatment or pre-heating were kept as references Test samples were cut out into 10 × 10 × 10 mm cubes from each heat treatment, and mounted in phenolic resin so that the exposed surface was perpendicular to the extrusion direction of the original bar. The exposed surface was ground to 600 mesh size using wet abrasive SiO 2 paper. Dezincification testing was performed in accordance with the Swedish standard SS-EN ISO 6509-1:2014 'Corrosion of metals and alloys -Determination of dezincification resistance of copper alloys with zinc' of the Swedish Standard Institute [24]. According to this, the mounted samples were submerged in aqueous copper (II) chloride solution (12.7 g CuCl 2 × 2 H 2 O/1000 ml H 2 O) at 75°C for 24 h ± 30 min in a beaker sealed in plastic foil. The volume of solution was 250 ml per 100 mm 2 exposed brass area. In addition, the test surfaces were orientated vertically and > 15 mm above the bottom of the beaker. By examining the resulting copper layer on the brass surface, the dezincification resistance of the material could be estimated.
To examine the depth of the copper layer, the exposed surface was sectioned parallel to the extrusion direction after exposure. The cross-section was ground down to fine 4000 mesh size using wet abrasive SiO 2 paper, followed by mechanical polishing using cloths with 3, 1 and 0.25 μm in diamond suspension successively. The depth of the dezincification attack (average depth and maximum depth) was examined in the optical microscope in accordance with SS-EN ISO 6509-1:2014 [24].
The volume fraction of β-phase precipitated in the samples was also measured on the plane parallel to the extrusion direction in the optical microscope. The examined surface had been etched in Klemm's solution for 15 seconds, giving the β-phase a distinct yellow colour. The area fraction of yellow pixels could thus be calculated using the software ImageJ on three representative pictures of the microstructure. Only the samples heat-treated for 1000 h and the samples with no heat treatment were investigated.
The presence of intermetallic arsenide particles in samples was investigated using scanning electron microscopy with energy-dispersive X-ray spectroscopy(EDS) and electron backscatter diffraction (EBSD) detectors. Only a few selected samples were investigated. Using the post-processing software Tango by Oxford Instruments, the EBSD results were overlapped with the EDS results. This made it possible to investigate where sites for nucleation of arsenide particles were located in the microstructure.
Thermodynamic calculations were performed in Thermo-Calc, utilising a database for brass developed by Swerea KIMAB. The calculated property diagram would display the simulated mole fraction of each stable phase at each heat treatment temperature, under the assumption that the system has reached equilibrium. This was done in order to assist the analysis of the microstructures.

Influence of annealing on dezincification resistance
The corrosion resistance was mapped in terms of annealing temperature and duration for the three alloys in order to relate the resulting microstructural precipitates to dezincification resistance. As represented in Figure 1, CW511L maintained high dezincification resistance for all the investigated annealing conditions. Analysis of this microstructure in optical and scanning electron microscopes revealed that neither β-phase nor arsenide particles had precipitated in the annealed CW511L samples. This means that the dezincification inhibitor arsenic remained in solid solution within the α-phase thus protecting it. Furthermore, the dezincification susceptible β-phase never precipitated from any heat treatment. Thus, it may be expected that CW511L should display a high dezincification resistance.
When comparing CW625N and CW626N in Figures 2 and 3, it is noted that high corrosion resistance was maintained for all annealing conditions at 250°C, as well as the shortest duration (2 h) at 300, 350 and 400°C. However, both these alloys displayed clear loss of corrosion resistance when annealed for an extended duration at 300-400°C.

