1 Introduction

Age-hardenable Al–Zn–Mg–Cu alloys are widely used in the aerospace and automotive industries for their desired mechanical properties such as light weight, high strength and good formability [1]. With increasing developments in these industries, the need for Al alloys with superior mechanical, thermal and electrical properties is rising. These properties can be tailored by adjusting the parameters of not only typical heat treatment steps, such as solution heat treatment (SHT) followed by artificial ageing (AA), but also more advanced thermo-mechanical treatments consisting of SHT, different types of plastic deformation and AA [2, 3]. Thermomechanical treatments consisting of SHT, cold rolling (CR) and AA as a strengthening method in age-hardenable Al alloys have been investigated by several groups [4,5,6,7,8,9]. Wang et al. [3] reported that an Al–Zn–Mg–Cu alloy cold rolled to 67% at peak-aged condition increased the ultimate tensile strength (UTS) from 588 to 634 MPa at the expense of a decreased elongation from 8.6 to 4.4%. Zhao et al. [10] reported a simultaneous increase of elongation from 3.3 to 7.4% and UTS from 550 to 615 MPa in a nanostructured Al–Zn–Mg–Cu aluminum alloy by severe plastic deformation of cryo-rolling followed by AA. Li et al. [11], Panigrahi et al. [12], Wang et al. [13] and Chen et al. [14] have also reported simultaneous enhancement of ductility and strength by CR followed by AA, suggesting that during AA of CR samples, recovery and precipitate formation occur simultaneously and if precipitate formation dominates over recovery, a simultaneous enhancement of ductility and strength is achieved. In spite of the reported positive effects of CR, it has not been widely applied to practical Al alloys because CR introduces deformation heterogeneity which means a non-uniform distribution of dislocations. This leads to inhomogeneous texture evolution and non-uniform distributions of macroscopic stress and strain. Dislocations act as preferred nucleation sites for precipitation [15] which leads to inhomogeneous distributions of precipitates in the CR conditions.

The accepted precipitation sequence in Al–Zn–Mg–Cu alloys is [2,3,4,5,6]:

$${\text{SSSS}} \to {\text{GPI}}/{\text{GPII zones}} \to \eta \prime \to \eta_{{{1} - {14}}} \to \eta \left( {{\text{MgZn}}_{{2}} } \right)$$

describing the evolution of precipitates after SHT and during the AA treatment. Directly after SHT, the alloy is in a supersaturated solid solution (SSSS). During the initial stages of either natural ageing or AA, Guinier–Preston (GP) zones form. Two types of GP zones have been reported in Al–Zn–Mg–Cu alloys, namely the spherical GPI zones and the plate-like GPII zones [16,17,18,19,20,21].

There exist fourteen different orientation relationships between the metastable η and the Al matrix, denoted η1 through η14. While the interior structures of η1-14 are closely related to that of the equilibrium Laves η-MgZn2 phase, their interfaces with the Al matrix vary due to the different orientation relationships. The predominant metastable η phases commonly observed after ageing are η1, η2 and η4 [21,22,23]. Marioara et al. [24], Xu et al. [25] and Bendo et al. [26] studied the structures of η′, η2 and η1 at the atomic scale by high-angle annular dark field-scanning transmission electron microscopy (HAADF-STEM). η1 and η2 have plate morphologies and can be understood as layers of rhombic units (R) stacked across their plate thicknesses along < 001 > η1// < 110 > Al and < 001 > η2// < 111 > Al, respectively. The stacking of the rhombic units in η1 and η2 zigzag as R and its 180° rotated self R−1 stack across the plate thickness, resulting in a R-R−1 sequence. This is identical with the stacking along < 001 > η of the equilibrium phase. However, in η2, the zigzag of the rhombic unit can be disrupted by the presence of R-R sequences. Due to the specific orientations of the η1 and η2 phases with the Al matrix, the stacking of the R units across the plate thickness can only be observed when viewing along < 110 > Al and < 211 > Al zone axes, respectively. The thicknesses of defect-free η2 plates have discrete values but could vary when the precipitates incorporate defects [24].

Peak hardness in 7xxx alloys is often associated with a high density of the semi-coherent plate-shaped η′ precipitates. The η' phase is often referred to as a precursor to the η2 and both phases have the same orientation relationship with the matrix; [120]η′//[\(\overline{1 }\) 10]Al, [100]η′//[11 \(\overline{2 }\)]Al and (001)η′//(111)Al as the habit plane. η′ consists of a stacking of one R unit and one orthorhombic unit (O) across its thickness (R–O), resulting in a fixed thickness. The O unit in η' is isostructural with the O unit in the T1 phase (in Al–Cu–Li alloys) [27] and in the Ω phase (in Al–Cu–Mg–Ag alloys) [28]. Plate-shaped η', η2, T1 and Ω phases have the same orientation with the Al matrix and the same type of interface consisting of a compositionally enriched {111}Al plane which connects with the terminal rhombs and/ or orthorhombs. η' is the thinnest precipitate in undeformed conditions in 7xxx alloys and consists of an R stacked with an O plus the two enriched interface layers, one on each side of the plate. The reason for the fixed thickness of η' and the discrete thicknesses of defect-free η2 plates is dictated by the need of the XYXY and XA'XA' layers of the R and O units respectively, to match the ABCABC stacking of {111}Al planes across the plate thickness [24].

