Cyclic transformation between nanocrystalline and amorphous phases in Zr based intermetallic alloys during ARB

Transcription

1 Intermetallics 15 (2007) 644e651 Cyclic transformation between nanocrystalline and amorphous phases in Zr based intermetallic alloys during ARB P.J. Hsieh a,b, Y.C. Lo a, C.T. Wang c, J.C. Huang a, *, S.P. Ju c a Institute of Materials Science and Engineering; Center for Nanoscience and Nanotechnology, National Sun Yat-Sen University, Kaohsiung 804, Taiwan, ROC b Department of Materials Science and Engineering, I-Shou University, Kaohsiung 840, Taiwan, ROC c Department of Mechanical and Electro-Mechanical Engineering, National Sun Yat-Sen University, Kaohsiung 804, Taiwan, ROC Available online 17 November 2006 Abstract By applying a simple room temperature accumulative roll bonding (ARB) process, the multilayered Zr/Ni and Zr/Ti elemental foils can be mixed and transformed into nanocrystalline and eventually amorphous bulk materials. The latest stage of the transformation from the nanocrystalline elemental phase measuring w2 nm to the amorphous alloying phase via gradual atomic mixing is examined, using high resolution transmission electron microscopy, Fourier transformation, and molecular dynamics simulation. Closer examination is focused on the following evolution during the subsequent ARB cycles after amorphization. Transformation from the amorphous phase back to the nanocrystalline elemental phase occurs during the next 3 ARB cycles. Upon further ARB cycles, the nanocrystalline Ni phase is again transformed into the complete amorphous state, exhibiting the cyclic transformation behavior between the nanocrystalline and amorphous phases. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: A. Nanostructured intermetallics; B. Glasses, metallic; C. Rolling; E. Simulations, atomistic; F. Electron microscopy, transmission 1. Introduction Severe plastic deformation processes, such as friction stir processing [1], mechanical alloying (MA) [2,3] and accumulative roll bonding (ARB) [4,5], have received considerable attention in recent years. High plastic strains and ultrafine or even amorphous structures with outstanding properties can be induced. In this study, the ARB process is of interest since this method could directly produce the nanocrystalline or amorphous phase in a plate form without the limitation of cooling rates. The MA and ARB methods belong to the same category of solid-state reaction and both methods force the atoms of adopted elements to diffuse in a solid state at low temperatures in order to obtain the metastable amorphous phases. According to previous reports, the initial microstructural evolution of the ARB process has some similarities with the early milling stage of the MA method [6]. But unlike * Corresponding author. Tel.: þ x4063; fax: þ address: jacobc@mail.nsysu.edu.tw (J.C. Huang). MA, the complete amorphization mechanisms during ARB, especially at the latest stage, have not been well developed. Hence, some of the ZreCu [3,6] or ZreNi [7] based alloys previously studied using the MA route were examined in our previous reports [8e12] using the ARB process in order to establish the graduate vitrification mechanisms during room temperature ARB. It reveals that the grain sizes of Zr based specimens are gradually refined from the micrometer scale down to a few nanometers with increasing repeated folding and rolling (F&R) steps, and the nanocrystalline phase retained is dispersed in the amorphous matrix. For the latest stage, sudden transformation of the crystalline phase into fully amorphous state was observed to proceed within a few ARB cycles. In this study, the high resolution transmission electron microscopy (HRTEM) and the related Fourier transformation software are used to cautiously characterize the cyclic transformation between the nanocrystalline and amorphous phases in the binary Zr based alloys. Recent efforts in molecular dynamical (MD) simulation studies on binary metallic systems [13] also reveal that the rapid /$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi: /j.intermet

