Letter pubs.acs.org/NanoLett
Plastic Deformation through Dislocation Saturation in Ultrasmall Pt Nanocrystals and Its in Situ Atomistic Mechanisms Lihua Wang,†,‡ Jiao Teng,§ Xuechao Sha,† Jin Zou,*,‡,∥ Ze Zhang,†,⊥ and Xiaodong Han*,† †
Institute of Microstructure and Property of Advanced Materials, Beijing Key Lab of Microstructure and Property of Advanced Materials, Beijing University of Technology, Beijing, 100124, China ‡ Materials Engineering, ∥Centre for Microscopy and Microanalysis, The University of Queensland, Brisbane, Queensland 4072, Australia § Department of Material Physics and Chemistry, University of Science and Technology Beijing, Beijing 100083, China ⊥ Department of Materials Science, Zhejiang University, Hangzhou, 310008, China S Supporting Information *
ABSTRACT: The atomic-scale deformation dynamic behaviors of Pt nanocrystals with size of ∼18 nm were in situ investigated using our homemade device in a high-resolution transmission electron microscope. It was discovered that the plastic deformation of the nanosized single crystalline Pt commenced with dislocation “appreciation” first, then followed by a dislocation “saturation” phenomenon. The magnitude of strain plays a key role on dislocation behaviors. At the early to medium stage of deformation, the plastic deformation was controlled by the full dislocation activities accompanied by the formation of Lomer dislocation locks from reaction of full dislocations. When the strain increased to a significant level, stacking faults and extended dislocations as well as Lomer−Cottrell locks appeared. The Lomer−Cottrell locks can unlock through transferring into Lomer dislocation locks first, and then Lomer dislocation locks were destructed under high stresses. The very high density dislocations and the frequent dislocation reactions through Lomer dislocations and Lomer−Cottrell locks may lead to work hardening in nanosized Pt. KEYWORDS: In situ atomic scale, nanocrystals, dislocation saturation, Lomer−Cottrell locks, Lomer dislocations, work hardening
S
whether or not the above proposed dislocation starvation mechanism was valid has been unresolved yet. Furthermore, most previous studies were carried out on metals with low to medium stacking-fault (SF) energies, such as Au,21,23,28−31 Cu,25,31,32 and Ni.12 On the basis of these experimental observations, it has been proposed that their plastic deformations are dominant by partial dislocations when the crystal size of metals is sufficiently small, such as below ∼100 nm.25,28−30,33,34 These observations are consistent with molecular dynamics simulations,35−39 resulting in wide acceptance that, for nanosized metals, their deformations are governed by partial dislocations, while full dislocations are prohibited. However, for those metals with high SF energies, such as Pt, how dislocations behave in the single crystalline form has been uncertain. In situ atomic-scale experimental evidence is critical for the understanding of the deformation mechanisms of nanosized materials.40−43 In this study, we used our homemade device that allows the slow and gentle deformation of nanosized
mall-sized metallic nanocrystals have attracted a great deal of interest because of their ultrahigh strength1−3 and unusual deformation phenomena compared with their bulk counterparts.4−7 For those submicrometer sized metals, it has been proposed that “dislocation starvation” accounts for the ultrahigh strength.1,2,8 This is because, in small-sized metals, dislocations escape faster through free surfaces than they can multiply, leaving the small-sized metals free of dislocations. This proposed deformation mode has been confirmed by theoretical predictions9−11 and in situ transmission electron microscopy (TEM) investigations in Au, Cu, and Ni crystals with grain sizes over 200 nm.12−20 An extreme example was the in situ TEM tests showed direct evidence of “dislocation starvation” in ∼200 nm sized Ni pillars,12 where the dislocation escape rate was faster than their nucleation rate throughout the deformation. Similar phenomena have been observed in ∼200 nm sized Cu crystals3,14,18 as well as submicrometer sized Al crystals.15 Many further experimental studies were conducted to investigate the deformation behaviors of the nanosized metals.21−27 The dislocation starvation mode in nanocrystals were demonstrated for relative large samples, in which the surface strains/stresses could be neglected. Meanwhile, for those small single crystal samples with sizes below ∼20 nm, © XXXX American Chemical Society
Received: April 5, 2017 Revised: July 5, 2017 Published: July 17, 2017 A
DOI: 10.1021/acs.nanolett.7b01416 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters metals and at the same time retains the TEM double-tilt capability to perform in situ atomic-scale observations.44,45 We monitored the atomic-scale dynamic processes of the plastic deformation behaviors of single crystalline Pt nanocrystals with sizes of ∼18 nm. The in situ atomic-scale deformation experiments were realized in an HRTEM. Figure 1a is the schematic view of our homemade device with a Pt thin film (our TEM specimen) and shows the extensor
Figure 2. A series of HRTEM images show the tensile process of the Pt nanocrystal. During tensile loading, the dislocation density increased. Dislocation activities are different at different deformation stages.
