•Discovery of SME in MGs•An entirely new mechanism is proposed for the SME in MGs•The origin of SME can be elucidated in the potential energy landscape framework Shape memory alloys have been widely used in aerospace, intelligent robots, medical devices, and automobiles due to their unique shape memory effect (SME). Their various functions rely on the reversible structural phase transformation from thermoelastic martensite. Here, we report the SME in monolithic metallic glasses (MGs) without the long-range periodic order, i.e., the pre-annealed MGs after shape elimination can be restored to their original shapes by reheating to the pre-annealing temperature. Also, the dynamics in pre-annealed MGs become retarded upon heating and reach the slowest at around the pre-annealing temperature, showing the non-monotonic relaxation behavior and low-energy state. Compared with martensitic phase change materials, shape memory MGs, in general, have higher strength, better corrosion resistance and biocompatibility, which are expected to achieve the unprecedented applications for realizing functional devices of broad interests. Shape memory effect (SME), mainly present in crystalline Ti-Ni alloys, is basically missing in metallic glasses (MGs) that lack the long-range periodic order of crystals. Here, we report experimental results of SME in annealed MGs, in which the low-energy configuration state recovery is observed by both differential scanning calorimetry and X-ray photon correlation spectroscopy. We elucidate the origin of SME in MGs under the potential energy landscape framework, i.e., after annealing, the energy of MGs enters into a deep basin and atoms are located in the low-energy configuration state. Albeit deviating from their relative stable configurations by temperature changes, atoms in the annealed MGs tend to return to the low-energy atomic configurations along certain trajectories as pre-annealing temperature is approached. These results could extend the application of MGs as functional materials by directionally manipulating their energy states via annealing and rejuvenation. Shape memory effect (SME), mainly present in crystalline Ti-Ni alloys, is basically missing in metallic glasses (MGs) that lack the long-range periodic order of crystals. Here, we report experimental results of SME in annealed MGs, in which the low-energy configuration state recovery is observed by both differential scanning calorimetry and X-ray photon correlation spectroscopy. We elucidate the origin of SME in MGs under the potential energy landscape framework, i.e., after annealing, the energy of MGs enters into a deep basin and atoms are located in the low-energy configuration state. Albeit deviating from their relative stable configurations by temperature changes, atoms in the annealed MGs tend to return to the low-energy atomic configurations along certain trajectories as pre-annealing temperature is approached. These results could extend the application of MGs as functional materials by directionally manipulating their energy states via annealing and rejuvenation. Shape memory effect (SME) is defined as the ability of a material to recover its original shape after certain treatments, such as in crystalline Ti-Ni alloy,1Gu H. Bumke L. Chluba C. Quandt E. James R.D. Phase engineering and supercompatibility of shape memory alloys.Mater. Today. 2018; 21: 265-277Crossref Scopus (75) Google Scholar, 2Huang X. Ackland G.J. Rabe K.M. Crystal structures and shape-memory behaviour of NiTi.Nat. Mater. 2003; 2: 307-311Crossref PubMed Scopus (272) Google Scholar, 3Tanaka Y. Himuro Y. Kainuma R. Sutou Y. Omori T. Ishida K. Ferrous polycrystalline shape-memory alloy showing huge superelasticity.Science. 2010; 327: 1488-1490Crossref PubMed Scopus (385) Google Scholar, 4Ahadi A. Sun Q. Stress-induced nanoscale phase transition in superelastic NiTi by in situ X-ray diffraction.Acta Mater. 2015; 90: 272-281Crossref Scopus (128) Google Scholar where SME is a result of reversible structural phase transformation from thermoelastic martensite.2Huang X. Ackland G.J. Rabe K.M. Crystal structures and shape-memory behaviour of NiTi.Nat. Mater. 