Bendo et al. [26] and Chung et al. [21] reported that C14 Laves phases usually contain various types of defects and bands of complicated planar defect structures. These often consist of the interconnecting rhombic unit and Mg6Zn7 elongated hexagons which play a key role to form defect structures. The elongated hexagons consist of the orthorhombic (O) unit in the center and half the rhombic unit (1/2 R) at each side of it [29]. Figure 1 shows that the planar defects are either chain-like configurations of elongated hexagons, or alternating sequences of elongated hexagons and rhombic η1 subunits. The orientation of η1 with the Al matrix can be written as [\(\overline{1 }\overline{2 }\) 0]η1//[001]Al, [100]η1//[110]Al and (001)η1//(\(\overline{1 }\) 10)Al and its habit plane is (010)η1//(001)Al [30]. The chain-like configuration runs parallel to the c-axis of the η1 precipitate having an atomic plane displacement of 1/6c (c = 0.857 nm for η-MgZn2), while the pair-wise configuration is at a certain angle to the c-axis having atomic plane displacement of 1/3c [31]. These configurations are generally present in platelet precipitates with high aspect ratios. In addition, there are precipitates with a random arrangement of elongated hexagons having low aspect ratios and forming round-shaped precipitates. Elongated hexagons are easily formed in precipitates in alloys with lower Zn/Mg ratios [26]. The precipitates may incorporate an additional flattened hexagonal sub-unit attributed to an internal stacking fault [21, 30] which results in structures with five-fold symmetry centers and Penrose-like tiling [31]. Yang et al. studied an over-aged condition and found highly disordered precipitates referred to as faulty C14 MgZn2 Laves phases. The formation mechanism of these highly disordered precipitates is presumably connected to relieving internal strain from lattice misfit between metastable η-type precipitates and the Al matrix by generating defect bands [31].

Fig. 1
figure 1

Illustration of MgZn2 Penrose bricks and Mg6Zn7 elongated hexagons subunits cell arrangements (atomic heights are not distinguished). a Systematic configuration. b Random configuration. The Zn atomic columns are drawn with yellow fill, and those with double occupancy are encircled by black lines. They appear with the brightest contrast in HAADF-STEM images. The Mg atomic columns are drawn with red fill [26, 31]

Full size image

The formation of a new phase called Y was recently reported as a quench-induced platelet precipitate in Al–Zn–Mg–Cu alloys. The formation of the phase depends on the quenching rate and the chemical composition of the alloy [32,33,34,35,36,37]. Zhang et al. [32] and Starink et al. [34] observed the Y phase in a 7150 Al alloy during step quenching and continuous cooling after hot rolling followed by SHT and AA. Liu et al. [35] and Tang et al. [36] also found Y phase precipitates in slowly-quenched 7085 and 7136 Al alloys, respectively. The Y phase exhibits a high aspect ratio and mainly contains Al, Zn, Cu and Mg. This phase is isostructural to the T1 (Al2CuLi) phase observed in Al–Cu–Li alloys having a thickness of c = 4dAl(111) [24]. Zhang et al. [32] found the orientation relationship of Y phase with the Al matrix as (001)Y //(111)Al and [110]Y //[112]Al, whereas Chung et al. [38] reported that (006)Y //(111)Al and [100]Y //[110]Al. Chemical compositions significantly affect the quenching sensitivity of Al–Zn–Mg–Cu alloys [39]. The high mobility of Cu atoms along dislocations results in heterogeneous nucleation and growth of the Y-phase on dislocations during slow cooling [32, 35]. Quenching sensitivity increases with increased total content of alloying elements (Zn + Mg + Cu) and decreased Zn/Mg ratios. In areas with a large number of dislocations the Y phase is preferentially nucleated instead of metastable η phases [33, 35, 40]. There is, however, limited information about the formation and nucleation mechanisms of the Y phase, especially for rapid quenching of pre-deformed Al alloys.

Pre-deformation before AA can alter the precipitation sequence [23, 41, 42]. Deschamps et al. [41] showed that the large density of dislocations introduced during plastic deformation alters the precipitation sequence. Therefore, coarser precipitates form on dislocations, often of the type associated with over-ageing rather than the main hardening phases [41, 42]. Allen et al. [43] and Deschamps et al. [44] found that in pre-deformed Al–Zn–Mg alloys, larger equilibrium η phases preferentially formed on dislocations.

In the present study, the effects of cold rolling with different reduction rates (30%–90% reduction) on AA behavior and mechanical properties of an Al–Zn–Mg–Cu alloy have been investigated by means of hardness measurements, tensile tests and HAADF-STEM. Atomically resolved HAADF-STEM was used to investigate the atomic structures of precipitates and clarify the type of precipitates and their behavior during ageing after cold rolling. This has been compared to the non-deformed condition.

2 Experimental Procedures

2.1 Materials and Heat Treatments

The samples used in this investigation were 10 mm thick plates of an Al–Zn–Mg–Cu alloy. Table 1 shows the chemical composition of the alloy. The samples were homogenized at 480 °C for 1 h in a salt bath followed by quenching in iced water. After homogenization, samples were pre-rolled to make sheets with desired thickness of 1.5 mm, 2.5 mm and 10 mm and then solution heat treated in a salt bath at 480 °C for 1 h again. The average quench rate from 480 °C to room temperature was estimated to be about 240 °C/s. Immediately after SHT, samples were cold rolled (CR) at room temperature (RT) with reduction ratios of 30%, 60% and 90% from their initial thickness. The specimens were CR from 1.5 mm to 1 mm (30%), 2.5 mm to 1 mm (60%), and 10 mm to 1 mm (90%). To achieve relatively uniform deformation and to avoid a strong temperature ramp during CR, a thickness reduction for each pass was maintained at less than 10%. The non-deformed (or as-quenched) and CR samples are abbreviated as ‘AQ’, ‘30%CR’, ‘60%CR’ and ‘90%CR’, respectively. AA was carried out in an oil bath at 120 °C for a duration of up to 14 days depending on the initial condition.