be estimated by measuring the elongated length with the initial length of these two referenced atomic columns. In this experiment, the Pt nanocrystal experienced a homogeneous strain up to ∼55.2%. Figure 2a is an HRTEM image captured under the very early stage of the tensile loading. Two full dislocations were revealed (marked by arrows), from which the dislocation density can be estimated as ∼5 × 1015 m−2 (see dislocations calculation details in Figure S2). During further straining, two dislocations glided away from the view area, as confirmed in Figure 2b. At the meaning time, several new dislocations nucleated, but they glided away with continuous loading (when comparing Figure 2a−c). Interestingly, the dislocation underwent “appreciation” stage, in which the dislocation nucleation rate was faster than its escape rate, leading to an increased density of dislocations. In addition, at the early stage of deformation, in situ observations show that the plastic deformation is controlled by the activities of full dislocations (nucleation and escape), rather than well-expected partial dislocations. When the plastic strain reached ∼13.8%, partial dislocations associated SFs were found, as shown in Figure 2c. With further tensile loading, Figure 2d demonstrates the increase of the dislocation density (estimated at a level of ∼4.4 × 1016 m−2) which is one order higher than the initial stage with an increased density of partial dislocations. Interestingly, under the continual tensile loading, Figure 2d−f indicates that the dislocation density becomes stationary, although individual dislocations continually nucleated and escaped, suggesting a balanced dislocation nucleation rate versus its escape rate. Nevertheless, several dislocation types, such as Lomer dislocation (LD) locks and Lomer−Cottrell (L−C) locks, and their formations and destructions were observed in our in situ atomic scale experiments. Figure 3 shows an example of the full dislocation nucleation and escape process at the early deformation stage (strain < ∼7%). Figure 3a shows two full dislocations, marked as “1” and “2”. During the deformation (Figure 3b), these two dislocations escaped, but dislocation “3” nucleated. In the same manner,
Figure 1. (a) Schematic view shows the Pt thin film was attached on a bimetallic extensor. (b) The bimetallic extensor with the attached thin film was loaded on a conventional TEM heat specimen, which applied strain on the sample. (c) HRTEM image of the Pt nanocrystal with a lateral width of ∼18 nm.
made of thermally actuated bimetallic strips fixed on a TEM Cu-ring grid using epoxy resin. The bimetallic strips have the length of ∼2.2 mm, and the distance between the two bimetallic strips is ∼20 μm. As shown in Figure 1b, under gentle heating, outward deflection of each bimetallic strip can be achieved to a maximum of less than 2 μm (according to our in situ TEM measurement), which is only 0.091% of the bimetallic strips’ length. Thus, this allows the realization of an approximate uniaxial tensile test of the Pt thin film (see more details in Figure S1). Figure 1c shows a [1̅10] HRTEM image of a single crystal Pt with a lateral width of ∼18 nm, in which a couple of dislocations can be seen. As indicated in Figure 1c, the tensile loading is along the ±[001] directions. During the vertical tensile loading, this area can be monitored in situ so that the deformation behavior can be directly witnessed at the atomic scale. Furthermore, from the HRETM images, the dislocation types and dislocation number can clearly identified, and thus we can directly calculate the dislocation density changes during the plastic deformation (see low magnified TEM image of the nanocrystals and the dislocations calculation details in Figure S2). Figure 2 is a series of in situ HRTEM images captured from the area shown in Figure 1c and shows the tensile process of the Pt nanocrystal. To track the deformation, we marked two atomic columns as the references; thus, the applied strain can B
DOI: 10.1021/acs.nanolett.7b01416 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 3. (a−c) In situ atomic-scale observation of full dislocation nucleation and escape processes at the early deformation stage (strain below ∼7%). (d−e) Enlarged HRTEM image taken from the red framed region of panels a and b shows the atomic-scale structure of dislocations “1−3” more clearly.