2 P.J. Hsieh et al. / Intermetallics 15 (2007) 644e nanocrystalline to amorphous transition occurred by means of only 4 cycles for a layered structure with 3 nm thickness. It is recognized that the atomistic simulation of the MD method provides a powerful tool to better investigate the atomic mutation of nanocrystalline phases measuring below 2 nm. The simulated results can be compared with the HRTEM images of the experimental ARB specimens. The many-body tight-binding potentials for the selected elements are used in the MD simulation. Meanwhile, the periodic boundary condition and volume change during repeating ARB cycles are not considered during the simulation of the atomic mixing behavior. 2. Experiment and simulation details In the ARB experiment, stacked foils of Zr and Ni (or Zr and Ti) of equivalent amount are repeatedly roll bonded at room temperature. The pure elemental foils, 80e100 mm in thickness, were first cut into pieces measuring 20 mm in width and 120 mm in length. About 10e13 layers of selected elemental foils were stacked in order to yield a multilayer specimen with total thickness near 1 mm. The stacked specimens were rolled by a 350-ton rolling machine (with the roller diameter and length of 140 and 220 mm, respectively) at a mean strain rate in the range of 0.4e0.8 s 1 [14]. The rolling speed for each F&R cycle is 30 mm/s and the processing time for each pass is around 1e2 s. The thickness reduction for each F&R cycle is set to about half of the initial value. Small parts of the specimens were sampled after various cycles for off line analyses. The layer (and grain size) refinement as well as the vitrification condition of the alloy structures were examined by scanning electron microscopy (SEM) and HRTEM with energy-dispersive X-ray spectrometry (EDS) with a beam size of 10 nm. Image intensifier and CCD camera coupled with TV system were used for the observation of interfaces between the nanocrystalline and amorphous phases. Meanwhile, a bi-layered structure consisting of Zr and Ni (or Zr and Ti) elemental layers, 20 nm in width, 20 nm in length, and 12 nm in thickness (for 2 layers), was set to be the starting alloy model in the MD simulation in order to trace the final cyclic transformation stage between nanocrystalline and amorphous phases during ARB. The processing ARB speed in the MD simulation is 0.05 nm/fs. The present simulation employs the Verlet algorithm [15] to calculate the trajectories of the atoms and the scaling method is adopted during the simulation to control the system temperature (at room temperature). Atomic interactions, among ZreZr, NieNi (TieTi), and ZreNi (ZreTi), should be all included in the consideration of the building of simulation model. Therefore, the many-body, tight-binding potential is adopted to model all of the atomic interactions as follows [16e18]: ( X ) 1=2 E i ¼ x 2 rij exp 2q 1 r j 0 þ X rij Aexp p 1 ; ð1þ r j 0 where x is an effective hopping integral, r ij is the distance between atoms i and j, and r 0 is the first-neighbor distance. The total band energy is characterized by the second moment of the d-band density of state and is shown in the first part of potential function. Meanwhile, the second part reveals the repulsive energy of the tight-binding potential [16]. The parameters A, p, q and x are determined by the experimentally obtained values of cohesive energy, lattice parameter, bulk modulus and 2 shear elastic constants (C 44 and C 0 ¼ 2(C 11 C 12 ), respectively). The interaction force on atom i is expressed as: F i ¼ X vei þ ve j rij : ð2þ vr jsi ij vr ij r ij The parameters of the tight-binding potential for Ni (Ti) and Zr [16e18] in the MD simulation are listed in Table 1. For the tight-binding potential used in this study, there are 5 parameters to be determined. Considering the parameters for the same element, for example, Zr or Ni, different researchers may obtain different sets of these 5 parameters, but these different parameters in tight-binding potentials can sufficiently model the same element [19]. In addition, the parameters of Ni and Zr in Ref. [18] are newer than those in Refs. [16] and [17], and using the newer parameters in tight-binding potentials can predict the more accurate cohesive energy. In this study, we decide to adopt the newer parameters of tight-binding potentials in Ref. [18] for pure Ni and Zr, and the cross-element parameters for NieZr from Refs. [16] and [17]. The variations of the cohesive energies, the radial distribution functions (RDF) and the average bond lengths are used to characterize the cyclic transition behavior between the nanocrystalline and amorphous phases. 3. Results and discussions 3.1. Transformation between nanocrystalline and amorphous phases As presented previously [8e12], with increasing F&R cycles, the accumulated shear stress forces the harder phases to break into smaller phases and finally into a much more homogeneous and mixed structure. The foils in the Zr 50 Ti 50 system with higher and more compatible hardness would undergo more efficient mutual deformation by the counterpart. Hence, the Zr 50 Ti 50 alloy was found to be more efficiently vitrified, as compared to Zr 50 Ni 50. After 80e100 F&R cycles, only a few nanocrystalline phases with nearly spherical shape and average size w2 nm remain and are dispersed in the dominant Table 1 Parameters used in tight-binding potential function Parameters A (ev) x (ev) p q r 0 (Å) Zr Ni Ti ZreNi ZreTi