Table 1 Chemical composition of the investigated alloy in mass and atomic percent
Full size table

2.2 Hardness and Tensile Tests

Specimens were prepared as 1 × 1 cm2 squares with 1 mm thickness, and both sides were polished. Vickers hardness measurements were conducted at RT by applying a load of 300 g for 15 s using a Matsuzawa MMT-X micro hardness tester. Average hardness values were calculated based on seven out of nine data points, where the highest and lowest values were excluded from the calculation. Tensile specimens were prepared in a dog bone shape with dimensions of 10 mm in width, 60 mm in length and 1 mm in thickness at a parallel portion. Tensile tests were performed using a Shimadzu AG–X plus tester machine at a strain rate of 6.07 × 10–4 s−1 at RT. The specimen surface was ground with up to 2500 grit emery papers to remove surface imperfections. The tensile axis of the specimens was oriented parallel to the rolling direction.

2.3 Transmission Electron Microscopy

TEM specimens were prepared by grinding the samples, from an approximate initial thickness of 1 mm down to 100 µm. Subsequently, 3 mm diameter discs were punched out. After that, the discs were electropolished in a Struers TenuPol-5 machine using an electrolyte of 1/3 nitric acid (HNO3) and 2/3 methanol (CH3OH). The electrolyte liquid was kept within a temperature range of − 20 °C to − 30 °C during polishing. The STEM images were acquired using a double-corrected JEOL ARM 200F microscope operated at 200 kV using a convergence angle of 27 mrad. The low angle (LA)ADF-STEM images were collected using an inner collection angle of 29 mrad, while for the HAADF-STEM images, the inner collection angle was set to 67 mrad. The thickness of the region of interest was estimated using electron energy loss spectroscopy (EELS). The number density of precipitates was calculated by the equation of N/A(t + D), where N represents the total number of counted precipitates in a given image (in all orientations), A is the area of the STEM image where the precipitates are counted, t is sample thickness obtained from EELS data, and D represents the average diameter of the counted precipitates. D was calculated from the lengths of edge-on precipitates which represent precipitate diameters when viewing along < 110 > Al orientation. Details are given by Andersen et al. [29]. In each condition, at least 10 images were used to count the precipitates. The number of counted precipitates for the AQ, 30%CR, 60%CR and 90%CR conditions were 143, 89, 90 and 140, respectively. In addition to counting precipitates, sizes of η′, η2, and η1 type precipitates were measured by using the knowledge of their specific orientations with the Al matrix when viewed edge-on along < 110 > Al. As these precipitates are platelets with low aspect ratios we assume they have a circular shape and consider the projected length as diameter.

3 Results

3.1 Influence of Pre-deformation on Age-Hardening

Figure 2 shows the hardness evolution during AA of all four conditions studied. The AQ condition exhibits a rapid initial increase, a secondary increase leading to peak hardness and a final decrease during AA. The rapid initial increase in hardness is from 85 to 123 HV within 5 min. Then there is a second increase leading to a hardness of 173 HV at 24 h. Finally, the AQ sample has a plateau and reaches the peak hardness of 174 HV at 264 h. The final stage is characterized by a decrease in hardness by over-ageing.

Fig. 2
figure 2

Micro-Vickers hardness curves of AQ and 30%CR, 60%CR and 90%CR conditions aged at 120 °C for various times

Full size image

The hardness values in the initial stages of the 30%CR, 60%CR, and 90%CR conditions were 160 HV, 173 HV, and 205 HV, respectively, as seen in Fig. 2. The hardness curves show a plateau for the 30%CR and 60%CR conditions, where the hardness remains relatively constant for the first 20 min. Subsequently, there is an increase in hardness, with the 30%CR sample reaching a peak hardness of 179 HV at 18 h, while the 60%CR condition reaches a peak hardness of 191 HV at 14 h followed by a hardness decrease with increasing ageing time. The 90%CR condition showed a tendency of hardness decrease to 196 HV within the first 20 min. A slight increase in hardness, reaching a peak of 204 HV, was observed at 14 h, after which the hardness started to decrease. It is observed that increased deformation decreases the time to reach peak hardness and decreases the age hardening response.

3.2 Tensile Test Results

The stress–strain curves for all samples before ageing and at peak-aged conditions are shown in Fig. 3 and Table 2. The AQ sample before ageing exhibits the highest elongation rate of 27%, and the lowest YS and UTS of 163 MPa and 357 MPa, respectively. After 24 h ageing, a significant increase in YS and UTS to 495 MPa and 513 MPa, with a slight decrease of the elongation to 23% can be observed. The AQ sample at the peak-aged condition (264 h) achieves the highest YS and UTS of 495 MPa, 535 MPa, with a slight decrease in the elongation to 22%. When 30%, 60%, or 90%CR was applied before ageing, the samples show a significant increase in YS to 385 MPa, 533 MPa and 654 MPa and UTS to 466 MPa, 561 MPa and 664 MPa, respectively, but a significant decrease in elongation to 14%, 3% and 2.9%, respectively, as shown in Fig. 3b. After ageing, at the peak-aged condition, 30%CR shows an increase in YS and UTS to 452 MPa, 503 MPa as well as a simultaneous increase in elongation rate to 17%. The 60%CR at peak-aged condition also shows a simultaneous increase of YS and UTS and elongation rate to 544 MPa, 564 MPa and 13%, respectively. In contrast, the 90%CR at peak-aged condition exhibits a decrease in UTS from 664 to 624 MPa, but a slight increase in elongation rate from 2.9% to 3.5%. Based on the tensile test results, UTS increases for the 30%CR and 60%CR samples, while all CR samples show an increase in elongation after ageing.