Figure 4. (a, b) In situ observation of a full dislocation nucleation. (b, c) With further straining, in situ observation of an LD locks formation. (c, d) The formed LD locks unlocked with further straining. (e−g) Enlarged HRTEM corresponds to the red framed region of panels b− d, respectively.
1 1 1 [01 1̅ ] + [101] → [110] 2 2 2 1 1 1 [10 1̅ ] + [011] → [110] 2 2 2
dislocation “3” escaped by increasing the tensile strain (Figure 3c). No dislocation interaction and dislocation locks can be detected at this early deformation stage. Figure 3d−e is enlarged HRTEM images taken from the red framed region of Figure 3a and b to show the detailed atomic structures of these dislocation cores. By employing the Burgers circuits around these dislocation cores,46−48 their Burgers vectors (b) can be determined with an extra half-plane on {111} planes, where configurations represent typical 60° full dislocations with b = (1/2)⟨110⟩. Our extensive HRTEM investigations indicate that, in the early stages of plastic deformation, 60° full dislocations are commonly introduced. When the tensile strain reaches ∼10% and the dislocation density increased, these 60° full dislocations glided on their glide planes and had a chance to react with each other to form LD locks.46,48 Figure 4a−d shows a typical example of the formation of an LD lock and its destruction process, and Figure 4e−g is respectively enlarged HRTEM images corresponding to the red framed region of Figure 4b−d to determine their Burgers vectors. At first, no dislocation was detected in the viewing area, as shown in Figure 4a. By increasing the strain, Figure 4b shows the nucleation of a full dislocation, and Figure 4e shows the detailed analysis of the dislocation core and confirms that it is a 60° full dislocation with b = (1/2)[011̅] or b = (1/2)[101̅]. With further straining, Figure 4c shows a new type of dislocation structure. Figure 4f is the corresponding enlarged HRTEM image from which the Burgers circuit shows two extra half planes on two crossing (111)̅ and the (111) planes, representing an LD lock with b = (1/2)[110] (also see another example in Figure S3).48,49 The LD lock was formed by the interaction of two 60° full dislocations with Burgers vectors b = (1/2)[011̅] and b = (1/2) [101] or b = (1/2)[101̅] and b = (1/2)[011], moving under applied stress on two intersecting glide planes of (111) and (111)̅ .49,50 The dislocation reaction can be expressed as
or (1)
The two 60° dislocations mutually knit with each other to reach a low energy configuration.51 The b2 criterion for dislocation energy per unit length indicates a considerable energy reduction after the reaction. With further loading, the comparison of Figure 4c and d reveals the destruction of the LD lock at the atomic scale, in which Figure 4d and its enlarged HRTEM image (Figure 4g) show the dislocation with a Burgers vector of b = (1/2)[011]̅ or b = (1/2)[101̅], which is identical to that shown in Figure 4b. This indicates that the LD lock was unlocked under high stress with the reaction of: 1 1 1 [110] → [01 1̅ ] + [101] 2 2 2 1 1 1 [110] → [10 1̅ ] + [011] 2 2 2
or (2)
When the plastic strain reached ∼15%, the partial dislocations associated with SFs were commonly observed. Also, extended dislocations were found. Figure 5 shows an example of the formation of an extended dislocation and its combination. Figure 5a shows a low strain case, and no dislocation was found in the view area. When the plastic strain reached ∼15%, Figure 5b shows a full dislocation “1” and a partial dislocation associated with an SF. Local Burgers circuit shows the partial dislocation being 90° with b = (1/6)[112̅]. With continued straining, Figure 5c shows a 30° partial dislocation with b = (1/ 6)[12̅ 1]̅ or b = (1/6)[211̅ ]̅ gliding into the viewing area. Since both partial dislocations are on the same glide plane and associated with the same SF, the entire dislocation configuration is an extended dislocation with the Burgers vector of b = (1/2)[011]̅ or b = (1/2)[101]̅ (note with “2”). With further C
DOI: 10.1021/acs.nanolett.7b01416 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