3 646 P.J. Hsieh et al. / Intermetallics 15 (2007) 644e651 Fig. 1. (a) HRTEM micrograph of Zr 50 Ni 50 after 80 ARB cycles, and the enlarged lattice images of the (b) Ni (in Zr 50 Ni 50 ), (c) Zr (in Zr 50 Ni 50 ), and (d) Ti (in Zr 50 Ti 50 ) nano-particles along with their associated Fourier transform and reconstructed images. amorphous matrix [11]. The residual nanocrystalline phases seem to become unstable with the further ARB passes and sudden transformation from the nanocrystalline to fully amorphous state occurred upon subjecting to a few more ARB cycles. It has been shown previously [12] using XRD and TEM diffraction that all the remaining nanocrystalline phases are unmixed elemental particles of pure Zr, Ni or Ti, i.e., no intermetallic compound was formed during the room temperature ARB process. Fig. 1 shows the examples of the HRTEM lattice images of Zr 50 Ni 50 or Zr 50 Ti 50 after 80 ARB cycles. In both alloy systems, the lattice images reveal that the elemental nanocrystalline phases are surrounded by the amorphous matrix. The smallest nanocrystalline size is w2 nm; the critical size of

4 P.J. Hsieh et al. / Intermetallics 15 (2007) 644e Fig. 2. The microstructural evolution and the associated two-dimensional Fourier transform of the bi-layered Zr 50 Ni 50 model subjected to various ARB cycles: (a) initial state; (b) 1, (c) 4, and (d) 6 F&R cycles. w2 nm appears to be the lower boundary for the stability of nanocrystalline phases in the current system. The local diffraction patterns for such nanocrystalline particles measuring w2 nm, obtained by the Fourier transformation as well as the reconstructed lattice images constructed by a second Fourier transformation after filtering the noise background, for the pure Ni, pure Zr, and pure Ti particles in such specimens are shown in Fig. 1(b)e(d). Over the final stage of ARB from 85 to 120 cycles, it was often observed that some of the crystalline characteristic peaks reappear after full vitrification, in other words, the fully amorphous phase would transform back to nanocrystalline pure elemental phase upon subsequent ARB passes, exhibiting cyclic transformation behavior. This was particularly apparent for the Ni phase with the FCC crystal structure in the Zr 50 Ni 50 system. The (111) Ni XRD peak was often the last peak to disappear (into fully amorphous state) and the first one to reappear (into nanocrystalline phase again). Parallel studies on the cyclic transformation are conducted using the nano-indentation on the fully amorphous Zr 50 Ni 50 alloys. Stress induced nanocrystallization near the indented region can also be seen MD simulation of microstructure evolution during ARB Figs. 2 and 3 reveal the MD simulated microstructural evolution over the final ARB stage and the associated two-dimensional Fourier transformation of the bi-layered Zr 50 Ni 50 and Zr 50 Ti 50 subjected to various ARB cycles. In the Zr 50 Ni 50 and Zr 50 Ti 50 systems, the nanocystalline structure is fully vitrified after 6 and 4 F&R cycles, respectively. A faster amorphization process is revealed in Zr 50 Ti 50, consistent with the experimental findings.