Influence of annealing on microstructure
Of the three alloys, CW511L showed negligible dezincification depth for all investigated samples exposed to CuCl 2 solution, as represented in Figure 1. As observed in the chemical analysis in Table 1, the main difference in composition in CW511L compared to CW625N and CW626N is its significantly lower aluminium and iron contents. Since the heat treatment, as well as the corrosion testing and sample preparation of CW511L samples, was identical to CW625N and CW626N, which displayed considerably lower dezincification resistance, it is probable that the difference in dezincification resistance is related to the higher aluminium or iron contents in the latter two alloys, which promote the precipitation of detrimental phases. EDS analysis in the SEM supports this; emitted As signals frequently overlapped with Al signals in CW625N and CW626N, as illustrated in Figure 4, indicating that there are traces of the intermetallic particles arsenic particles in the form of aluminium arsenides, AlAs. These reportedly are related to dezincification and grain boundary attack      [1,13,14,19,20]. Hence, it is possible that the higher aluminium content contributes to more frequent precipitation of intermetallic AlAs particles.
Based on the work of Wessman et al. [14], Olivier et al. [9] and Claesson and Rod [18], it was expected that arsenic would form intermetallic particles with iron. However, as illustrated in Figure 5, in the EDS analysis, it was observed that emitted iron signals did not overlap with arsenic. Instead, it overlapped in conjunction with other elements, such as chromium, silicon and phosphorus. This suggests that the iron did not form particles with arsenic, and so did not contribute to the depletion of dezincification resistance to the same degree as aluminium. The reason for this could be that the iron content was too low to precipitate as iron arsenide particles and/or that aluminium has a higher affinity for arsenic compared to iron.
By complementing the EDS signals with the EBSD image as displayed in Figure 6, it was observed that both the iron-and arsenic-enriched areas were visible along crystallographic defects, including grain boundaries, twins and adjacent to lead particles. This lends support to the aforementioned mechanism of dezincification by grain boundary attack [13,20] since it indicates that the grain boundaries have lost dezincification resistance due to depletion of arsenic in solid solution with the αphase. The results exemplified in Figure 6 also indicate that in addition to grain boundaries, the presence other defects such as twins or surfaces of undissolved particles (Pb in this instance) can provide nucleation sites for arsenic particles to form and thus result in loss of dezincification resistance.
With regard to β-phase, in CW625N and CW626N samples annealed at 250 and 400°C, no traces of βphase was observed in the etched microstructure using optical microscopy. It was only observable within samples annealed at 300°C, and to a lesser extent 350°C, for both CW625N and CW626N as illustrated in Figures 2  and 3. Since no β-phase was observed in the reference samples, it is plausible that this observed β-phase was caused as a result of the stress-relieving heat treatment.
For CW625N and CW626N samples annealed at higher temperatures (350-400°C), the area fraction of β-phase decreased, yet the corrosion resistance continued to decline while the relative frequency of As particles increased. This strongly supports the view that it is the formation of As particles that are predominantly responsible for the decreased corrosion resistance at higher annealing temperatures.
At 300-350°C, both β-phase and arsenic particles were observed in the microstructure of annealed samples and it is thus not clear the extent to which each phase contributes to the loss of dezincification resistance. However, since the CW625N annealed at 300°C for 1000 h resulted in a large deterioration compared to the corresponding CW626N sample, it is possible that the slightly larger area fraction of β-phase in CW625N contributed to the larger loss of dezincification resistance.
No β-phase was observed in CW511L. However, examination of the thermodynamic properties of βphase in Thermo-Calc in Figure 7 implies that β-phase should be stable enough to precipitate in a homogeneous steady state in this alloy system, even to a larger extent than in CW626N, which is not in agreement with the experimental results. There is a possibility that the database for calculations is inadequate and gives an incorrect contribution from the aluminium component. In terms of the zinc content alone, it would be expected that CW511L would be the most prone to form β-phase according to the simulated results. Alternatively, this might be an effect of kinetics whereby aluminium accelerates the formation rate of β-phase due to differences in diffusivity.
In addition, the simulations in Figure 7 show that the stability of β-phase decreases as temperature increases. For CW625, this can explain why samples with annealing temperature above 300°C experimentally display less β content. CW626 behaves similarly experimentally, yet its simulation indicates that β-phase should not form at 300°, thus supporting that the database for calculations is inadequate. Slow kinetics due to low annealing temperature could explain why no visible β-phase is formed at 250°C as the simulation indicates.
In terms of what might cause CW625N to differ from CW626N, calculations in Figure 8 show that a higher copper/zinc ratio coincides with a lower area fraction of β-phase. This provides a slight indication that the copper/zinc ratio of the system can control the visible area fraction of β-phase, and thus the dezincification resistance. There is, however, a need to investigate the effect of this factor more systematically to properly evaluate its impact on the dezincification resistance. The findings of this study will prove useful for the brass industry in the endeavour to adapt their processes to new low-lead and lead-free brass components. The next step in completing the transition to the lead-free ecosystem is to demonstrate how the absence of lead in the new alloys affects the machinability of them. Additionally, one unresolved mechanistic aspect of this work is if the formation of intermetallic arsenic particles also depletes the arsenic atoms from the bulk of the grains in addition to the grain boundaries.

Conclusions
Brass alloys CW511L, CW625N and CW626N were stress-relief annealed from 2 to 1000 h at temperatures in the range 250-400°C in order to map how precipitated phases impact the dezincification behaviour.
Stress-relieving heat treatments at temperatures higher than 250°C decrease the corrosion resistance only for the new low-lead alloys that contained aluminium, as a result of significant precipitation of βphase and intermetallic AlAs particles. For the alloy without aluminium or iron, the corrosion resistance remained completely intact through the annealing process, regardless of temperature or annealing duration.
Intermetallic particles are formed adjacent to grain boundaries as well as on other crystallographic defects, including twins and undissolved lead particles. The thermodynamic simulations indicate that aluminium or iron may accelerate the formation of β-phase and that lower copper/zinc ratio may be related to increased precipitation of β-phase.