© 2007 Nature Publishing Group LETTERS Tensile ductility and necking of metallic glass H. GUO 1 , P. F. YAN 1 , Y. B. WANG 1 , J. TAN 1 , Z. F. ZHANG 1 , M. L. SUI 1 * AND E. MA 2 1 Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences, Shenyang 110016, China 2 Department of Materials Science and Engineering, The Johns Hopkins University, Baltimore, Maryland 21218, USA *e-mail: mlsui@imr.ac.cn Published online: 19 August 2007; doi:10.1038/nmat1984 Metallic glasses have a very high strength, hardness and elastic limit. However, they rarely show tensile ductility at room temperature and are considered quasi-brittle materials 1,2 . Although these amorphous metals are capable of shear flow, severe plastic instability sets in at the onset of plastic deformation, which seems to be exclusively localized in extremely narrow shear bands ∼10 nm in thickness 3–13 . Using in situ tensile tests in a transmission electron microscope, we demonstrate radically different deformation behaviour for monolithic metallic-glass samples with dimensions of the order of 100 nm. Large tensile ductility in the range of 23–45% was observed, including significant uniform elongation and extensive necking or stable growth of the shear offset. This large plasticity in small-volume metallic-glass samples did not result from the branching/deflection of shear bands or nanocrystallization. These observations suggest that metallic glasses can plastically deform in a manner similar to their crystalline counterparts, via homogeneous and inhomogeneous flow without catastrophic failure. The sample-size effect discovered has implications for the application of metallic glasses in thin films and micro-devices, as well as for understanding the fundamental mechanical response of amorphous metals. In sharp contrast to crystalline metals that have large tensile ductility including significant uniform elongation, monolithic metallic glasses show little or no macroscopically observable tensile strain at room temperature 1–13 . Under compressive loading with or without confinement, plasticity is often observable, but is always highly inhomogeneous. The strains are concentrated in narrow shear bands that are not only few in number but also tend to run wild to cause early failure. It is thus believed that the vast majority of the metallic-glass sample volume does not contribute to plastic deformation 1 , and severe strain localization is the only deformation mode at temperatures well below the glass-transition temperature. In the following, we demonstrate qualitatively different behaviour in small-volume metallic glasses. The behaviours common to ductile crystalline metals, including uniform elongation, necking and stable shear, can all happen when the sample dimensions of the metallic glasses are brought into the submicrometre to nanometre range. Hints for important changes in deformation modes have emerged recently in micrometre-sized samples 14–16 (Z.W. Shan et al., unpublished). To observe the entire sequence of deformation stages in small samples, we carried out in situ tensile straining experiments in a transmission electron microscope (TEM) on several monolithic metallic-glass samples with dimensions in the 100 nm range. The material studied was a typical bulk metallic glass, Zr 52.5 Cu 17.9 Al 10 Ni 14.6 Ti 5 , prepared using copper-mould casting 10 . This alloy was studied previously in conventional mechanical tests 10 : the total plastic strain to failure was ∼1% in compression and nearly zero in tension before fracture (the elastic strain was ∼1.7%). A slice with dimensions of 45 μm (thickness) × 700 μm (width) × 3.3 mm (length) was cut from the bulk sample and glued to a brass substrate for in situ tension straining, as shown in Fig. 1a. Small test samples with a gauge section (the straight portion) of about 100 nm × 100 nm × 250 nm were fabricated near the centre of the upper edge of the slice (Fig. 1b), using the dual-focused-ion-beam (FIB) micromachining technique 17 , as shown in the schematic diagrams in Fig. 1b–d. The final sample sets, as shown in scanning electron microscope (SEM) images (Fig. 1e,f) viewed from the angles in Fig. 1b,c respectively, were subjected to the in situ tensile straining experiments at a strain rate of about ∼5 × 10 −4 s −1 . The sample design, testing schemes and electron-beam-heating effects are discussed in the Methods section. Figure 2 shows a series of video frames presenting the typical behaviour of sample I during the in situ tension experiment. Interestingly, measurements of the lengths of the gauge section (marked by dashed horizontal white lines) indicate that the sample uniformly elongated, up to a strain as high as 15% (Fig. 2b). This is the point when the first sign was observed for non-uniform deformation starting at a location slightly above the middle of the gauge section. One shear band was initiated, and the shear offset became obvious at the stage shown in Fig. 2c, where the total elongation reached 24%. This strain increment (from 15 to 24%) is partly a result of the slow growth of the shear offset, without the rapid fracture common in conventional metallic-glass test samples 8 , and also of the preferential thinning of the material in the middle (Fig. 2c). This necked region (marked with an arrow in Fig. 2d) narrowed gradually and considerably, also contributing elongations without fracture, as shown in Fig. 2d,e. At the stage shown in Fig. 2e, the total tensile strain reached 45% (if we discount the strains non-uniformly concentrated in the thin neck in the middle, the rest of the gauge section experienced an elongation of 29%). We emphasize that the large strain here was not achieved through the formation of multiple shear bands 9–13,18–22 . Samples II and III both had some unevenness on the sample (side) surfaces after FIB cutting. These ‘notches’ in the virgin samples before testing served as stress concentrators and encouraged necking to start early during the in situ tensile straining experiments. Figure 3 shows a series of photos from the videotape, showing sample II at different straining stages. Necking was the dominant deformation mode (from Fig. 3b–e), starting from the pre-existing notch (marked with an arrow in Fig. 3a). The necked region had a very small volume, a complex stress state, and nature materials ADVANCE ONLINE PUBLICATION www.nature.com/naturematerials 1