5 648 P.J. Hsieh et al. / Intermetallics 15 (2007) 644e651 Fig. 3. The microstructural evolution and the associated two-dimensional Fourier transform of the bi-layered Zr 50 Ti 50 model subjected to various ARB cycles: (a) initial state, (b) 1, (c) 4, and (d) 6 F&R cycles. Meanwhile, the variations of the MD simulated alloy potential energy (representing the atomic interaction enthalpy energy) with increasing F&R cycles are shown in Fig. 4. In Fig. 4(a), the system potential energy for Zr 50 Ni 50 starts to decline, or becomes more stable with unlike atom mixing, from the 3rd F&R cycle due to the severe structure change during the transition period and becomes saturated at the 6th cycle. This result is consistent with the strongly negative mixing enthalpy of ZreNi (DH m ¼ 49 kj/mol). The decreasing potential enthalpy energy leads to the lower Gibbs free energy of the mixed amorphous phase. In contrast, the potential energy variation for Zr 50 Ti 50 in Fig. 4(b) does not show the similar trend. No obvious variation of potential energy is observed in Zr 50 Ti 50 because of the near zero mixing enthalpy for Zr and Ti atoms and the unique characteristic of the complete dissolubility. It appears that the faster mixing of unlike atoms in Zr 50 Ti 50 is not a result of the potential or mixing enthalpy, but is due to the same HCP structure and the compatible initial hardness for Zr and Ti. The (0002) basal planes of both Zr and Ti lying on the rolling plane could be effectively sheared through with each other, accelerating the thickness reduction and mutual atomic mixing Cyclic transformation between nanocrystalline and amorphous phases It has been shown that it needs only 4 or 6 cycles for the current bi-layered structure, with 5e6 nm in each layer thickness, to fully transform into an amorphous state, similar to the simulated results of Lund and Schuh [13]. However, the phase stability [20] of the accumulative roll bonded amorphous alloys with further F&R passes has not been discussed and is of concern. Thus, the MD simulation continues to model the structure until 20 rolling cycles. In the Zr 50 Ti 50 system, the amorphous structure, though with very slight fluctuation in the radial distribution functions, was basically maintained

6 P.J. Hsieh et al. / Intermetallics 15 (2007) 644e (a) Average potential energy (ev) (b) Average potential energy (ev) F&R cycles F&R cycles Fig. 4. The variation of potential energy of (a) Zr 50 Ni 50 and (b) Zr 50 Ti 50 alloys subjected to different F&R cycles. till the end of all rolling processes. In contrast, the fully amorphous structure in the Zr 50 Ni 50 system remains over the 6th and 7th F&R cycles, and exhibits the first cyclic transformation into the nanocrystalline FCC Ni (with grain size of about 3 nm) at the 8th cycle, and is transformed back to fully amorphous from the 10th F&R cycle. Fig. 5(a) reveals the associated Fourier transform of the pure FCC Ni in the [111] zone axis. The Ni (111) plane texture tends to be stress induced and lie along the rolling plane. The re-crystallized Ni phase size is very close to the observed critical phase size of w2 nm before vitrification. And a phase with such a small size cannot maintain its crystalline structure upon further rolling cycles. Thus, the nano-sized FCC Ni re-states to collapse from the 9th F&R cycle and returns back to amorphous structure at the end of the 10th F&R cycle. The second minor cyclic transformation occurs again at the 12th F&R cycle. Fig. 6 reveals the variations of RDF and the relative packing density of Zr 50 Ni 50 alloy subjected to various F&R cycles. The RDF in Fig. 6(a) exhibits sharper peaks at the 8th and 9th cycles, marked by the circle, and the packing density is seen to increase as the appearance of nanocrystalline Ni phase at the 8th, 9th and 12th F&R cycles. The corresponding morphologies at the 8th and 9th cycles shown in Fig. 5(a) and (b) indicate that the crystalline structure, surrounded by circles, reappear during ARB cycles. Because the shape of the simulation system after the ARB cycle is not cubic, it is more convenient to choose 1 cube inside the simulation model after each cycle to calculate the packing density of this system. Consequently, the packing density is defined as the ratio of the total atom number inside this cube to the volume of this cube. The MD simulation was also performed on the atomic reconstruction during nano-indentation. Similar to the experimentally observed re-crystallization of the amorphous alloys under the applied load from the indenter, the MD simulated microstructures also reveal this trend, implying the stress induced nanocrystallization. Fig. 7 shows the local Fig. 5. The microstructural evolution and the associated two-dimensional Fourier transform of the bi-layered Zr 50 Ni 50 model subjected to (a) 8 and (b) 9 F&R cycles.