partials from two intersections with extended 60° dislocations “1” and “2” on two intersecting glide planes of (111) and (111̅), respectively (see another example of L−C locks formation in Figure S6). The detailed description of the L−C lock can be found in ref 52. In Figure 6b, the extended dislocation “2” becomes a full 60° dislocation by reducing the distance between their two partial dislocations with continued straining, and the distance between two partial dislocations of dislocation “1” also decreasing. Under further straining, the extended dislocation “1” also changed into a full 60° dislocation with the same manner as the extended dislocation “2,” leading to the original L−C lock changed into an LD lock, and the LD lock is unlocked (see Figure 6c) and disappeared (see Figure 6d) with extensive loading. The LD destruction process is the same as that observed in Figure 4. It should be noted that this is the first experimental evidence of unlocking an L−C lock by transferring the L−C locks into LD locks first, then destroying the LD locks under high stress. Atomistic and dislocation dynamics simulations have shown that the LD and L−C locks have practically the same strength, and these stable junctions are dislocation obstacles and therefore a source of strain hardening.50,51 To clarify the plastic deformation mechanism of the ultrasmall-sized Pt nanocrystals, Figure 7a presents statistical data of the dislocation density as a function of the strain level (see the dislocation calculation details in Figure S2). It clearly shows that the plastic deformation can fit into two regions. At the low strain region, the dislocation nucleation rate is obviously faster than the dislocation escaping rate, leading to an increased density of dislocations from ∼5.0 × 1015 m−2 to ∼4.4 × 1016 m−2. After that, the dislocation density maintains almost constant at the high strain region. Figure 7b presents statistical data that shows the proportion of 60° full dislocations, partial dislocations, and dislocation locks (include LD and L−C locks) under different strain levels. At the early stage, the plastic deformation was controlled by full dislocation activities. After the strain reached ∼10%, a clear transition in deformation modes occurred, where the plastic deformation was switched from fully controlled by 60° full dislocations to plasticity governed by both 60° full and LD, then ultimately to the coexistence of 60° full and 90° partial dislocations, as well as LD and L−C locks. From Figure 7b, we can conclude that, although three types of dislocations were observed, the proportion of 60° full dislocations has been always above ∼67%, indicated that 60° full dislocations dominated the plastic deformation in Pt nanocrystals. Different from the previous studies and dislocation starvation theories which showed dislocations escape quicker through free surfaces than they multiply in small-sized face-centered cubic (FCC) metals (leaving the metals free of dislocations for those (sub) micrometer-sized metals during the plastic deformation),2,3,8,9,12−14,16,18,20,28 our in situ atomic-scale evidence in ∼18 nm sized Pt nanocrystal shows different plastic deformation mechanisms. First, the plastic deformation underwent the dislocation “appreciation” stage, in which the dislocation nucleation rate is faster than their escape rate, leading to an increased density of dislocations. After the dislocation density reached a certain value, the dislocation escape rate became equivalent to their nucleation rate and approached the dislocation “saturation”. Second, 60° full dislocations are the dominant deformation mechanism in the ∼18 nm sized Pt nanocrystal. This finding is different from the previous studies, in which the plastic deformation mechanisms
Figure 5. In situ observation of SFs and extended dislocation that resulted from the leading partial and trailing partial dislocation emission after the plastic strain surpassed ∼15%. (a, b) In situ observation of a 90° leading partial dislocation. (b) With continued straining, a 30° trailing partial dislocation nucleating behind the previous 90° partial dislocation, developing an extended dislocation. (d) With further loading, the trailing partial runs faster than the leading partial, eliminates the stacking fault between the extended dislocations, and leads to the dislocation change into a 60° full dislocation.