Fig. 3
figure 3

a Engineering stress–strain curves of AQ sample for various ageing conditions. b Engineering stress–strain curves of 30%CR, 60%CR and 90%CR for various ageing conditions

Full size image
Table 2 The tensile properties of the investigated alloy after different ageing treatments
Full size table

3.3 Transmission Electron Microscopy

3.3.1 Precipitate Evolution During Ageing

Figure 4 shows simultaneously acquired HAADF-STEM and LAADF-STEM images of AQ and CR samples in the peak-aged condition along the < 110 > Al zone axis. The LAADF-STEM images exhibit diffraction contrast in addition to Z contrast, hence dislocations and other deformation induced defects are visible, while the HAADF-STEM images exhibit mostly Z-contrast, thus visualizing the precipitate distribution. In the AQ sample at the peak-aged condition shown in Fig. 4a, the precipitates are uniformly distributed within the matrix, and have an average size of 6.9 nm in projected length and 2.2 nm in thickness. The LAADF-STEM images of the AQ sample show similar features as the HAADF-STEM images due to the negligible number of dislocations and defects in AQ as shown in Fig. 4e.

Fig. 4
figure 4

a–d Simultaneously acquired HAADF-STEM images a–d and LAADF-STEM images e–h at peak-aged condition in AQ samples and different cold rolling ratios along <110> orientation. a, e AQ for 264 h, b, f 30% CR for 18 h, c, g 60%CR for 14 h and d, h 90%CR for 14 h

Full size image

An HAADF-STEM image of 30%CR in the peak-aged condition is shown in Fig. 4b. Here, both coarse precipitates and fine precipitates co-exist. Measurement of 20 coarse precipitates, as exemplified by red circles, give an average size of 15.7 nm in length and 3.8 nm in thickness, while a measurement of 20 fine precipitates as exemplified by yellow circles, resulted in an average size of 8.7 nm in length and 3.1 nm in thickness. In the corresponding LAADF-STEM image of 30%CR shown in Fig. 4f, regions of higher and lower densities of dislocations are clearly distinguishable. A comparison of Fig. 4b and f clearly shows that the coarse precipitates preferentially nucleated heterogeneously on the deformation induced dislocations, while the finer precipitates nucleated in undeformed areas in-between the dislocations.

HAADF-STEM and LAADF-STEM images of 60%CR and 90%CR in the peak-aged condition are shown in Fig. 4c,d and g,h, respectively. In 60%CR and 90%CR, it is observed that the volumes with high density dislocations increase with increasing CR. Moreover, the size of the coarse precipitates nucleated on dislocations increased with increasing CR, while the size of fine precipitates nucleated in undeformed areas was similar regardless of CR ratio. Total overall number density of precipitates in CR conditions significantly decreased with increasing CR ratio.

The calculated number densities and average sizes of precipitates in the AQ and different CR conditions are shown in Table 3 and Fig. 5. The η′ and η2 precipitates are low aspect ratio platelets with {111}Al habit planes. Therefore, when viewed along < 110 > Al, half of them will be observed edge-on on two of the four {111}Al planes, and half will be observed at oblique angles showing oval projections of their faces. The ratio of edge-on precipitates on {111}Al planes over the rest of precipitates for AQ at 24 h ageing and peak-aged conditions was 1:1.3 and 1:1.5, respectively. For the 30%, 60% and 90% CR samples at their respective peak-aged conditions the ratio was 1:1 regardless of CR degree. It means that the AQ conditions must contain other types of η phases (with different orientations in the Al matrix) in addition to η′ and η2 precipitates.

Table 3 Number density, average diameter and average thickness of platelets viewed edge-on along a < 110 > Al projection
Full size table
Fig. 5
figure 5

Distribution of precipitate diameters measured in the AQ, 30%CR and 60%CR peak-aged conditions

Full size image

3.3.2 Types of Precipitates in the Non-deformed Condition

Figure 6 shows HAADF-STEM images of the AQ sample at the peak-aged condition along the < 110 > Al zone axis. Two types of plate-shaped precipitates, indicated by arrow 1 and arrow 2, were observed in the wide view in (a). Based on their widths and {111}Al habit plane, they can be regarded as η′ and η2, respectively. The η′ phase is shown enlarged in (b) and has a thickness of 6 atomic layers while the η2 phase is shown enlarged in (c) and has a thickness of 24 atomic layers. In addition to η′ and η2 phases, precipitates indicated by arrow 3 in Fig. 6a were also observed. Such precipitates will be referred to as a disordered η phase in this investigation with examples shown in Fig. 7. Figure 7a,b show HAADF-STEM images of disordered η phases viewed along < 110 > Al zone axis. Atomic scale examination revealed that the upper parts of both precipitates exhibits the structure of C15 and C14 MgZn2 stackings of the rhombic unit (R) along [\(\overline{1 }\) 10]Al as indicated with the yellow lines, which is identical to the orientation relationship of η1 phase in the Al matrix. The C14 stacking consists of alternating R-R−1 layers as in the equilibrium η phase, while C15 has the rhombic units stacked in R-R layers. The lower parts of the precipitates have incorporated Mg6Zn7 elongated hexagons in random orientations, with one indicated by green fill in Fig. 7a. In addition, the precipitate in Fig. 7b contains five-fold symmetry indicated by blue fill and constructed by five rhombic units sharing a corner. These disordered η precipitates have an equiaxed rather than a plate-shaped morphology. The elongated hexagon and the five-fold symmetry features are enlarged in Fig. 7c,d. Similar defects consisting of elongated hexagons and five-fold symmetry have also been observed in precipitates imaged along < 112 > Al, see for example Fig. 3 and 4 in reference [31]. This indicates that such defects are not only associated with the orientation of η1 phase in the Al matrix.