7 650 P.J. Hsieh et al. / Intermetallics 15 (2007) 644e651 (a) 50 (b) 40 RDF Relative density R (Å) F&R cycles Fig. 6. The variation of (a) RDF and (b) volume-based number density distribution of the Zr 50 Ni 50 alloy subjected to various F&R cycles. microstructures of Zr 50 Ni 50 near the indenter. The Ni atoms under loading tend to cluster more and form the FCC nanocrystalline phase. The cyclic phase transformation of the 2 nm nanocrystalline particles to the amorphous phase during ARB is considered to be caused by the strain or stress accumulation process, similar to the Zr 67 Cu 33 alloy made by MA method [3]. No thermal factor can be taken into account since the ARB is processed at room temperature. The thermodynamically related glass forming ability parameters, such as higher negative mixing enthalpy DH m and larger size mismatch Dr, do not seem to play the dominant roles for the vitrification during room temperature ARB. The faster vitrified Zr 50 Ti 50 possesses lower DH m and Dr (w0 kj/mol and 8%) as compared with Zr 50 Ni 50 ( 49 kj/mol and 23%). The controlling factors for solid-state amorphization during ARB are thus the relative hardness and crystal structure. 4. Conclusions The critical stage for the cyclic phase transformation between the nanocrystalline (w2 nm) and amorphous phases during room temperature ARB are examined by HRTEM and MD simulation. Both results obtained from the Zr 50 Ni 50 Fig. 7. MD simulated microstructures near the indenter at different stages. The Ni atoms tend to form nanocrystalline clusters under the applied load.

8 P.J. Hsieh et al. / Intermetallics 15 (2007) 644e and Zr 50 Ti 50 systems reveal that the nanocystalline structure would be quickly vitrified within a few cycles at the final ARB process. The Zr 50 Ti 50 would vitrify faster than Zr 50 Ni 50, due to the compatible hardness and HCP crystal structure. The thermodynamically related glass forming ability parameters, such as higher negative mixing enthalpy DH m and larger size mismatch Dr, do not seem to play the dominant roles for the vitrification during room temperature ARB. Cyclic transformation in Zr 50 Ni 50 is evident from the HRTEM and MD results, with the FCC Ni nanocrystalline phase disappearing and reappearing with increasing F&R cycles, as well as under the applying load during nano-indentation. Acknowledgement The authors gratefully acknowledge the sponsorship by National Science Council of Taiwan, ROC, under the project no. NSC E References [1] Mishra RS, Mahoney MW, McFadden SX, Mara NA, Mukherjee AK. Scripta Mater 2000;42:263. [2] Koch CC, Cavin OB, Mckamey CG, Scarbrough JO. Appl Phys Lett 1983;43:1017. [3] Sherif El-Eskandarany M, Inoue A. Metall Mater Trans 2002;A33:135. [4] Saito Y, Tsuji N, Utsunomiya H, Sakai T, Hong RG. Scripta Mater 1998;39:1221. [5] Sagel A, Sieber H, Fecht HJ, Perepzko JH. Acta Mater 1998;46:4233. [6] Sherif El-Eskandarany M, Inoue A. Metall Mater Trans 2002;A33:2145. [7] Eckert J, Schultz L, Hellstern E, Urban K. J Appl Phys 1998;64:3224. [8] Hsieh PJ, Huang JC, Hung YP, Cho SI, Jang JSC. Mater Chem Phys 2004;88:364. [9] Hsieh PJ, Hung YP, Huang JC. Scripta Mater 2003;49:173. [10] Hsieh PJ, Hung YP, Chou SI, Huang JC. Mater Trans JIM 2004;45: [11] Hsieh PJ, Huang JC, Jang JSC, Tsao CYA. J Metastable Nanocryst Mater 2005;24e25:351. [12] Hsieh PJ, Lo YC, Huang JC, Ju SP. Intermetallics 2005;14:924. [13] Lund AC, Schuh CA. Acta Mater 2004;52:2123. [14] Saito Y, Utsunomiya H, Tsuji N, Saka T. Acta Mater 1999;47:579. [15] Haile JM. Molecular dynamics simulation. Canada: John Wiley & Sons; [16] Massobrio C, Pontikis V, Martin G. Phys Rev Lett 1989;62:1142. [17] Massobrio C, Pontikis V, Martin G. Phys Rev B 1990;41: [18] Cleri F, Rosato V. Phys Rev B 1993;48:22. [19] Kallinteris GC, Papanicolaou NI, Evangelakis GA. Phys Rev B 1997; 55:2150. [20] Nagase T, Umakoshi Y. Scripta Mater 2003;48:1237.