loading, Figure 5d illustrates the 30° partial dislocation gliding faster than the 90° partial dislocation, eventually eliminating the SF and leading to a 60° full dislocation (marked as dislocation “2”). At the same time, a new full dislocation “3” nucleated. During the straining, such extended dislocations were frequently observed (see other examples in Figures S4 and S5). Beside the partial dislocations resulted in extended dislocations, the L−C lock formation and destruction process can be also witnessed. Figure 6a shows a typical example of an L−C lock consisting of four SF segments pinned by the L−C lock core. The L−C lock was formed by the reaction of two
Figure 6. In situ observation of the Lomer−Cottrell (L−C) locks destruction process. (a) A typical example of an L−C lock formed by the reaction of two extended 60° dislocations “1” and “2.” (b) The extended 60° dislocation “2” becomes a full 60° dislocation by decreasing the distance between the leading and trailing partial dislocations, and splitting distance of dislocation “1” decreases. With continued straining, the extended dislocation “1” also changes into a full 60° dislocation, leading to the original L−C lock changed into an LD lock, and then the LD lock unlocked (c) and disappeared (d). D
DOI: 10.1021/acs.nanolett.7b01416 Nano Lett. XXXX, XXX, XXX−XXX
Letter
Nano Letters
Figure 7. (a) Statistical data of the dislocation density versus axial strain. At first, the dislocation density increases as the strain increases. (b) Statistical data show the proportion of 60° full dislocations, partial dislocations, and dislocation locks (include LD and L−C locks) under different strains.
that contains large fraction of surface atoms, there exists a high surface stress concentration on the surface of these small sized nanocrystals. According to the previous theoretical calculation and experimental results,63,64 there exists high surface strain in these small sized nanocrystals, and the local strain concentration can even approach ∼15%, which may assist the fast dislocation nucleation easily. At the same time, the surface strain may create stress gradient from the surfaces to the centers of the Pt nanocrystals, and the corresponding geometry necessary dislocations may resist to move. Furthermore, our statistical results show that the strain level has an obvious impact on the dislocation types, in which the plasticity was initiated by 60° full dislocations; then LD locks and partial dislocations can only be observed at relatively larger strains. According to previous theoretical predictions,65,66 the full dislocation is initiated by the emission of a leading partial dislocation, then followed by a trailing partial dislocation. After the full dislocation nucleated, the splitting distance (r) between the leading and the trailing partial dislocations depends on the local resolved shear stress (σ) and the SF energy (γ) of the metals according to the below formula:65
in nanosized single crystalline FCC metals were controlled by partial dislocation activities. For examples, Seo et al.21,22 revealed that coherent twin propagation resulting from partial dislocations can lead to the superplasticity in Au, Pd, and AuPd nanowires, and Yue et al.25 revealed that partial dislocations mediated plasticity when the single crystal Cu nanowire size was < ∼150 nm. Furthermore, we uncovered that the magnitude of strain has an impact on the dislocation activities in the Pt nanocrystals, leading to alternations of dislocation types for different strain levels, as well as the L−C and LD locks formations which possibly may lead to work hardening in the nanocrystals. It was further uncovered the unlock of L−C locks by transferring into LD locks first, and then LDs destructed under high stress. These phenomenons are obvious different from the dislocation activities in nanocrystalline metals, in which the grain boundary types and structures can significantly affect the dislocation characters which can be quite distinctive with their single crystalline form.53−56 The dislocation nucleation rate is faster than its escape rate in Pt nanocrystals may arise from the following factors. First, the Peierls−Nabarro