Fig. 6
figure 6

a HAADF-STEM image from the AQ material at peak-aged condition viewed along < 110 > Al zone axes. b and c Enlarged images of precipitates indicated by arrows 1 and 2, respectively

Full size image
Fig.7
figure 7

HAADF-STEM images of disordered η phases indicated by arrow 3 in Fig. 6a in AQ samples at peak-aged conditions along < 110 > Al zone axes, a, b Two representative disordered η phases, c Enlarged image indicated by red frame in Fig. 7a, d Enlarged image indicated by red frames in Fig. 7b

Full size image

3.3.3 Types of precipitates in the cold rolled samples

Figure 8 shows HAADF-STEM images of 30%CR at peak-aged condition along the < 211 > Al zone axis. Most of the precipitates in this condition are identified as η′ and η2 phases based on their sizes and {111}Al habit plane. Figure 8a shows a smaller η2 having C14 (R-R−1) stacking of the rhombic unit along its thickness, while Fig. 8b shows a coarse η2 precipitate containing both C14 and C15 (R-R) stackings. The defect-induced five-fold domains were also observed in the coarser precipitates, but only formed at the incoherent interface with the Al matrix, as shown in Fig. 8c.

Fig. 8
figure 8

HAADF-STEM images of precipitates at peak-aged condition in 30%CR viewed along a < 211 > Al orientation. a Fine η2 precipitate. b Coarse η2 precipitate. c Coarse η2 precipitate with five-fold symmetry at the edge interface

Full size image

HAADF-STEM images of 60%CR and 90%CR at peak-aged conditions along < 211 > Al zone axes are shown in Fig. 9a–f. Fine and coarse precipitates in both 60%CR and 90%CR shown in Fig. 9a,d have coexistence of C14 and C15 stackings. However, the coarse precipitates show increased amounts of stacking faults with increasing precipitate size and CR degree as shown in Fig. 9b,e. Five-fold symmetries are also observed, see Fig. 9e,f. The round disordered η precipitates are not observed in CR samples.

Fig.9
figure 9

HAADF-STEM images of precipitates at peak-aged conditions in 60%CR and 90%CR viewed along the < 211 > Al orientation. a, d Representative fine η2 precipitates images. b, e Coarse η2 precipitates containing stacking faults and R–R stacking. c, f Coarse η2 precipitates containing five-fold symmetry configurations at edge interfaces

Full size image

Thin plate-like precipitates having high aspect ratio and {111}Al habit planes were observed along < 110 > Al and < 211 > Al directions in addition to η′ and η2 precipitates at peak-aged condition of 30%CR and 60%CR. Figure 10 shows HAADF-STEM images of these precipitates, identified as the plate-like Y phase [32]. In Fig. 10a viewed along the < 110 > Al orientation, it is observed that the precipitate has a thickness of 4d(111)Al. When viewed along the < 211 > Al direction, it is observed that the precipitate imaged in 30%CR consists of a single rhombic layer connecting to the Al matrix by the same compositionally modified {111}Al planes the as η′ and η2 precipitates [24], see Fig. 10b. The precipitate imaged in the 60%CR condition consists of a single orthorhombic layer connecting to the Al matrix by the same compositionally modified {111}Al planes as the η′ [24], see Fig. 10c. Thus, the Y phase is the thinnest type of precipitate observed in the Al–Zn–Mg system and has the same orientation relationship and the same type of interface with the Al matrix as η′ and η2. The Y phase is not found in the 90%CR condition. However, it is reasonable to assume that the Y phase may also form in 90%CR, but due to its low number density (and relatively small volume studied), it was not observed.

Fig. 10
figure 10

HAADF-STEM images showing Y phase in 30%CR and 60%CR at peak-aged conditions viewed from both < 110 > Al and < 211 > Al orientations. a Image of precipitate in 30%CR along < 110 > Al b Image of precipitate in 30%CR along < 211 > Al. c Image of precipitate in 60%CR along < 211 > Al

Full size image

4 Discussion

It is well known that pre-deformation has a significant effect on the mechanical properties of Al alloys by combined strengthening effects from precipitates and deformation induced dislocations. In this work, the effect of CR has been studied in an Al–Zn–Mg–Cu alloy by hardness measurements, tensile tests and HAADF-STEM analysis.