2003; 2: 307-311Crossref PubMed Scopus (272) Google Scholar,5Ogawa Y. Ando D. Sutou Y. Koike J. A lightweight shape-memory magnesium alloy.Science. 2016; 353: 368-370Crossref PubMed Scopus (107) Google Scholar Consequently, such materials are widely applied in the robotics, aerospace, biomedical, and automotive industries. However, in metallic glass (MGs), SME is basically missing due to the lack of a long-range periodic order for characterizing crystals. In metallic glass matrix composites, which have been developed to overcome the low fracture toughness and catastrophic failure of MGs,6Hays C.C. Kim C.P. Johnson W.L. Microstructure controlled shear band pattern formation and enhanced plasticity of bulk metallic glasses containing in situ formed ductile phase dendrite dispersions.Phys. Rev. Lett. 2000; 84: 2901Crossref PubMed Scopus (1322) Google Scholar, 7Hofmann D.C. Suh J.Y. Wiest A. Duan G. Lind M.L. Demetriou M.D. Johnson W.L. Designing metallic glass matrix composites with high toughness and tensile ductility.Nature. 2008; 451: 1085-1089Crossref PubMed Scopus (1231) Google Scholar, 8Hofmann D.C. Shape memory bulk metallic glass composites.Science. 2010; 329: 1294-1295Crossref PubMed Scopus (172) Google Scholar SME was also mentioned due to the appearance of ductile B2 phase, such as in CuZr-based MG.9Pauly S. Liu G. Wang G. Das J. Kim K.B. Kühn U. Kim D.H. Eckert J. Modeling deformation behavior of Cu-Zr-Al bulk metallic glass matrix composites.Appl. Phys. Lett. 2009; 95: 101906Crossref Scopus (76) Google Scholar, 10Song K.K. Pauly S. Zhang Y. Gargarella P. Li R. Barekar N.S. Kühn U. Stoica M. Eckert J. Strategy for pinpointing the formation of B2 CuZr in metastable CuZr-based shape memory alloys.Acta Mater. 2011; 59: 6620-6630Crossref Scopus (113) Google Scholar, 11Şopu D. Albe K. Eckert J. Metallic glass nanolaminates with shape memory alloys.Acta Mater. 2018; 159: 344-351Crossref Scopus (31) Google Scholar Moreover, experimental results for some polymers suggest that SME may result from conformational changes in certain molecular chains.12Zhao Q. Qi H.J. Xie T. Recent progress in shape memory polymer: new behavior, enabling materials, and mechanistic understanding.Prog. Polym. Sci. 2015; 49: 79-120Crossref Scopus (888) Google Scholar, 13Hornat C.C. Yang Y. Urban M.W. Quantitative predictions of shape-memory effects in polymers.Adv. Mater. 2017; 29: 1603334Crossref Scopus (48) Google Scholar, 14Behl M. Lendlein A. Shape-memory polymers.Mater. Today. 2007; 10: 20-28Crossref Scopus (1009) Google Scholar It implies that even though without a B2 phase, monolithic MGs could also exhibit SME if a certain preferential low-energy structure is achieved, especially at specific temperatures below Tg (the glass transition temperature). As temperature changes, this low-energy state could be remembered by atoms once the original temperature at which it developed is reached again, possibly inducing a macroscopic shape recovery in MGs. To confirm this scenario and to avoid oxidation at high temperatures, Au- and Pd-based MGs were investigated (see experimental procedures for details) with the following experiments designed. (1) The as-prepared ribbon was coiled on a rod and annealed at 0.9Tg (Tg ∼ 358 K at 20 K/min for as-prepared Au55Cu25Si20 MG) for 3 days under vacuum to realize a spring-like shape, as shown in Figure 1A. This state is defined as “I.” (2) The spring-shaped sample was straightened and fixed on a steel sheet at both ends to prevent it from retracting (Figure 1B) at room temperature (RT = 303 K). It was reheated from RT to a near below Tg under vacuum, and then quickly cooled down to RT again. The free-standing sample state after removing any restrictions is defined as “II” and hereafter also called the flash-annealed sample. (3) The flash-annealed sample (II) was vacuum annealed at 0.9Tg for another 3 days. The final state is defined as “III.” Figures 1C and 1D show three stages of samples undergoing shape fixation (I), shape elimination (II), and shape recovery (III) for Au55Cu25Si20 and Pd40Ni40P20 MGs, respectively. In addition, we bent the Au55Cu20Ag5Si20 MG into the “ZJU” logo (Figure S1) and repeated the above three processes, as shown in Figure 1F. The original shape of the logo was restored to some extent after 0.9Tg re-annealing. Moreover, we found that this SME also occurred in bulk MG samples. Figure 1E shows the shape fixation (I), shape elimination (II), and shape recovery (III) for an Au-based (Au60Cu15.5Ag7.5Si17) bulk MG15Zhang W. Guo H. Chen M.W. Saotome Y. Qin C.L. Inoue A. New Au-based bulk glassy alloys with ultralow glass transition temperature.Scr. Mater. 