stress (lattice resistance to dislocation motion) for Pt is about ∼10−4G (G is the shear modulus),57 which is about one order higher than those in Cu and Au.58 Thus, the Peierls−Nabarro stress will be much higher in Pt than those in Cu and Au to drive dislocations to move. The low Peierls−Nabarro stress in Cu and Au will facility the fast motion and escaping of the dislocations; thus, dislocation starvation may easily to be observed. Second, most previous studies were carried out with metals having low SF energies,22−26,29−34,59 in which the plastic deformation was governed by partial dislocation or twinning; while the SF energy curve of Pt is very different from other metals,60,61 this will significantly impact its dislocation behaviors. Accordingly, we anticipate that, for those metals with similar SF energy curves as Pt (such as Al), their plastic deformation should be governed by full dislocation activities in large size crystals. Third, the dislocations on different slip systems were activated, and the dislocations locks were easily formed, which can impede the dislocation motion. Fourth, the nanocrystal does not have a uniform cross-section compared to nanowire and nanopillars. The stress concentration would cause stress/strain gradient, which is the source of geometrical necessary dislocations that can stay more easily during plastic deformation.62 Finally, for the ultrasmall sized nanocrystal
r=
r0(γ ) k1b2 /γ = 1 − σ / σ (γ ) 1 − σ /k 2γ
(3)
where r0(γ) = k1b2/γ is the equilibrium splitting distance at σ = 0; and σ(γ) = k2γ is the resolved shear stress at which the splitting distance becomes infinitely large. The constants k1 and k2 depend on the elastic moduli of the material and the particular types of the two Shockley partials.46,65 For Pt with very high γ, full dislocations have been always observed at the early stage of plastic deformation, because σ is not very high in the low dislocation density nanocrystals (as seen in Figure 2a, 2b). While at the later stage of deformation, the strong strain hardening and the high density of dislocations stored in the nanocrystal make σ very high, which lead to the increased r. Thus, we can see extended dislocation and even single partial dislocations with SFs in the nanocrystals. In conclusion, the atomic-scale tensile processes of Pt nanocrystals with a size of ∼18 nm were captured in situ. The plastic deformation commenced with the dislocation appreciation first, followed by the dislocation “saturation”. The dislocation type changed under different strain levels, indicating that the strain level plays a key role in the dislocation types, the E
DOI: 10.1021/acs.nanolett.7b01416 Nano Lett. XXXX, XXX, XXX−XXX
Nano Letters
■
dislocation lock formation, and destruction. It is uncovered that L−C locks unlocked by transferring the L−C locks into LD locks first, followed by unlocking LD under high stresses. These results add fundamental insight of the deformation mechanisms in small nanosized metals, particularly the ultrathin metallic nanocrystals with high SF energies, and will also provide guidance for designing and manufacturing nanodevices. Methods. In Situ Atomic-Scale Deformation. The in situ tensile experiment was conducted with a specially designed double-tilt deformation stage operated inside a JEOL-2010F TEM (200 kV) at a strain rate of ∼10−3 s−1. To prepare the nanocrystal sample, a Pt crystalline thin film with ∼30 nm in thickness was deposited by magnetron sputtering on a (001)oriented NaCl single crystal substrate (3 × 3 cm2) at 300 °C. Under an optical microscope, the Pt thin film-deposited NaCl substrate was attached on a bimetallic extensor by using epoxy resin. After etching away the NaCl substrate with water, the Pt thin film became free-standing but remained on the bimetallic extensor. The Pt nanocrystalline thin film on the bimetallic extensor was further thinned using Fischione 1040 Nanomill, from which single crystalline Pt nanocrystals (e.g., Figure 1c) can be obtained with a thickness of ∼20 nm. The bimetallic extensor was loaded on a conventional double-tilt TEM heat holder. During in situ deformation, the temperature controller was accurately increased (