4.1 Influence of Cold Rolling on Age-Hardening Behavior

Higher initial hardness and strength of the CR conditions compared to that of the AQ conditions before AA treatment are owing to the accumulation of dislocations. According to the Taylor relationship which is based on dislocation theory [12], factors affecting the strength of AQ and CR can be calculated as

$$\sigma_{{\text{y}}} = \, \sigma_{{\text{o}}} + \, \alpha {\text{M}}^{{\text{T}}} {\text{Gb}}\rho^{{{1}/{2} }} + \frac{{{\text{Gb}}ln\left( {x/b} \right)}}{{2\pi \left( {l - x} \right)}}$$
(1)

where σo, α, G, b, MT, ρ, x and l are the friction stress, a constant, shear modulus, Burgers vector, Taylor factor, dislocation density, average size of precipitates and interparticle spacing, respectively. In this equation, the second term is associated with dislocation strengthening behavior and the third term is associated with precipitation hardening. In AQ before AA, the strength contributions from both the second and the third terms are negligible, while in CR before AA, the third term is negligible. When the AQ sample is subjected to peak-age treatment, significant increases in hardness and YS (204% and 304%, respectively), were observed. This is due to the contribution from precipitates with small size, high number density, less interparticle spacing and uniform distribution which is related with the third term of Eq. (1). During AA of the AQ sample, excess quenched-in vacancies have enabled diffusion of solute atoms to form fine precipitates. These result in a rapid increase of strength in the early stages of ageing and further increase of strength at peak-aged condition, by the formation of η′ precipitates. An optimal balance of coherent/semi-coherent precipitates of mostly η′ phase at the peak-aged condition acts as obstacles for dislocation movement, resulting in maximum strength of the AQ sample. The negligible number density of dislocations in the AA condition leads to a relatively high elongation rate.

The tensile tests of the CR conditions show that the YS of 60%CR is about √2xYS of 30%CR and YS of 90%CR is about √3xYS of 30%CR which are directly proportional to the square root of the dislocation density implying that the dislocation density increases linearly with the CR degrees. It suggests that the increased strength of the CR samples is associated with deformation induced dislocation densities and that dislocation strengthening has a significant contribution to the strength enhancement in all CR samples. During the subsequent AA, there is an initial incubation stage. This may be explained by the suppression of the GP zone formation due to the lower supersaturation of solute atoms as compared to the AQ samples. At the same time, the hardness plateau might occur due to combined effects of dislocation annihilations leading to recovery and precipitate formations balancing the hardness value at the early stage of ageing. The dislocations act as heterogeneous nucleation sites for precipitates by reducing the energy barrier as well as promoting diffusion of solute atoms [44]. This reduces the time to reach the peak-aged condition by a factor of ten as compared to the AQ sample and also coarsening of precipitates on dislocations. In 30%CR at peak-aged condition, the increases in hardness and UTS were observed to be 112% and 108%, respectively, as compared to AQ at peak-aged condition. Less improvements in hardness and tensile strength for 60%CR and 90%CR at peak-aged condition compared to that of AQ were observed. An explanation of this could be the combined effect of recovery (decreasing the number density of dislocations) and coarsening (decreasing the number density of precipitates). With increasing CR degrees, the age hardening response related to the third term of Eq. (1) decreased due to increased interparticle spacing and coarsening of precipitates on dislocations. The significant decrease of elongation by CR is due to the high level of pre-deformation induced dislocations. Therefore, the materials’ ability to accumulate new dislocations will be low, which will drastically reduce the elongation as compared to the AQ condition. Increase of strength and ductility for 30%CR and 60%CR at peak-aged conditions as compared to their respective initial non-heattreated CR conditions, indicate strengthening by precipitate formation and availability of more sites for dislocation accumulation leading to a higher plasticity during AA due to dislocations annihilation. In the 30%CR and the 60%CR samples at peak-aged condition, precipitation hardening is dominating over recovery resulting in an increase in hardness, tensile strength and elongation during AA (c.f. Fig. 2). In the 90%CR sample at peak-aged condition, recovery is dominating over precipitation hardening resulting in increased elongation but decrease in hardness and tensile strength». This work agrees well with previous research [8]. The peak-aged AQ sample exhibits homogeneous precipitation, while the peak-aged CR samples contain both coarse precipitates nucleated on dislocations, as well as finer precipitates nucleated in undeformed areas in-between the dislocations. The average particle size in the CR conditions increased and their number density decreased as compared to the AQ condition. It is suggested that both dislocations and precipitate strengthening play a role in enhancing the strength in CR samples. However, the age hardening response in the CR conditions is significantly less than in the AQ condition due to the high number density of dislocations, causing low density of coarse heterogeneous precipitation.

4.2 Influence of Cold Rolling on the Precipitation

4.2.1 Precipitation and Precipitate Morphology in the Non-deformed Condition

The precipitates in the AQ sample at peak-aged condition were investigated by atomically resolved HAADF-STEM and LAADF-STEM. The precipitates were predominantly η', η2 and disordered η phases. Zou et al. [45] explained that the ratio of edge-on over face-on precipitates in his work was 1:1 when viewed along the < 110 > Al zone axis assuming that most of precipitates are η′ or η2 and there were equal amounts of η′ or η2 phase on each equivalent (111)Al plane. In our study, the number of projected round-shaped particles is higher than that of edge-on plate-like particles, with a ratio of 1.51:1. Therefore, according to the ratio, in addition to η′ or η2 phases, other types of metastable η phases accounted for approximately 20% of the total number of precipitates during AA. Most of these other phases consisted of disordered η displaying irregular arrangements of elongated hexagons known as Mg6Zn7 [26] connecting with rhombic units of C14 MgZn2 that zig-zag in a R-R−1 stacking sequence. The zig-zag direction was along < 110 > Al which is typical for the orientation relationship of the η1 phase in the Al matrix. These precipitates are more round shaped with low aspect ratio, rather than plate shaped with high aspect ratio.