2009; 61: 744-747Crossref Scopus (55) Google Scholar with dimensions of about 14 × 5 × 1 mm3. It underwent bending at 0.9Tg (Tg ∼ 368 K at 20 K/min) for 3 days to form a curved shape with an angle of about 13.4°, then was straightened by annealing under stress at 363 K for 20 min to get state “II,” and finally free recovered at 0.9Tg for another 3 days to realize state “III.” The restoration of bending angle indicates the occurrence of shape recovery. All observed results in Figure 1 indeed reveal the existence of the SME in MGs, independent of their initial shapes and compositions. Considering the contribution of anelastic strain recovery16Zener C.M. Siegel S. Elasticity and anelasticity of metals.J. Phys. Chem. 1949; 53: 1468Crossref Google Scholar,17Ju J. Jang D. Nwankpa A. Atzmon M. An atomically quantized hierarchy of shear transformation zones in a metallic glass.J. Appl. Phys. 2011; 109: 053522Crossref Scopus (84) Google Scholar to the SME, we conducted time- and temperature-dependent sample length changes for a flash-annealed sample (stage II shape elimination) in Figures 2 and S2. It was observed that the shape recovery of the flash-annealed sample was quite limited when it was held at 303 K for 1 day and 313 K for a further day. However, when the temperature increased to 323 K, the pre-annealing temperature Ta for the Au55Cu25Si20 MG, the whole sample length could be almost recovered within 5 h. While with the temperature continually increasing to 328 K, the sample length hardly changes. These results indicate that, within the experimental timescale, anelastic strain recovery at low temperatures is very limited and can be accelerated to achieve SME at near or above Ta. The differential scanning calorimetry (DSC) curves in Figure 2B show that the enthalpy overshoot just above Tg disappears in the flash-annealed MG, but returns after shape recovery at Ta or higher temperatures, suggesting that the local atomic configurations could be significantly altered from stage II to stage III, between which the energy barrier could also be large (see Figure 2C). Therefore, to achieve the SME, heating the sample to (or near) the Ta is required. Peering into the mechanism of SME in the amorphous state is of great challenge due to the unsolved complex amorphous structure. Without long-range periodic order, the origin for the SME in MGs might be different from that in crystalline materials. As shape recovery in MGs must involve atomic movements, tracing how atoms move in MGs could provide some hints to understand the observed SME in MGs. Owing to the recent development of the X-ray photon correlation spectroscopy (XPCS) technique, a high-flux and partially coherent X-ray beam can be used to monitor structural dynamics on interatomic or interparticle length scales in disordered systems.18Sutton M. Mochrie S.G.J. Greytak T. Nagler S.E. Berman L.E. Held G.A. Stephenson G.B. Observation of speckle by diffraction with coherent X-rays.Nature. 1991; 352: 608-610Crossref Scopus (321) Google Scholar When the coherent X-ray beam is scattered by disordered materials, the intensity fluctuation of speckles changing with time is recorded, by which the temporal structural rearrangements can be tracked, providing a new insight into atomic dynamic behaviors in MGs,19Mezei F. Knaak W. Farago B. Neutron spin echo study of dynamic correlations near liquid-glass transition.Phys. Rev. Lett. 1987; 58: 571-574Crossref PubMed Scopus (311) Google Scholar, 20Leitner M. Sepiol B. Stadler L.M. Pfau B. Vogl G. Atomic diffusion studied with coherent X-rays.Nat. Mater. 