Ultra-large single crystals by abnormal grain growth

Producing a single crystal is expensive because of low mass productivity. Therefore, many metallic materials are being used in polycrystalline form, even though material properties are superior in a single crystal. Here we show that an extraordinarily large Cu-Al-Mn single crystal can be obtained by abnormal grain growth (AGG) induced by simple heat treatment with high mass productivity. In AGG, the sub-boundary energy introduced by cyclic heat treatment (CHT) is dominant in the driving pressure, and the grain boundary migration rate is accelerated by repeating the low-temperature CHT due to the increase of the sub-boundary energy. With such treatment, fabrication of single crystal bars 70 cm in length is achieved. This result ensures that the range of applications of shape memory alloys will spread beyond small-sized devices to large-scale components and may enable new applications of single crystals in other metallic and ceramics materials having similar microstructural features.

, where R and 0 R are average grain radii at annealing time t and that at t = 0, respectively, and k is constant. The grain growth exponent, n, is theoretically 2 in a single-phase condition, but n determined in the whole time range was 12 in the present study. The k for n = 2 can be determined for annealing time up to 3.6 ks to be 6.97 × 10 -14 m 2 s -1 , as shown in the inset. The lower growth rate of the subgrains is mainly due to the low mobility of low-angle boundaries.

Accuracy of misorientation
The error of the electron backscatter diffraction (EBSD) experiments should be considered because the accuracy of the absolute orientation in EBSD is said to be 0.5° to 1°2 , 3 . In order to determine the average orientation of each subgrain, at least 200 points in one subgrain were averaged. The standard error σM is given by where σ is the standard deviation and n is the number of data. We use σ = 1° and n = 200 as the worst case, and σM is calculated to be 0.07°. Therefore, the maximum error of difference between average orientations is 0.14°, and the average θ can be written as 0.46º ± 0.14° and 1.12º ± 0.14° for one low-temperature cycle (LTC) and five LTCs, respectively. It is also important not to miss the sub-boundaries in the grain reference orientation deviation (GROD) mapping. Supplementary Fig. 1c shows the GROD map with the minimum misorientation of 0.11° detected in this study.

Grain selection during AGG
It is an important issue whether or not grain selection for abnormal grain growth (AGG) occurs. Every grain boundary receives stored energy for AGG and every grain can grow unless some grain boundaries are pinned by precipitates or have low boundary energy. As seen in the microstructures of Fig. 4a (5 min and 10 min), only a limited number of grains start to grow in finer grained matrix, namely, grain selection and a typical microstructural feature of AGG. It was confirmed in the present case that no special grain-boundary-character distribution develops, in contrast with the usual AGG. We also observed that AGG occurs at temperatures several tens of Kelvin higher than the α solvus temperature.
Careful observation of the microstructure of the normal grains in Fig. 2c reveals that the orientation deviation is lower in the vicinity of high-angle boundaries and that the density of subgrains around the high-angle boundaries is relatively lower than in the central part of grains.
This typical microstructure may be due to formation of a low density zone (LDZ) of α precipitates near grain boundaries in the two-phase temperature region as seen in Fig. 5.
Therefore, AGG may not immediately start until the grain boundaries cross over the LDZ of subgrains by normal grain growth (NGG), namely, some "incubation" time for AGG is required.
This means that grain selection for AGG occurs due to microstructural inhomogeneity in the initial NGG process. In order to judge this possibility, we prepared a sample with the different microstructure obtained by cyclic heat treatment (CHT) with rapid cooling and heating, which has a relatively thin LDZ of subgrains, as shown in Supplementary Fig. 3a. The specimen slowly heated to the β single phase region at 800 °C after rapid cooling has a relatively small and uniform grain structure, as seen in Supplementary Fig. 3b. It is interesting to note that most grains hardly include subgrains, as shown in Supplementary Fig. 3c, meaning that since AGG occurs simultaneously for many grains, the AGG grains immediately come in contact with one another, resulting in formation of a homogeneous microstructure with relatively small grains.
This result means that the width of the LDZ of subgrains, resulting from the LDZ of α precipitates, is one of the most important factors for development of the inhomogeneous AGG and that the slow cooling in the CHT, which yields a coarse precipitation structure, is very important. In this consideration, it should be pointed out that the heating process in the CHT is also very important. The "incubation" time for the grain boundaries to cross over the LDZ of subgrains is obviously related to the dissolution process of the α precipitates. Since being able to easily move in advance of other boundaries with pinning precipitates, some grain boundaries, where α precipitates are resolved early, may be able to cross over the LDZ of subgrains earlier.
This means that the slower the CHT heating rate, the more remarkable the microstructural inhomogeneity. Actually, in our experiments, larger abnormal grains were obtained by decreasing both the cooling and heating rates of the CHT.
When a limited number of previously formed abnormal grains (AGs) grow in the surrounding normal grains after the following CHT, the large AGs may initially start to selectively grow by the NGG mode caused by the effect of the curvature of the grain boundary.
After the grain boundaries of the AGs cross over the LDZ of subgrains, however, drastic growth by the AGG mode actually starts. Therefore, the same large AGs are always selected as growing grains in further CHTs. Such grain selection, however, may no longer occur in a bamboo structure. Nevertheless, the total number of grains continues to decrease through further CHT due to the disappearance of grain boundaries at the ends of the sample and the unification with neighbouring grain boundaries.