The formation mechanism of the disordered η phase is unclear. The interplanar spacings of the Al matrix are 0.405 nm for (100)Al, 0.286 nm for (110)Al and 0.234 nm for (111) Al. For the MgZn2-η phase the spacings are 0.856 nm for (001)η, 0.429 nm for (100)η and 0.248 nm for (120)η, respectively [20]. From the data of interplanar spacings and orientation relationships, lattice misfits by evolution of precipitates on Al matrix are calculated as shown in Table 4. Chung et al. [21] and Bendo et al. [23] have calculated the relatively small lattice misfit as 3.4% and 0.34%, respectively, which are vertical to the (001)η1 basal plane based on the periodical interplanar spacing of d(001)η1 corresponding to 3d(110)Al. However, if one compares four Zn rich layer of η1 phase(d(001)η1 layer) corresponding to four Al layer of 4d(110)Al in order to retain the high coherency between matrix of (110)Al layer and η1 precipitate of 1/4(001)η1 layer, then the lattice misfit is recalculated by 25% of d(001)η1//4d(110)Al. It indicates that the η1 precipitate has relatively large lattice misfit of d(001)η1//4d(110)Al as well as d(100)η1//d(001)Al and d(120)η1//d(110)Al. Different orientations of the elongated hexagon structures are observed in the disordered η1 precipitates. The different orientations help to reduce misfit stress with the Al matrix in different directions. For example, elongated hexagon chains parallel to the c-axis of η1 have atomic plane displacements of 1/6c relieving the strain parallel to the c-axis, while elongated hexagons with rotated R units due to incorporation of C15 stacking have atomic plane displacements of 1/3c relieving strain along both c-axis and a-axis [31]. These are systematic configurations which can reduce lattice misfit stress as well as preserve plate-shaped morphology. The highly disordered η precipitates containing a high number of randomly tilted elongated hexagons and five-fold domains with round-shaped (equiaxed) morphology transform from η1 to relieve strain inside these precipitates. Alternatively, they may be an early stage in the precipitate evolution towards the η1 phase due to suppressed formation of η1 precipitates by the low AA temperature below 150 °C [26]. Also, the precipitate morphology is very sensitive to the Zn/Mg ratio of the alloy. Bendo et al. [26] reported that with a decreasing Zn/Mg ratio from 2 to 1.6, the aspect ratio of precipitates decreased from 8 to 2. They also found that the number density of elongated hexagons in η1 increased from 0 to 20%. In our study the Mg/Zn ratio is close to 1. It is therefore reasonable to infer that the disordered η phase with a high number of elongated hexagons with random orientations is preferentially transformed from η1. In the η2 precipitates, due to the low lattice misfits of d(100)η2//3d(220)Al and of d(120)η2//3d(224)Al, only elongated hexagons parallel to the c-axis relieving strain along this direction were observed and no disordered η precipitates with random tilt elongated hexagon structures were found. In conclusion, the strength in the AQ samples is generated not only by the η′ and η2 phases on {111}Al planes, but also by the disordered phase.

Table 4 Calculated lattice misfits between precipitate and matrix plane
Full size table

4.2.2 Precipitation and Precipitate Morphology in CR Samples

For the CR samples, it was found that the peak-aged conditions contain η', η2 and a low number of the newly identified Y phase. During CR, dislocations are non-uniformly generated and accumulated in the matrix, resulting in distorted regions with high dislocation density that are surrounded by undistorted regions with low dislocation density. This non-uniform distribution of dislocations leads to size inhomogeneity, as well as a low number density of precipitates. It is presumable that coarse precipitates (> 15 nm) in the distorted regions make only small contributions to age-hardening, while the low number of fine precipitates (< 6–8 nm) in undistorted regions has a stronger influence on age-hardening. Due to these effects, the age hardening response of the CR samples is low compared to that of the AQ samples. The non-uniform distribution of dislocations and coarsening of precipitates also significantly reduce ductility. The high growth rate of precipitates on dislocations in CR conditions resulted in an accelerated coarsening kinetics and large precipitate sizes. Inside the coarse precipitates, a high number of defects was observed. The number of stacking faults and stepwise growth with different thicknesses along the lateral direction in η2 precipitates increase with the CR degree due to accelerated nucleation and high level of stored energy by deformation induced dislocations. All coarse precipitates have a plate-shape morphology. The aspect ratio of the coarse precipitates decreased from 4.0 to 3.1 with increasing CR from 30%CR to 60%CR, indicating that widening is preferred over lengthwise growth in these precipitates due to the growth inhibition imposed by the pentagonal features [46].

Unlike in the AQ sample, disordered η phases with equiaxed morphology were not observed in the CR samples. This was supported by the fact that the ratio between edge-on and face-on η precipitates was approximately 1 in these samples. The absence of the disordered η phase in CR conditions may be due to the accelerated nucleation and growth rate of η2 phase precipitates as platelets with {111}Al habit planes. This effect could provide sufficient diffusion paths for solute atoms resulting in η2 rather than η1 with habit planes parallel to {100}Al. Due to the relaxation of lattice misfit between precipitates and matrix at the dislocation core, disordered η phases transformed from η1 are hardly seen.

In this investigation, Y phase precipitates having either rhombohedral or orthorhombic appearance as viewed along < 112 > Al were found in 30%CR and 60%CR at peak-aged conditions, respectively, as shown in Fig. 10. All previous reports indicate that the Y phase only precipitates during slow quenching in alloys with low Zn/Mg ratio and high Cu contents [40, 41, 43]. This is because Y phase formation needs both sufficient time and preferred diffusion paths like dislocations, due to slow bulk diffusion rate of Cu compared to Zn and Mg. To the authors’ knowledge, the Y phase has not previously been reported to form in pre-deformed Al–Zn–Mg–Cu alloys with SHT followed by rapid quenching. We suggest the origin of the deformation induced Y phase in our case is due to the combined effects of two factors:

  1. 1.