2009; 8: 717-720Crossref PubMed Scopus (81) Google Scholar, 21Ruta B. Chushkin Y. Monaco G. Cipelletti L. Giordano V.M. Pineda E. Bruna P. Relaxation dynamics and aging in structural glasses.AIP Conf. Proc. 2013; 1518: 181-188Crossref Scopus (19) Google Scholar, 22Leitner M. Sepiol B. Stadler L.M. Pfau B. Time-resolved study of the crystallization dynamics in a metallic glass.Phys. Rev. B. 2012; 86: 064202Crossref Scopus (26) Google Scholar especially the process of shape recovery on the atomic scale. Although several experimental investigations have shown that the memory effect in MGs results from the fluctuations in enthalpy change23Li M.X. Luo P. Sun Y.T. Wen P. Bai H.Y. Liu Y.H. Wang W.H. Significantly enhanced memory effect in metallic glass by multistep training.Phys. Rev. B. 2017; 96: 174204Crossref Scopus (6) Google Scholar,24Luo P. Li Y.Z. Bai H.Y. Wen P. Wang W.H. Memory effect manifested by a boson peak in metallic glass.Phys. Rev. Lett. 2016; 116: 175901Crossref PubMed Scopus (39) Google Scholar and a large value of activation entropy,25Song L. Xu W. Huo J. Li F. Wang L.M. Ediger M.D. Wang J.Q. Activation entropy as a key factor controlling the memory effect in glasses.Phys. Rev. Lett. 2020; 125: 135501Crossref PubMed Scopus (15) Google Scholar no macroscopically obvious SME has ever been observed in monolithic MGs and detected from a dynamics perspective. Here, we take Au-based MGs as an example to investigate the origin of the SME in MGs. Figure 3A schematically illustrates the experimental setup for XPCS in reflectivity geometry (see experimental procedures for details), and Figure 3B shows the temperature dependence of the intensity autocorrelation function, g2(q,t)below Tg for Au55Cu25Si20 MGs with different thermal histories. For the as-prepared sample,26Xu T. Wang X.D. Dufresne E.M. Ren Y. Cao Q. Zhang D. Jiang J.Z. Anomalous fast atomic dynamics in bulk metallic glasses.Mater. Today Phys. 2021; 17: 100351Crossref Scopus (3) Google Scholar the decay curves initially shift to the short-time direction with increasing temperature, indicating temperature-induced fast dynamics. This trend reverses at 318 K and the shape of the g2(q,t) curve at 323 K shows a remarkable change. From 323 to 328 K, the dynamics become even slower. When the temperature reaches 333 K, the curves start to move to short time again, i.e., atomic dynamics become faster with increasing temperature. Surprisingly, the 3-day 323 K annealed MG (similar to the sample at stage I in Figure 1C) exhibits different dynamic behaviors. Pre-annealing treatment makes the dynamics of the sample slow down at 303 K compared with the as-prepared MG. Unexpectedly, dynamics become even slower as the temperature increases, reaching the slowest level at around the corresponding annealing temperature (Ta = 323 K). When the temperature goes above Ta, the dynamics become faster. To confirm this abnormal phenomenon in the 3-day 323 K annealed MG upon heating, we measured another 3-day 318 K annealed MG and found that the whole process is reproducible, i.e., the slowest dynamics indeed appear near 318 K, as shown in the bottom diagram of Figure 3B. Generally, atomic dynamics in MGs strongly depend on the competition between thermal activation (favoring fast dynamics) and free volume annihilation (favoring slow dynamics). Temperature-dependent relaxation times obtained by fitting g2(q,t) curves using the Kohlrausch-Williams-Watts (KWW) equation for as-prepared and annealed Au55Cu25Si20 MGs are plotted in Figure 3C. In the as-prepared MG, fast dynamics at low temperatures are linked with the existence of excess free volume, causing the atoms to move easily under thermal activation. When the temperature increases above 318 K, excess free volume annihilation in the as-prepared MG becomes significant, resulting in slow dynamics. Thermal activation becomes dominant again at above 328 K, whereby the relaxation time shortens. In contrast, abnormal dynamical behavior occurs for annealed MGs (annealed at 323 or 318 K for 3 days), i.e., relaxation time does not decrease but gradually increases with temperature rising from RT to 323 or 318 K. In addition, due to less free volume, overall structural relaxation time is several times longer than that for the as-prepared sample. A unique characteristic peak appears near Ta, as shown in Figure 3C. By means of two-timecorrelation functions (TTCFs), dynamic evolutions can be visualized directly, as illustrated in Figures 4A and 4B for as-prepared MGs and 3-day 