    The high density of dislocations from CR provides short and easy diffusion paths for solute Cu atoms. The dislocations act as nucleation sites for the Y phase.

  2. 2.

    The low Zn/Mg ratio and high Cu content make the evolution of Y precipitates easier due to rapid quenching and a low temperature ageing (120 °C). The dependence of Y phase formation on CR ratio is unclear and Y phase precipitates are not distributed homogeneously in 30% and 60% CR conditions.

The Y phases in 30%CR and 60%CR have the same orientation relationship with the matrix as the η′ and η2 type precipitates. However, the observed Y precipitate in 30%CR has a projected rhombohedral (R) structure consisting of 3 layers (XYX [24]) and 2 interfaces of compositionally modified {111}Al, making it 4d(111)Al wide. The observed Y precipitate in 60%CR has a projected orthorhombic (O) structure also consisting of 3 layers (XA'X [24]) and 2 interfaces of compositionally modified {111}Al, making it also 4d(111)Al wide, see Fig. 10. It is interesting to notice that the Y variant containing the orthorhombic unit is isostructural with the T1 phase in the Al–Mg–Li system. The Y variant containing the rhombic unit has not been previously reported.

The η′ phase consists of one rhombic unit on top of the orthorhombic structure and with the same 2 interfaces of compositionally modified {111}Al, making it 6d(111)Al wide [24]. In this respect the Y phase is the thinnest platelet observed in Al–Zn–Mg alloys. However, in undeformed matrix η′ platelets should be the thinnest because the rhombic unit and the orthorhombic structures consist of two alternating layers (XY for the rhombic unit, and XA' for the orthorhombic unit) that must match with the ABC alternation of {111}Al planes along the precipitate thickness [24]. Therefore, in the case of η′ we have the following sequence through the thickness: AXA'XYXA, as the identical top and bottom X layers face identical A layers of the matrix. However, in the case of the Y phase the sequence through thickness becomes AXYXB for the rhombic unit, and AXA'XB for the orthorhombic unit, resulting on one side of the plate an X connecting to an A, and on the other side an X connecting to a B, see Fig. 10. The only explanation for the existence of the Y phase therefore is the presence of a dislocation in the Al matrix that will disrupt the ABC sequence of the {111}Al planes. In the author's opinion the Y phase can only nucleate on dislocations, even when the other conditions are met, such as a low Zn/Mg ratio and a high Cu content.

5 Conclusion

The effects of pre-deformation of an Al–Zn–Mg–Cu alloy with a low Zn/Mg ratio have been investigated by means of mechanical properties and precipitate evolution during AA at 120 °C.

  1. (1)

    The CR conditions showed higher mechanical strength than the AQ condition for all ageing times, and this effect increased with higher deformation ratios due to work hardening. A higher age hardening response was observed for the AQ sample compared to that of the CR samples. The uniformly distributed SSSS containing high levels of quenched-in vacancies combined with negligible numbers of dislocations in the AQ sample enhanced precipitation of fine, homogeneous coherent/semi-coherent precipitates during ageing. However, in the CR samples, the pre-existing non-uniformly distributed dislocations lead to formation of coarser precipitates during ageing, having significantly reduced number densities as compared to the AQ condition.

  2. (2)

    The AQ sample at peak-aged condition contained η', η2 and disordered η phases. The structure of the disordered η phase displays irregular arrangements of rhombic units and elongated hexagons of Mg6Zn7 configuration connected with rhombic units. This type of precipitate either transforms from the η1 phase to relieve strain from its lattice misfit with the matrix, or it may be an early stage having high numbers of defects such as elongated hexagon and five-fold symmetry. The defects disturb the coherency of the disordered η phase with the matrix, leading to round-shaped (equiaxed) morphology. The formation of η1 precipitates might be suppressed by the low AA temperature at 120 °C and low Zn/Mg ratio of the alloy. Therefore, the strengthening precipitates in the AQ sample are not only the η′ and η2 phases with {111}Al habit plane, but also the disordered η phase.

  3. (3)

    The CR samples at peak-aged conditions contained η' and η2 phases in large amounts, and the newly identified Y phase in low amounts. The disordered η phase was not observed in the CR samples presumably due to the accelerated nucleation rate by the deformation induced dislocations which provide sufficient diffusion paths for solute atoms. This results in more defect-free precipitates with periodic atomic arrangements. The nucleation of η', η2 and Y must be enhanced by the presence of dislocations as all these phases have the same coherency with the Al matrix and grow as platelets with {111}Al habit planes.

  4. (4)

    The Y phase is a platelet containing either one rhombohedral or one orthorhombic unit connecting to the Al matrix by one compositionally modified {111}Al habit plane at each side. This gives the Y phase the same type of interface and the same orientation relationship with the matrix as the η′ and η2 phases. The Y phase is the thinnest observed platelet in the Al-Mg-Zn alloy, with 4d(111)Al thickness. The Y phase was observed on dislocations in the 30 and 60% CR samples with rapid quenching and it is considered a deformation induced phase. Due to the specific stacking of its layers, it can be demonstrated that the formation of Y needs a dislocation that would disrupt the ABC stacking of the {111}Al planes in Al matrix.