323 K annealed Au55Cu25Si20 MGs, respectively. For as-prepared MGs, in which aging dynamics are governed by a fast and stationary regime at the initial stage, showing a constant and narrow diagonal at 318 K. On the contrary, the width of the intensity profile along the main diagonal for the 3-day 323 K annealed Au55Cu25Si20 MG gradually becomes broader, and reaches a maximum at 323 K, corresponding to the slowest dynamics. For the annealed Au55Cu20Ag5Si20 MG, similar XPCS results (see Figures 4C and S3) further confirm the existence of the similar essence in dynamics. For further revealing the structural information behind SME and different dynamic behaviors affected by thermal histories, we carried out high-energy synchrotron radiation X-ray diffraction (XRD) and DSC measurements for the as-prepared and 3-day 323 K annealed Au55Cu25Si20 MGs. Conventional XRD patterns for all studied ribbon samples are given in Figure S4. Apparently, both MGs are fully amorphous without significant differences in structure factor S(q) (Figure S5A) and pair distribution function G(r) (Figure S5B). By enlarging the first peak of S(q) and G(r), the annealed MG shows sharper peaks with higher intensities, and a slight peak shift toward larger q and lower r values compared with the as-prepared one, suggesting that pre-annealing treatment slightly favors ordering and close packing in MGs. Onset crystallization temperature (Tx) and melting temperature (Tm) are almost identical for both MGs as shown by the DSC curves in Figure S5C, whereas Tg slightly increases after pre-annealing. However, the treatment of 3-day annealing at 323 K creates a pronounced endothermic peak at around Tg, as observed in Figure 2B, resulting from the enthalpy recovery of relatively low-energy atomic configurations.27Van den Beukel A. Sietsma J. The glass transition as a free volume related kinetic phenomenon.Acta Metall. Mater. 1990; 38: 383-389Crossref Scopus (407) Google Scholar As a result, atoms after sub-Tg annealing for 3 days could be trapped in a deep valley in the potential energy landscape (PEL).28Fukuhara M. Inoue A. Nishiyama N. Rubberlike entropy elasticity of a glassy alloy.Appl. Phys. Lett. 2006; 89: 101903Crossref Scopus (13) Google Scholar, 29Debenedetti P.G. Stillinger F.H. Supercooled liquids and the glass transition.Nature. 2001; 410: 259-267Crossref PubMed Scopus (3464) Google Scholar, 30Saika-Voivod I. Poole P.H. Sciortino F. Fragile-to-strong transition and polyamorphism in the energy landscape of liquid silica.Nature. 2001; 412: 514-517Crossref PubMed Scopus (344) Google Scholar These low-energy atomic configurations in 3-day annealed Au55Cu25Si20 MGs can be erased by heating the pre-annealed MGs to a supercooled liquid region at 358 K at 2 K/min (above Tg, hereafter, termed the 358 K erased sample), as illustrated in Figure 3C. XPCS measurements of the 358 K erased sample reveal a nearly continuous decline in relaxation time upon heating, while the abnormal sluggish dynamical behaviors at the pre-annealing temperature disappear. The 358 K erased sample exhibits the longest relaxation time at 303 K compared with other MGs, which is likely attributed to the annihilation of excess free volume during the slow cooling process with a cooling rate of about 2 K/min. To understand SME in MGs (Figures 1 and 2) and this novel dynamic behavior in a general scenario, we adopted the PEL concept,31Doliwa B. Heuer A. What does the potential energy landscape tell us about the dynamics of supercooled liquids and glasses?.Phys. Rev. Lett. 2003; 91: 235501Crossref PubMed Scopus (93) Google Scholar,32Heuer A. Exploring the potential energy landscape of glass-forming systems: from inherent structures via metabasins to macroscopic transport.J. Phys. Condens. Matter. 2008; 20: 373101Crossref PubMed Scopus (327) Google Scholar which has been widely applied to investigate many physical phenomena, such as glass transition,33Niblett S.P. de Souza V.K. Jack R.L. Wales D.J. Effects of random pinning on the potential energy landscape of a supercooled liquid.J. Chem. Phys. 2018; 149: 114503Crossref PubMed Scopus (7) Google Scholar,34Blank-Burian M. Heuer A. Shearing small glass-forming systems: a potential energy landscape perspective.Phys. Rev. E. 2018; 98: 033002Crossref Scopus (10) Google Scholar liquid-to-liquid transition,35Wei S. Yang F. Bednarcik J. Kaban I. Shuleshova O. Meyer A. Busch R. Liquid-liquid transition in a strong bulk metallic glass-forming liquid.Nat. Commun. 2013; 4: 2083Crossref PubMed Scopus (134) Google Scholar,36Lan S. Ren Y. Wei X.Y. Wang B. Gilbert E.P. Shibayama T. Watanabe S. Ohnuma M. Wang X.L. Hidden amorphous phase and reentrant supercooled liquid in Pd-Ni-P metallic glasses.Nat. Commun. 2017; 8: 14679Crossref PubMed Scopus (79) Google Scholar and crystallization,37Richard D. Speck T. Crystallization of hard spheres revisited. I. Extracting kinetics and free energy landscape from forward flux sampling.J. Chem. Phys. 2018; 148: 124110Crossref PubMed Scopus (14) Google Scholar as well as mechanical deformation38Harmon J.S. Demetriou M.D. Johnson W.L. Samwer K. Anelastic to plastic transition in metallic glass-forming liquids.Phys. Rev. Lett. 2007; 99: 135502Crossref PubMed Scopus (199) Google Scholar and physical aging of MGs.39Lüttich M. Giordano V.M. Le Floch S. Pineda E. Zontone F. Luo Y. Samwer K. Ruta B. Anti-aging in ultrastable metallic glasses.Phys. Rev. Lett. 2018; 120: 135504Crossref PubMed Scopus (35) Google Scholar,40Ketov S.V. Sun Y.H. Nachum S. Lu Z. Checchi A. Beraldin A.R. Bai H.Y. Wang W.H. Louzguine-Luzgin D.V. Carpenter M.A. Greer A.L. Rejuvenation of metallic glasses by non-affine thermal strain.Nature. 2015; 524: 200-203Crossref PubMed Scopus (436) Google Scholar Since the as-prepared MG freezes rapidly from its melt, its energy level is most likely located at a relatively high-energy basin. Upon heating, atoms relax relatively easily because of more free volume in their surroundings. At a certain temperature, atoms could cross the energy barrier to trigger the annihilation of excess free volume at a certain temperature, resulting in slightly slow dynamics. When the temperature continues to increase to Tg, cooperative α-relaxation is activated, leading to the fast dynamics. However, a 3-day sub-Tg annealing treatment can relax atoms in MGs substantially and annihilate most of free volume, causing atoms to reach their quasi-equilibrium sites. In other words, the energy in the annealed MG system is most likely located at a local deep minimum in PEL, as shown in Figure 5. When the annealed MG is cooled to RT or flash heated to a near below Tg temperature and then quickly cooled down to RT, atoms could be slightly or largely deviated from their quasi-equilibrium positions. During reheating, atoms potentially move back to their relatively low-energy quasi-equilibrium positions in configurations when pre-annealing temperature is reached. In the low-energy valley, the atomic motions are largely trapped thus the relaxation time increases, showing the slowest dynamics near pre-annealing temperature. As a result, on a macroscopic level, the lengths of spring-shaped MG samples (stage III) are recovered. The enthalpy at each step of shape change was tracked by DSC for all the studied MGs (Figures 2 and S6), through which one can discriminate the disappearance and recovery of the endothermic peak near above Tg, corresponding to the damage and recovery of local configurations. These results reveal that macroscopic SME, accompanied with local atomic configuration recovery, indeed occurs in annealed MGs. Besides, it was found that, once the annealed MG (stage I) has been heated to a temperature above Tg, the extended sample almost forgets its original shape during re-annealing at Ta (erased sample). Accordingly, the endothermic peak near above Tg for enthalpy recovery becomes unpronounced for the erased sample, which is similar to the DSC curve for the as-prepared MG (Figure S7). In polymers, the SME was reported a few years ago, which mainly depends on the movement of molecular chains.12Zhao Q. Qi H.J. Xie T. Recent progress in shape memory polymer: new behavior, enabling materials, and mechanistic understanding.Prog. Polym. Sci. 2015; 49: 79-120Crossref Scopus (888) Google Scholar,13Hornat C.C. Yang Y. Urban M.W. Quantitative predictions of shape-memory effects in polymers.Adv. Mater. 2017; 29: 1603334Crossref Scopus (48) Google Scholar Above transition temperature (Tt), molecular chains are active to move under external force, forming a preferential orientation. When cooled below Tt, they “freeze” to retain a temporary shape even though the force has been removed. If the temperature goes above Tt again, the mobility of molecular chains is reactivated, and they spontaneously revert to their original state. Analogously, SME in MGs is also based on two states. A permanent state is created by long-time sub-Tg (Ta) pre-annealing to position atoms into a low-energy atomic configuration state in the PEL. With changing temperature or external force, atoms could deviate from their low-energy positions, and thus form a temporary shape. When the temperature increases to Ta again, atoms tend to return to their initial low-energy states. Basically, SME in polymers needs to occur above the glass transition temperature (Tg), melting temperature (Tm), or less commonly the liquid crystal clearing temperature (Tcl),12Zhao Q. Qi H.J. Xie T. Recent progress in shape memory polymer: new behavior, enabling materials, and mechanistic understanding.Prog. Polym. Sci. 2015; 49: 79-120Crossref Scopus (888) Google Scholar whereas SME in MGs reported here appears below Tg without phase transformation (Figure S8). Naturally, the discovery of SME in MGs has inspired their potential applications. For example, the temperature at which SME appears in Au-based MGs is close to that of the human body, making them suitable for everyday objects, such as spectacle frames that could automatically restore their original shape after use. Besides, a shape memory alloy bath temperature control valve could intelligently suspend hot water supply once water temperature has exceeded the human body temperature to avoid scalding. Moreover, SME detected in the widely studied Zr-based MG41Li Y.H. Zhang W. Dong C. Qiang J.B. Yubuta K. Makino A. Inoue A. Unusual compressive plasticity of a centimeter-diameter Zr-based bulk metallic glass with high Zr content.J. Alloy. Compd. 2010; 504: S2-S5Crossref Scopus (36) Google Scholar without any noble metals (Figure S9) and the multiple repeatability of SME (Figure S10) make shape memory MGs promising materials for engineering applications. In general, shape memory MGs could possess features, such as higher strength, elastic limit, and corrosion resistance, than those of crystalline material-based shape memory alloys, prospectively having longer service life and less damage to human tissues when used as biodegradable implants.42Zberg B. Uggowitzer P.J. Loffler J.F. MgZnCa glasses without clinically observable hydrogen evolution for biodegradable implants.Nat. Mater. 2009; 8: 887Crossref PubMed Scopus (711) Google Scholar In summary, all results obtained here demonstrate the existence of SME in monolithic MGs. After long-time annealing treatments below the glass transition temperature, MGs can be restored to their original shapes by heating to the pre-annealing temperatures again. We elucidate the origin of the SME in MGs, i.e., upon heating atoms in the pre-annealed MGs can return to their low-energy atomic configurations at around pre-annealing temperature along certain trajectories, exhibiting SME on a macro scale. Moreover, the dynamics in pre-annealed MGs get slow upon heating and reach the slowest level at around the pre-annealing temperature detected by advanced XPCS. This phenomenon is extremely exciting because the SME in monolithic MGs has never been reported before, which differs from the martensitic transformation for crystalline shape memory alloys, and also differs from the glass transition or melting for non-metallic shape memory polymers (SMPs). Compared with martensitic phase change materials and SMPs, shape memory MGs, in general, have higher strength, better corrosion resistance, and are non-toxic, which can be used as profitable engineering and biomedical materials. These findings reported here can provide a comprehensive understanding of SME, which could be expected to lead to unprecedented applications of MGs and even other disordered materials for realizing functional devices of broad interests. With the development of artificial intelligence, shape memory materials will receive extensive attention.